Student Projects

Humanoid robots are attracting significant interest for their potential to transform manufacturing and home services. At the same time, advances in learning from demonstrations have enabled human-like feats of dexterity and agility in robotics [1-8]. However, humanoids remain limited to relatively structured tasks and lack the robustness and speed needed for unstructured environments. Building on our prior success in enabling quadruped robots to climb ladders [9], this project aims to extend ladder-climbing capabilities to humanoid robots using modern learning from demonstration techniques to facilitate training and develop natural motions. Unlike previous attempts at humanoid ladder climbing, which were slow and sensitive to perturbations [10, 11], our goal is to develop a more agile and reliable system—unlocking critical real-world applications such as disaster response and industrial inspection.

[1] Peng, X.B., Abbeel, P., Levine, S. and Van de Panne, M., 2018. Deepmimic: Example-guided deep reinforcement learning of physics-based character skills. ACM Transactions On Graphics (TOG), 37(4), pp.1-14.

[2] Peng, X.B., Ma, Z., Abbeel, P., Levine, S. and Kanazawa, A., 2021. Amp: Adversarial motion priors for stylized physics-based character control. ACM Transactions on Graphics (ToG), 40(4), pp.1-20.

[3] Li, C., Vlastelica, M., Blaes, S., Frey, J., Grimminger, F. and Martius, G., 2023, March. Learning agile skills via adversarial imitation of rough partial demonstrations. In Conference on Robot Learning (pp. 342-352). PMLR.

[4] Li, C., Blaes, S., Kolev, P., Vlastelica, M., Frey, J. and Martius, G., 2023, May. Versatile skill control via self-supervised adversarial imitation of unlabeled mixed motions. In 2023 IEEE international conference on robotics and automation (ICRA) (pp. 2944-2950). IEEE.

[5] Peng, X.B., Guo, Y., Halper, L., Levine, S. and Fidler, S., 2022. Ase: Large-scale reusable adversarial skill embeddings for physically simulated characters. ACM Transactions On Graphics (TOG), 41(4), pp.1-17.

[6] Tessler, C., Kasten, Y., Guo, Y., Mannor, S., Chechik, G. and Peng, X.B., 2023, July. Calm: Conditional adversarial latent models for directable virtual characters. In ACM SIGGRAPH 2023 Conference Proceedings (pp. 1-9).

[7] Li, C., Stanger-Jones, E., Heim, S. and Kim, S., 2024. Fld: Fourier latent dynamics for structured motion representation and learning. arXiv preprint arXiv:2402.13820.

[8] Watanabe, R., Li, C. and Hutter, M., 2025. DFM: Deep Fourier Mimic for Expressive Dance Motion Learning. arXiv preprint arXiv:2502.10980.

[9] D. Vogel, R. Baines, J. Church, J. Lotzer, K. Werner, and M. Hutter. Robust ladder climbing with a quadrupedal robot. 2025 IEEE International Conference on Intelligent Robots and Systems (accepted; preprint available on arXiv)

[10] J.Vaillant, et al., “Multi-contact vertical ladder climbing with an HRP-2 humanoid,” Autonomous Robots, vol. 40, no. 3, pp. 561–580, Mar. 2016.

[11] H.Yoneda, et al.,“Vertical ladder climbing motion with posture control for multi-locomotion robot,” in 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems. Nice: IEEE, Sep. 2008, pp. 3579–3584.

• Literature research
• Design the task simulation environment
• Develop autonomous control policies for ladder climbing
• Deploy the control policy on real hardware (hardware already provided)

• Excellent knowledge of physical simulation
• Deep knowledge of coding in Python and C++
• Ability to work independently and creatively
• Experience with robot sim2real preferred
• Experience with learning from demonstration algorithms

Robert Baines (rbaines@ethz.ch) Alex Hansson (ahansson@mavt.ethz.ch) Chenhao Li (chenhao.li@ai.ethz.ch)

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

Background

Visual SLAM (Simultaneous Localization and Mapping) enables robots to navigate unknown environments by building 3D maps using visual data from onboard cameras. While these maps are optimized for robotic perception and localization, they often lack the structural clarity and semantic richness of CAD models that humans rely on for spatial understanding. Additionally, they often differ in scale, completeness, and accuracy due to sensor noise and perception limitation. These mismatches create a communication gap between autonomous systems and human operators.

Goals

To bridge this gap, this thesis explores the problem of registering maps generated by SLAM with CAD floor plans. By establishing the projection between the robot-centric and human-centric representations, the approach will improve human-robot interaction, environment interpretation, and integration of autonomous mobile robots into existing workflows that rely on structured 2D CAD maps. Additionally, the thesis will explore how such alignment can support refining SLAM maps using CAD priors.

Proposed Method

The alignment will be formulated as a non-linear projection problem between the two map representations [1], while preserving local and global consistency of given correspondences. The problem can be solved using optimization techniques. As an optional continuation, the accurate geometric and semantic information in the CAD structure can be utilized to correct the distortion in the SLAM map. The approach will be evaluated on a real robot. The exact method remains open to refinement and innovation. It’s encouraged to explore alternative methods.

Expected outcome

● A novel spatial registration algorithm to align SLAM-generated maps with architectural CAD models

● A quantitative and qualitative analysis of the proposed method.

● Strong self-motivation and curiosity for solving challenging robotic problems

● Previous experience in the fields of optimization, computer vision and VSLAM

● Excellent programming skills in C++ and Python

● Experience with Linux, ROS, and typical development tools such as git are advantageous

● An excellent academic record is desirable, but may be compensated by expert knowledge in the areas mentioned above

If you are interested, please send your transcripts and CV to Zimeng Jiang (zimeng.jiang@sevensense.ch) and Roxane Merat (roxane.merat@sevensense.ch).

The goal of this project is to develop a lidar-visual-inertial odometry approach that integrates visual and lidar measurements into a single unified representation. The starting point, inspired by FAST-LIVO2 [1], will be to investigate methods for efficiently combining visual patches from camera images with a set of geometric primitives extracted from FAST-LIO2 [4], a lidar-inertial odometry pipeline. The performance of the resulting approach will be evaluated in comparison with existing lidar-visual-inertial odometry approaches.

References: [1] C. Zheng et al., “FAST-LIVO2: Fast, Direct LiDAR-Inertial-Visual Odometry,” IEEE Transactions on Robotics, 2024. [2] J. Lin and F. Zhang, “R3LIVE++: A Robust, Real-time, Radiance reconstruction package with a tightly-coupled LiDAR-Inertial-Visual state Estimator,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. [3] T. Shan, B. Englot, C. Ratti, and D. Rus, “LVI-SAM: Tightly-coupled Lidar-Visual-Inertial Odometry via Smoothing and Mapping,” in IEEE International Conference on Robotics and Automation, 2021. [4] W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang, “FAST-LIO2: Fast Direct LiDAR-inertial Odometry,” arXiv, 2021.

• WP1: Literature review of work on lidar-visual-inertial odometry.
• WP2: Develop a lidar-visual-inertial odometry approach with a single unified representation.
• WP3: Evaluate the performance of the approach in comparison with existing work.

Experience with C++ and ROS.

Please send CV and transcripts to Rowan Border (border.rowan@ucy.ac.cy) and Ruben Mascaro (rmascaro@ethz.ch).

The objective of the thesis is to develop an algorithm capable of real-time collision avoidance for drones, ensuring safe operation even in environments with other air traffic, such as airplanes, helicopters, paragliders, and similar obstacles. The tasks include selecting and evaluating existing neural networks and algorithms, or designing a custom solution tailored to our requirements. The project also involves working with a high-performance hardware setup, with support for real-time execution on an NVIDIA Jetson Orin.

Avientus provides a flexible and dynamic research environment with a budget for necessary hardware, such as cameras and sensors. Students are encouraged to experiment with and choose tools, hardware, and software frameworks that best suit their approach. This thesis offers a unique opportunity to work on cutting-edge technology in a startup environment, contributing directly to the future of automated drone logistics.

If you are passionate about robotics, computer vision, and real-time systems, and are excited to tackle challenges in the aviation domain, we’d love to hear from you!

• Strong competencies in computer vision & neural networks
• Good knowledge in programming (Python and / or C++)
• General interest in VTOL drones

Please send your CV and transcripts to info@avientus.ch and rmascaro@ethz.ch

Neural motion planners [1, 2] distill expert trajectories from classical motion planners [3, 4] into fast, reactive, and generalizable policies for robotic arms. However, existing neural planners do not account for the presence of in-hand objects, leading to significant limitations including (a) the influence of additional gravitational forces and (b) increased complexity in collision geometry. This project extends such methods to handle novel objects and tools with unknown geometry. Using a camera mounted on the robot's end-effector, the system first performs 3D reconstruction of the target object [5]. The resulting shape representation is fused into the state space of a reinforcement learning policy trained to reach and manipulate objects while avoiding self-collisions and obstacles in the environment. Unlike prior work [6], which uses conservative bounding box representations, our method leverages precise reconstructed geometry for fine-grained, tool- and object-aware collision avoidance. The goal is to learn a single policy that generalizes across different tools and objects, adapting its behavior based on their geometry. The policy will be validated both in simulation and on hardware, first with a stationary arm and then on a mobile manipulator combining a legged base and a robotic arm.

[1] A. Fishman, A. Murali, C. Eppner, B. Peele, B. Boots, and D. Fox, “Motion policy networks,” in Conference on Robot Learning. PMLR, 2023, pp. 967–977 [2] Murtaza Dalal, et al. “Neural MP: A Generalist Neural Motion Planner”, 2024 [3] Ratliff N, et al.. “CHOMP: gradient optimization techniques for efficient motion planning”, 2009, IEEE International Conference on Robotics and Automation [4] M.P. Strub et al., “Adaptively informed trees (ait*): Fast asymptotically optimal path planning through adaptive heuristics,” in 2020 IEEE ICRA [5] Hong, Y. et al.. See-Then-Grasp: Object Full 3D Reconstruction via Two-Stage Active Robotic Reconstruction Using Single Manipulator. Appl. Sci. 2025 [6] J. Lee et al., "Learning Fast, Tool-Aware Collision Avoidance for Collaborative Robots," in IEEE Robotics and Automation Letters, vol. 10, no. 8, pp. 7731-7738, Aug. 2025

1.Literature review on visual object reconstruction and shape representations (e.g. NeRF, Gaussian Splatting, SDFs). 2.Train an RL policy that uses both robot state and reconstructed object geometry to perform collision-aware manipulation. 3.Evaluate the policy on unseen tools and objects, testing generalization and robustness. 4.Deploy the pipeline on a physical arm and mobile manipulator.

1.Good knowledge of Python 2.Knowledge of Reinforcement Learning and Isaac Sim (Prefered) 3.Experience with robotic hardware
4.Understanding of object reconstruction methods

Sport monitoring has many benefits including injury prevention and performance optimization. Current methods for activity monitoring in sports mostly use camera-based motion tracking or the use of inertial measurement unit-based systems that are limited in their measurement space or are obtrusive to the activities of the user. Smart clothing offers a solution that makes monitoring of biomechanics during individual and team sports possible in different conditions.

We have made textile-based wearable technology for unobtrusive monitoring of movement and completed a study in which we acquired real-world data from our technology during sport activities. Using machine learning methods, this project aims to extract useful information from our data such as activity type and fatigue. This project will offer valuable experience in data processing, machine learning, working with real-world data, and cutting-edge wearable technologies.

Your Profile

• Background in electrical engineering, computer engineering, mechanical engineering, or related fields

• Prior experience with data analysis (signal processing, machine learning algorithms, or similar) and machine learning

• Independent worker with critical thinking skills and problem solving skills

• Investigate methods to classify different sport-specific activities or predict biomechanical measures based on real-world data.
• Develop and test models to assess the level of fatigue during sports.
• Improve the model and compare it with other measurement methods.

Prof Dr Carlo Menon and Dr Chakaveh Ahmadizadeh will supervise the student and the research will be performed at the Biomedical and Mobile Health Technology lab (www.bmht.ethz.ch) at ETH Zurich, Switzerland.

To apply, use the button below to tell us why you want to do this project ("motivation") along with the type of project you are applying for (e.g., master project or thesis) and your timeline for the project; attach a mini CV with your current program of study, your grades and any other info you deem relevant--maybe the name and phone number of a postdoc or a professor willing to be your reference; and make any further comments ("additional remarks").

Movement monitoring is a rapidly growing field with many applications in healthcare and fitness. Current methods for activity monitoring mostly use camera-based motion tracking or the inertial measurement unit-based systems that are limited in their measurement space or are obtrusive to the activities of the user. Smart clothing offers a solution that makes monitoring of biomechanics during individual and team sports possible in different conditions.

We develop textile-based wearable technologies for unobtrusive monitoring of movement based on capacitive, resistive, or inductive sensors. Signals from these sensors are often acquired using benchtop LCR meters. However, it is crucial for our wearable applications to replace such solutions with portable and unobtrusive data acquisition methods. This project aims to design, develop, and test portable data acquisition systems for data recording from multiple capacitive sensors. This project will offer valuable experience in electronics, prototyping, embedded software, and cutting-edge wearable technologies.

Your Profile

• Background in electrical engineering, computer engineering, mechanical engineering, or related fields

• Prior experience with electronics and prototyping (printed circuit board, electronic circuit testing and troubleshooting, or similar)

• Independent worker with critical thinking skills and problem solving skills

• Identify areas for improvement in current data acquisition solutions.
• Develop and test electronics and software for signal recording from capacitive sensors.
• Implement developed solutions into textile-based wearable technologies for movement monitoring.

Prof Dr Carlo Menon and Dr Chakaveh Ahmadizadeh will supervise the student and the research will be performed at the Biomedical and Mobile Health Technology lab (www.bmht.ethz.ch) at ETH Zurich, Switzerland.

To apply, use the button below to tell us why you want to do this project ("motivation") along with the type of project you are applying for (e.g., master project or thesis) and your timeline for the project; attach a mini CV with your current program of study, your grades and any other info you deem relevant--maybe the name and phone number of a postdoc or a professor willing to be your reference; and make any further comments ("additional remarks").

In this project, we will characterize and evaluate the use of piezoelectric sensors to monitor blood pressure using AI approaches. Data will be acquired using two sensors placed along a major artery, which will detect the pulse wave. From this, key physiological features will be extracted such as arterial wall displacement, diameter variation, and blood flow velocity, and pulse transit time which may be used in computational models of blood pressure estimation (e.g., Bramwell–Hill model). Machine learning and deep learning approaches will be used for blood pressure estimation and compared to established computational models. This project will offer valuable experience in data processing, machine learning, and working with real-world health data.

Goals • Implement physiological feature extraction algorithms • Implement mathematical, ML and DL models for blood pressure estimation

Tasks • Literature review (20%) • Data processing and key physiological feature extraction (20%) • Implement and calibrate mathematical models for blood pressure estimation (10%) • Implement ML and DL models for blood pressure estimation (30%) • Test and evaluation of models and comparison with gold standard (10%) • Report and presentation (10%)

Prof Dr Carlo Menon, Dr. Corin Otesteanu (corin.otesteanu@hest.ethz.ch) will supervise the student and the research will be performed at the Biomedical and Mobile Health Technology lab (www.bmht.ethz.ch) at ETH Zurich, Switzerland.

To apply, use the button below to tell us why you want to do this project ("motivation") along with the type of project you are applying for (e.g., master project or thesis) and your timeline for the project; attach a mini CV with your current program of study, your grades and any other info you deem relevant--maybe the name and phone number of a postdoc or a professor willing to be your reference; and make any further comments ("additional remarks").

Wearable triboelectric nanogenerators (TENGs) are emerging as energy-harvesting and self-powered sensors that can transduce biomechanical motion into electrical signals. In this project, we will characterize and evaluate TENGs for real-time joint motion estimation and fatigue analysis. Controlled experiments will be conducted during repetitive tasks using an optical motion capture system or inertial measurement unit (IMU) as gold standard. The dataset will consist of TENG voltage outputs synchronized with 3D kinematic data from the reference systems. This project will offer valuable experience in data processing, machine learning, and working with real-world health data.

Goals • Characterize the TENG sensor signal for different joint movements • Implement ML and DL models for joint angle reconstruction • Benchmarking shallow, deep, and transformers based models for performance

Tasks • Literature review (10%) • Characterize the TENG sensor signal in controlled settings (10%) • Implement ML and DL models for angle and angular velocity reconstruction (20%) • Data collection from 10-15 participants during a standardized protocol (30%) • Test and evaluation of models and comparison with gold standard (20%) • Report and presentation (10%)

Prof Dr Carlo Menon, Dr. Corin Otesteanu (corin.otesteanu@hest.ethz.ch) and Dr. Satyiaranjan Bairagi (satyaranjan.bairagi@hest.ethz.ch) will jointly supervise the student and the research will be performed at the Biomedical and Mobile Health Technology lab (www.bmht.ethz.ch) at ETH Zurich, Switzerland.

To apply, use the button below to tell us why you want to do this project ("motivation") along with the type of project you are applying for (e.g., master project or thesis) and your timeline for the project; attach a mini CV with your current program of study, your grades and any other info you deem relevant--maybe the name and phone number of a postdoc or a professor willing to be your reference; and make any further comments ("additional remarks").

The project aims to evaluate the effectiveness of wearable technology in identifying anxiety in individuals. This research makes use of publicly available datasets of individuals watching anxiety inducing videos . Data gathered from these devices include continuous measurements over 1 hour of respiration, ECG and electrodermal activity data. By developing machine learning methods, the aim is to continuously detect anxiety levels in an individualized manner. This project will offer valuable experience in data processing, machine learning, and working with real-world health data. Moreover, it has the potential to meaningfully impact the users wellbeing.

Goals • Quantitatively analyze the effect of anxiety inducing videos on several biomarkers extracted from wearable data • Investigate different machine learning models to classify anxiety levels (low, normal, high) from wearable device data. Tasks • Literature review (10%) • Data analysis (loading data, data filtering) (30%) • Design and implement shallow machine learning model for anxiety level classification (20%) • Design and implement a recurrent neural network for anxiety level classification (20%) • Test and evaluation of models (10%) • Report and presentation (10%)

Prof Dr Carlo Menon and Dr. Corin Otesteanu (corin.otesteanu@hest.ethz.ch) will supervise the student and the research will be performed at the Biomedical and Mobile Health Technology lab (www.bmht.ethz.ch) at ETH Zurich, Switzerland.

To apply, use the button below to tell us why you want to do this project ("motivation") along with the type of project you are applying for (e.g., master project or thesis) and your timeline for the project; attach a mini CV with your current program of study, your grades and any other info you deem relevant--maybe the name and phone number of a postdoc or a professor willing to be your reference; and make any further comments ("additional remarks")

6 DoF pose estimation of a vessel model using 2D X-ray image

Derick Sivakumaran, sderick@ethz.ch

Volumetric bioprinting is a powerful technology for tissue engineering that enables the fabrication of cell-laden constructs with exceptional speed, resolution, and geometrical freedom.

We offer student projects that can be tailored to fit your specific interests, ranging from 3D bioprinting to the development and characterization of testing platforms for engineered muscles. These research projects aim to combine expertise in biomaterials, 3D printing, and cell culture techniques with the final goal of generating functional living actuators.

Possible project goals:

• Optimize xolographic printing with living and synthetic materials.
• Design and print different bioactuator geometries.
• Culture and characterize engineered muscle constructs.
• Optimize/develop platforms to assess the bioactuator's performance.

abadolato@ethz.ch

Wearable biosensing platforms are increasingly explored for continuous, noninvasive health monitoring. This project focuses on developing a wireless sensor system based on resonant metamaterials combined with soft, ion-selective interfaces that can extract biomarkers from skin-interfaced fluids. By leveraging changes in electromagnetic properties, the sensor can detect specific physiological signals without requiring batteries or complex electronics. Students will investigate flexible materials and resonator designs, gaining interdisciplinary experience across electronics, materials science, and biomedical sensing

• Experimental work on hydrogel synthesis and functionalization • Fabricate and integration on flexible substrate • Wireless measurement, calibration, and performance evaluation • Write a scientific project report

Prof Dr Carlo Menon, Dr. Muhammad Zada and Dr. Weifeng Yang will supervise the student, and the research will be performed at ETH Zurich’s Biomedical and Mobile Health Technology research group (www.bmht.ethz.ch) in the Balgrist Campus in Zurich, Switzerland.

In recent years, the advent of learning-based methods has led to interesting novel structure from motion modules that provide end-to-end 3D structure, camera pose, and camera intrinsics estimation from arbitrary image sets [1,2]. While it is reasonable to assume that these representation implicitly impose geometric priors onto regularly encountered scene elements (e.g. chairs, tables, etc.), the imposition of these priors gets lost during the typical global back-end optimization step [3,4].

The imposition of geometric priors in back-end optimization is a topic with long-standing history, and nowadays encompasses both classical techniques such as piece-wise planarity or Manhattan world assumptions, as well as modern learning-based representations such as neural shape priors [5]. While certainly interesting, these approaches suffer from two important limitations:

• Classical approaches are limited to a small set of pre-defined regularities that may be encountered in a certain class of environments (typically man-made). It is furthermore difficult to decide without further cues which parts of an environment should be regularized. • More modern approaches such as deep shape priors are typically trained on specific classes of objects, only, and are thus limited to certain object-level application.

The goal of the present thesis is to investigate the following topics:

• Flexible discovery of regularities (surface regularities, repetitive scene elements) in large scale outdoor reconstruction scenarios using LLM/VLMs • Extraction of geometric surface properties from grounded open-set features • Automatic formulation of regularization priors using LLMs • Inclusion into back-end reconstruction for improved global reconstruction results

The proposed thesis will be conducted at the Robotics and AI (RAI) Institute, a new top-notch partner institute of Boston Dynamics pushing the boundaries of control and perception in robotics. For an example of the recent achievements of the institute, please consider our online video channel: Stunting with Reinforcement Learning

The down-stream task of the present project will be the creation of geometric models with high fidelity surfaces for inclusion into simulation environments.

Selection will be highly selective. Potential candidates are invited to submit their CV and grade sheet, after which students will be invited to an on-site interview.

References [1] Grounding image matching in 3D with Mast3r, Vincent Leroy, Yohann Cabon, and Jerome Revaud [2] VGGT: Visual Geoemtry Grounded Transformer, Wang, Chen, Karaev, Vedaldi, Rupprecht, Novotny [3] MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priors, Murai, Dexheimer, Davison [4] VGGT-SLAM: Dense RGB SLAM Optimized on the SL(4) Manifold, Dominic Maggio, Hyungtae Lim, Luca Carlone [5] DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation, Park, Florence, Straub, Newcombe, Lovegrove

• Literature research • Familiarization with modern 3D reconstruction pipeline • Exploration of LLM/VLM for automatic geometric regularity detection • Implementation and inclusion into back-end optimization • Testing and Validation

• Excellent knowledge of Python and C++ • Knowledge in Computer vision • Experience in SLAM/reconstruction • Experience in applying learning-based representations • Interest in recent LLM/VLM architectures

lkneip@theaiinstitute.com aliniger@theaiinstitute.com

Most manipulation research focuses on using only the gripper mounted on an arm for interaction. However, robots, like humans, should be able to use different parts of their morphology (base, elbow, hips, feet) for interaction. A common challenge in learning these multi-modal behaviors is mode collapse due to limited expressivity in the policy and/or learning algorithm.

Recently, the conditional denoising diffusion process has shown powerful generative modeling capabilities for image (DALL-E) and behavior generation. The goal of this project is to develop a deeper understanding of these models and investigate their ability to learn interactions with multiple end-effectors on the robot [1]. Additionally, we want to improve upon the vanilla diffusion policy [2] and study the model's generalizability to novel situations.

References:

• Sleiman, Jean-Pierre, et al. "Versatile multi-contact planning and control for legged loco-manipulation." Science Robotics (2023).
• Chi, Cheng, et al. "Diffusion policy: Visuomotor policy learning via action diffusion." arXiv:2303.04137 (2023).

• Literature research on diffusion policies
• Designing simulation environment in Orbit and generating data
• Training and improving diffusion policy architecture for multi-modality in interaction data
• Performing ablations and comparisons to alternate methods

• Highly motivated and autonomous student
• Experience with machine learning and controls
• Excellent knowledge of Python and PyTorch
• Experience with simulators and robot hardware is a bonus

Please send your CV and transcript to the following:

• Mayank Mittal (mittalma@ethz.ch)
• Jean-Pierre Sleiman (jsleiman@ethz.ch)

Innovations using metamaterials have been made in recent years, offering promising approaches for creating stretchable and adaptable structures in wearable electronics. Innovative ways of integrating wearables into our daily life and maximizing the functionality of such structures require new fabrication methods that are both cost-effective and versatile. In the present project, a new approach utilizing 3D printed molds to create stretchable auxetic structures is being developed. Soft substrates will be patterned to form auxetic structures that can be integrated into textiles. Simultaneously, the sensing elements will be embedded into the soft substrate using an insert molding technique. The resulting sensors are designed for motion monitoring, making them suitable for integration into garments and other wearable platforms. The ability of these sensors to maintain performance under mechanical deformation makes them particularly advantageous for wearable health monitoring systems.

• Develop a method for the production of capacitive auxetic structures
• Fabricate functional wearable sensors
• Write a project report

Tasks

• Literature review (10%)
• Optimization of the fabrication of capacitive sensors (50%)
• Characterization of the produced sensors (30%)
• Data collection, reporting and presentation (10%)

Your profile

• Background in Applied Physics, Mechanical Engineering, Materials Science, Electronics Engineering, Biomedical Engineering, Health Science or related fields is desirable but not mandatory.
• Independent worker with critical thinking and problem-solving skills

Prof. Dr. Carlo Menon and Pierre Kateb will supervise the student and the research will be performed at ETH Zurich’s Biomedical and Mobile Health Technology research group (www.bmht.ethz.ch) in the Balgrist Campus in Zurich, Switzerland.

To apply, use the button below to tell us why you want to do this project ("motivation"); attach a CV with your current program of study, your grades and any other info you deem relevant.

Join us at Skaaltec to design and implement a robust CI/CD pipeline for our embedded software development. In this project, you will leverage GitHub workflows to automate building, testing, and deployment of firmware. The focus will be on improving software quality and speeding up development cycles in a high-reliability embedded environment.

You will be responsible for:

• Setting up automated code checks and unit tests (e.g., triggered on pull requests)
• Ensuring that the codebase builds without errors or warnings across configurations
• Integrating a testing framework for embedded unit testing
• Developing a secure firmware build and release pipeline, including signing and deployment steps

This is a great opportunity to work on modern software development infrastructure, applied to embedded systems with real-world constraints.

Requirements

• Familiarity with GitHub Actions and CI/CD pipelines
• Experience in embedded C programming, debugging, and cross-compilation
• Comfort working in a Unix-based development environment

Preferred Experience

• Zephyr RTOS and the ZTEST framework
• Basic knowledge of firmware signing and secure boot concepts

Electrical/Software Engineer, Computer Scientist or proved equivalent experience

aronco@ethz.ch

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This project targets performance and feature enhancements to a firmware module that manages data transfers from flash memory to a host PC over USB (CDC/ACM). The goal is to improve system reliability, throughput, and communication efficiency for data logging or diagnostic use cases.

You will work on:

• Implementing and evaluating a caching layer for the flash translation mechanism to reduce bus utilization
• Designing and developing a lightweight USB communication protocol with data transmission and acknowledgment features
• Enhancing the firmware state machine to optimize the flow of data from flash to USB, ensuring robust handling of edge cases

You will gain hands-on experience with low-level embedded development and system design, contributing to a real and evolving codebase used in a real medical device.

Requirements

• Solid experience with embedded C programming and debugging
• Understanding of state machine design and resource-constrained systems

Preferred Experience

• Some familiarity with flash memory interfaces and wear-leveling techniques
• Experience with Zephyr RTOS

Electrical/Software Engineer, Computer Scientist or proved equivalent experience

Documentation is a crucial part of any high-quality software project. In this internship, you will help improve the usability and maintainability of Skaaltec’s embedded firmware codebase by integrating automated documentation generation using Doxygen.

The goal is to set up a complete documentation pipeline, both locally and in our GitHub-based CI/CD workflows. You will:

• Configure Doxygen for the project, including layout, style, and output formats (e.g., HTML, LaTeX/PDF)
• Write and revise in-code documentation (docstrings) for key modules and APIs
• Ensure documentation is automatically generated and published as part of the CI/CD process
• Help define a structure that makes the documentation easy to navigate for new developers and collaborators

This project will help bring clarity and accessibility to a growing embedded codebase used in real-world applications, while giving you hands-on experience with tooling, automation, and technical writing.

Requirements

• Basic familiarity with Git and GitHub workflows
• Experience reading and writing embedded C code
• Interest in technical documentation and developer tooling

Preferred Experience

• Previous use of Doxygen or similar documentation generators
• Exposure to CI/CD tools like GitHub Actions
• Understanding of software architecture and code organization in embedded systems is a plus

Electrical/Software Engineer, Computer Scientist or proved equivalent experience

Background

Microrobots have immense potential in fields such as biomedicine and environmental remediation. However, their development has been hindered by limitations in integrating multiple functional components effectively. Current top-down fabrication methods, e.g. photolithography or 3D printing, struggle to combine diverse functional components, restricting the versatility and performance of microrobots.

To overcome these challenges, this project will develop a novel bottom-up microfluidic assembly method, enabling the creation of multifunctional microrobots with unprecedented precision and flexibility. This innovative approach has the potential to redefine microrobot fabrication and expand their applications significantly.

Ideal Skills and Experience (not mandatory):

• Experience or knowledge in microfluidic devices design and operation.

• Prior experience in chemistry lab.

References

M. Hu et al. "Shaping the assembly of superparamagnetic nanoparticles." Mater. Horiz. 9.6 (2022), 1641.

B. J. Nelson & S. Pané “Delivering drugs with microrobots.” Science 382.6675 (2023), 1120.

T. Moragues et al. “Droplet-Based Microfluidics”. Nature Reviews Methods Primers 3.1 (2023), 32.

The goal of this project is to develop a novel bottom-up microfluidic assembly method for creating multi-functional microrobots with enhanced precision and flexibility. This approach aims to overcome current limitations in integrating diverse functional components, paving the way for advanced applications in biomedicine and environmental remediation.

Our project is highly interdisciplinary and embodies a high-impact, high-reward research approach. Your work could lead to pioneering discoveries and applications in microrobotics. If you are interested, please contact Dr. Minghan Hu (minghu@ethz.ch) and Chao Song (chao.song@chem.ethz.ch) for more details about the Master thesis.

Hip osteoarthritis (OA) can cause pain and disability, from minor activity limitations to severe participation restrictions. Joint stiffness and reduced range of motion are among the most common functional impairments, which impact function and quality of life. A restricted hip internal rotation (often symptomatic) is one of the typical findings in the clinical examination of patients with hip OA. Morphology deformities (cam/pincer) of the hip joint, which have been found be common in young active individuals, represent potential risk factors for hip OA. Individuals diagnosed with femoroacetabular impingement syndrome (FAIS), as an example, young athletes with cam morphology involved in vigorous sport activities, may present with a limitation in hip internal rotation. Recently, it has been found in 244 asymptomatic young males that the standardised assessment (with an examination chair) of internal hip rotation did not allow the exclusion of a cam morphology, but might be used to rule one in. These preliminary results can support the use of a standardised assessment of internal hip rotation in screening of youth and adolescents involved in high-risk sports (i.e., ice hockey, gymnastics, martial arts) and in the documentation of hip range of motion in individuals with other hip disorders. The ETH SMS Lab has developed and pilot tested a new mobile device (called “mHIRex”) to examine hip internal rotation while also recording the torque needed for/produced by the hip against the movement (see Figure 1). This device was found to be accurate in the measurement of both hip motion and torque. As only a small number of subjects (24 healthy individuals, 14 elite athletes, and 4 patients with diagnosed FAIS) were involved in the feasibility study, no conclusions from the collected data could be drawn. The next step would be to assess hip internal rotation in a large cohort of consecutive patients with hip disorders (incl. OA, FAIS) at the Schulthess Clinic.

• Get familiar with the relevant literature, talk to clinicians
• Elaborate the measurement device
• Run study, analyse data, present results

• Exchange with very experienced clinicians
• Being involved in translational science

• Background in Human Movement Science
• Interested in clinical biomechanics

Interested? Please send your motivation, CV and latest transcript to Peter Wolf, pwolf@ethz.ch

Kirigami is the Japanese art of cutting paper, similar to origami, which is the art of folding paper. Innovations using kirigami-inspired designs and metamaterials have been made in recent years, offering promising approaches for creating stretchable and adaptable structures in wearable electronics. Innovative ways of integrating wearables into our daily life and maximizing the functionality of such structures require new fabrication methods that are both cost-effective and versatile. In the present project, a new approach utilizing laser-cutting and 3D printing for creating kirigami-inspired antennas on flexible substrates is being developed. Soft substrates will be patterned to form kirigami structures that can be integrated into textiles. Simultaneously, the antenna element will be embedded into the soft substrate using an insert molding technique. The resulting kirigami antennas are designed for applications such as wireless communication and motion monitoring, making them suitable for integration into garments and other wearable platforms. The ability of these antennas to maintain performance under mechanical deformation makes them particularly advantageous for wearable health monitoring systems.

• Develop a method for the production of kirigami
• Fabricate functional wearable antennas
• Write a project report

Tasks

• Literature review (10%)
• Optimization of the fabrication of kirigamis (50%)
• Characterization of the produced kirigami (30%)
• Data collection, reporting and presentation (10%)

Your profile

• Background in Applied Physics, Health Science, Mechanical Engineering, Materials Science, Electronics Engineering, Biomedical Engineering or related fields is desirable but not mandatory.
• Independent worker with critical thinking and problem-solving skills

Prof Dr Carlo Menon, Dr. Muhammad Zada and Pierre Kateb will supervise the student and the research will be performed at ETH Zurich’s Biomedical and Mobile Health Technology research group (www.bmht.ethz.ch) in the Balgrist Campus in Zurich, Switzerland.

To apply, use the button below to tell us why you want to do this project ("motivation"); attach a CV with your current program of study, your grades and any other info you deem relevant.

Prior work has successfully synthesized multi-terrain skills into a single controller using RL fine-tuning [1]. This project seeks to explore the ultimate scalability of this method: can a policy achieve universal competence through continued training? The primary objective is to incrementally fine-tune the policy on a growing collection of challenging real-world and simulated terrains. This process will systematically evaluate performance to determine if and when generalization capabilities begin to plateau. A further objective could be to explore procedural terrain generation. This would involve creating terrains that specifically target the policy's weaknesses, providing an efficient path to improved robustness. This project will provide key insights into the requirements for creating truly generalist robotic controllers.

References

• [1] Rudin, N., He, J., Aurand, J., & Hutter, M. (2025). Parkour in the Wild: Learning a General and Extensible Agile Locomotion Policy Using Multi-expert Distillation and RL Fine-tuning. arXiv preprint arXiv:2505.11164.

• Literature Research
• Implementation of a scalable finetuning environment
• Scientific evaluation of the scalability of RL finetuning

• Excellent knowledge of Python
• Experience with Reinforcement Learning

Please send your CV, transcript and a short motivation to:

cschwarke@ethz.ch junzhe@ethz.ch

Differentiable simulation [1] has demonstrated significant improvements in sample efficiency compared to traditional reinforcement learning approaches across various applications, including legged locomotion [2]. This project seeks to explore another key advantage of differentiable simulation: its capability for more precise optimization. The study will focus on a tracking task involving ALMA, a quadrupedal robot equipped with a robotic arm. The primary objectives are to develop a differentiable simulation environment for the robot and evaluate its advantages over traditional reinforcement learning methods. By utilizing the gradients provided by the simulation, control policies will be optimized to improve tracking performance. The work involves creating a tailored differentiable simulation, systematically comparing its performance with reinforcement learning techniques, and analyzing its impact on accuracy and real-world applicability. This project provides an opportunity to contribute to advanced research in robotics by combining theoretical insights with practical implementation.

References

• [1] H. J. Suh, M. Simchowitz, K. Zhang, and R. Tedrake, “Do differentiable simulators give better policy gradients?” in InternationalConference on Machine Learning. PMLR, 2022, pp. 20 668–20 696.
• [2] Schwarke, C., Klemm, V., Tordesillas, J., Sleiman, J. P., & Hutter, M. (2024). Learning Quadrupedal Locomotion via Differentiable Simulation. arXiv preprint arXiv:2404.02887

• Literature research
• Implementation of a differentiable simulation environment for ALMA
• Training and evaluation of tracking policies

• Excellent knowledge of Python
• Background in Simulation or Learning

Please send your CV, transcript and a short motivation (4-5 sentences max.) to:

cschwarke@ethz.ch vklemm@ethz.ch mittalma@ethz.ch

Stroke and traumatic brain injury (TBI) are debilitating conditions that often result in long-term physical and cognitive impairments. While rehabilitation therapy in the clinic plays a vital role in promoting recovery, maintaining patient adherence to therapy protocols, especially outside clinical environments, poses significant challenges. Leveraging mobile health technologies is an opportunity to enhance patient engagement and adherence to rehabilitation activities, for example, through smart push notifications triggered by the amount of therapy patients do with robotic devices.

In this project, we aim to investigate the impact of AI-driven push notifications on adherence to physical therapy among stroke and TBI patients. Collaborating with interdisciplinary teams, including clinicians, software developers, and researchers, the student will explore the effectiveness of push notifications in promoting sustained engagement with the at-home rehabilitation plan. The project will involve developing a smart push notifications system for our RehabCoach mobile application, collecting and analyzing data on the influence of push notifications on adherence to physical therapy in an unsupervised setting, and comparing it with real-time data of patients' interactions with ReHandyBot, the upper-limb rehabilitation robot used for therapy.

The student will conduct a comprehensive literature review to understand existing research on the efficacy of push notifications in healthcare contexts and their potential application in stroke and TBI unsupervised rehabilitation. Using this knowledge, they will design and implement a push notification system tailored to the needs of stroke and TBI patients in the RehabCoach app. The system will allow for personalized notification schedules and content customization based on patient preferences and therapy goals. By exploring the role of push notifications in promoting adherence to rehabilitation therapy, this project aims to contribute to the development of innovative solutions for improving patient outcomes in stroke and traumatic brain injury rehabilitation.

1. Conduct a literature review on push notification effectiveness in healthcare and rehabilitation.
2. Collaborate with clinicians and researchers to understand patient needs and therapy objectives.
3. Design and develop a push notification system for the RehabCoach app and integrate it with the existing LLM-based chatbots.
4. Implement data collection mechanisms to track patient interaction and adherence rates.
5. Analyze collected data to assess the impact of push notifications on rehabilitation therapy adherence.
6. Iterate the notification system based on feedback from patients and clinicians.
7. Document findings and contribute to research publications in relevant academic journals or conferences.

We are seeking a highly motivated master’s student or graduate with a background in software development, computer science, or a related field. The ideal candidate should have a keen interest in mobile health technologies and interdisciplinary research. Strong programming skills and the ability to work collaboratively in a team setting are essential. Additionally, candidates with an understanding of healthcare systems and patient-centered design principles will be preferred.

Alexandra Retevoi: alexandra.retevoi@hest.ethz.ch

Work packages

Literature research

Understand the training pipeline of the paper Robotic World Model: A Neural Network Simulator for Robust Policy Optimization in Robotics.

Explore the possibility of using a first-order gradient in optimizing the policy.

Requirements

Strong programming skills in Python

Experience in machine learning frameworks, especially model-based reinforcement learning.

Publication

This project will mostly focus on simulated environments. Promising results will be submitted to machine learning conferences, where the method will be thoroughly evaluated and tested on different systems (e.g., simple Mujoco environments to complex systems such as quadrupeds and bipeds).

Related literature

Hafner, D., Lillicrap, T., Ba, J. and Norouzi, M., 2019. Dream to control: Learning behaviors by latent imagination. arXiv preprint arXiv:1912.01603.

Hafner, D., Lillicrap, T., Norouzi, M. and Ba, J., 2020. Mastering atari with discrete world models. arXiv preprint arXiv:2010.02193.

Hafner, D., Pasukonis, J., Ba, J. and Lillicrap, T., 2023. Mastering diverse domains through world models. arXiv preprint arXiv:2301.04104.

Li, C., Stanger-Jones, E., Heim, S. and Kim, S., 2024. FLD: Fourier Latent Dynamics for Structured Motion Representation and Learning. arXiv preprint arXiv:2402.13820.

Song, Y., Kim, S. and Scaramuzza, D., 2024. Learning Quadruped Locomotion Using Differentiable Simulation. arXiv preprint arXiv:2403.14864.

Li, C., Krause, A. and Hutter, M., 2025. Robotic World Model: A Neural Network Simulator for Robust Policy Optimization in Robotics. arXiv preprint arXiv:2501.10100.

Please include your CV and transcript in the submission.

Chenhao Li

https://breadli428.github.io/

chenhli@ethz.ch

Simulation-based training of locomotion and environment interaction policies have recently shown tremendous success in pushing the abilities of real-world robots. Using massive parallelization, simulation-based learning enables robots to quickly learn new skills without involving the time and hardware investments attached to trying out things in the real world. However, one existing challenge is that such simulators are currently focusing on physics while the simulation of perception readings is often limited to simple geometry. In order to for example support end-to-end, vision-based models, we'd like to add realistic image rendering in complex, realistic environments to such simulators.

In this project, we'd like to explore a data-driven approach to add such capabilities to a simulator. Specifically, neural rendering methods have made large progress in recent years and their use in simulators for training and validation is now actively being investigated. Challenges that need to be addressed are given by 1) runtime considerations for efficient use inside a simulator, 2) artifact-free rendering of novel views, and 3) the imposition of physical constraints such as watertight meshes or the structural stability of static environment reconstructions.

The project is conducted at The AI Institute, a recently established top robotics research institute created by the founders of Boston Dynamics.

References

[1] Neuralangelo: High-Fidelity Neural Surface Reconstruction, CVPR 2023 [2] ViPlanner: Visual Semantic Imperative Learning for Local Navigation, ICRA 2024 [3] OmniRe: Omni Urban Scene Reconstruction, arxiv 2024 [4] 3D Gaussian Splatting for Real-Time Radiance Field Rendering, Siggraph 2023

-Literature research -Adding existing rendering functionality to simulator (similar to [2]) -Incorporate gaussian splatting based rendering into simulator -Improve gaussian splatting for the use in simulators (render quality, mesh extraction, …) -Setup validation pipeline for simulator (validate end-to-end policy, or VIO, …)

Excellent knowledge of Python Computer vision experience Knowledge of neural rendering methods

Alexander Liniger (aliniger@theaiinstitute.com) Igor Bogoslavskyi (ibogoslavskyi@theaiinstitute.com)

Please include up-o-date CV and transcript

Event cameras are a relatively new, vision-based exteroceptive sensor relying on standard CMOS technology. Unlike normal cameras, event cameras do not measure absolute brightness in a frame-by-frame manner, but relative changes of the pixel-level brightness. Essentially, every pixel of an event camera independently observes the local brightness pattern, and when the latter experiences a relative change of minimum amount with respect to a previous value, a measurement is triggered in the form of a time-stamped event indicating the image location as well as the polarity of the change (brighter or darker) [2]. The pixels act asynchronously and can potentially fire events at a very high rate. Owing to their design, event cameras do not suffer from the same artifacts as regular cameras, but continue to perform well under high dynamics or challenging illumination conditions.

Event cameras currently enjoy growing popularity and they represent a new, interesting alternative for exteroceptive sensing in robotics when facing scenarios with high dynamics and/or challenging conditions. The focus of the present thesis lies on 3D motion estimation with event cameras, and in particular aims at event-driven, computationally efficient methods that can trigger motion hypotheses from sparse raw events. Initial theoretical advances in this direction have been presented in recent literature [3,4,5], though these methods are still limited in terms of the assumptions that they make. The present thesis will push the boundaries by proposing novel both geometry and learning-based representations.

The proposed thesis will be conducted at the Robotics and AI Institute, a new top-notch partner institute of Boston Dynamics pushing the boundaries of control and perception in robotics. Selection is highly competitive. Potential candidates are invited to submit their CV and grade sheet, after which students will be invited to an on-site interview.

[1] Event-Based, 6-DOF Camera Tracking from Photometric Depth Maps, TPAMI 40(10):2402-2412, 2017

[2] The Silicon Retina, 264(5): 76-83, 1991

[3] A 5-Point Minimal Solver for Event Camera Relative Motion Estimation. In Proceedings of the International Conference on Computer Vision (ICCV), 2023

[4] An n-point linear solver for line and motion estimation with event cameras. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2024.

[5] Full-DoF Egomotion Estimation for Event Cameras Using Geometric Solvers, Arxiv: https://arxiv.org/html/2503.03307v1

● Literature research

●​ Extend the mathematical foundation for sparse event-based motion estimation

●​ Propose novel detectors that extend operability from lines and constant velocity motion to full 6 DoF motion estimation from either points or lines, or other specific object trajectories such as ballistic curves

●​ Investigate learning-based, sparse event-based motion detectors to handle more general cases

●​ Apply the technology to real-world data to track fast ego-motion or ballistic object motion in the environment

●​ Excellent knowledge of C++

●​ Computer vision experience

●​ Knowledge of geometric computer vision

●​ Plus: Experience with event cameras

Laurent Kneip (lkneip@theaiinstitute.com)

Please include your CV and up-​to-date transcript.

As 3D reconstruction [2,3], real-time data-driven rendering [4,5], and learning-based control technologies [6,7] are becoming more mature, recent efforts in reinforcement learning are moving towards end-to-end policies that directly consume images in order to generate control commands [8]. However, many of the simulated environments are limited to a composition of rigid objects. In recent years, the inclusion of differentiable particle-based simulation borrowed from computer graphics has enabled the inclusion of non-rigid or even fluid elements. Ideally, we can generate such representations from real world data in order to extend data-driven world simulators to arbitrary new objects with complex physical behavior.

The present thesis focuses on this problem and aims at reconstructing soft objects in terms of their geometry, appearance, and physical behavior. The goal is to make use of the Material Point Method (MPM) in combination with vision-based cues and physical priors in order to reconstruct accurate 3D models of soft objects. The developed models will finally be included into an RL learning environment such as Isaac-Gym in order to train novel manipulation policies for soft objects.

The proposed thesis will be conducted at the Robotics and AI Institute, a new top-notch partner institute of Boston Dynamics pushing the boundaries of control and perception in robotics. Selection is highly competitive. Potential candidates are invited to submit their CV and grade sheet, after which students will be invited to an on-site interview.

[1] Modeling of Deformable Objects for Robotic Manipulation: A Tutorial and Review, Front. Robot. AI, 7, 2020

[2] Global Structure-from-Motion Revisited, ECCV 2024

[3] MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priors, CVPR 2025

[4] NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis, CVPR 2020

[5] 3D Gaussian Splatting for Real-Time Radiance Field Rendering, SIGGRAPH 2023

[6] Learning to walk in minutes using massively parallel deep reinforcement learning, CoRL 2022

[7] Champion-Level Drone Racing Using Deep Reinforcement Learning, Nature, 2023

[8] π0: A Vision-Language-Action Flow Model for General Robot Control, Arxiv: https://arxiv.org/abs/2410.24164

● ​Literature research

●​ Design of suitable reconstruction method based on visual data and physical priors

●​ Dataset collection and testing

●​ Cross-validation against contact-based reconstruction methods

●​ Embedding into Isaac-Gym for training novel manipulation policies

● ​Excellent knowledge of Python or C++

●​ Computer vision experience

●​ Interest in optimization with physics representations

Laurent Kneip (lkneip@theaiinstitute.com)

Sina Mirrazavi (smirrazavi@theaiinstitute.com)

Please include your CV and up-​to-date transcript.

The proposed project aims to develop a flexible energy harvesting device that utilizes human body to scavenge ambient electromagnetic (EM) energy. With the rapid expansion of wireless technologies and ubiquitous EM radiation in urban environments, this project explores a novel and sustainable approach to convert ambient EM energy into usable electrical power for low-energy wearable electronics. Unlike conventional rectenna-based harvesters, this approach leverages the human body as a functional dielectric medium to enhance polarization-induced charge accumulation at the electrode interface.

Methodology

• Materials selection and device fabrication.

• Electrical performance characterization is based on oscilloscope and other equipment.

• PCB circuit design and fabrication.

• Low contact impedance interface device design and fabrication.

• Electrical and mechanical characterization and performance tuning: evaluate the output electrical signal under different parameters.

• Design and processing of back-end processing circuits.

• Application exploration: Integrate the designed device with wearable devices.

• Write scientific project report / research articles (if possible)

Tasks

• Literature review (10%)

• Device design and fabrication (30%)

• Characterization and performance tuning (20%)

• Application implementation (30%)

• Reporting and presentation (10%)

Your Profile

• Background in Applied Physics, Mechanical Engineering, Chemical Engineering, Materials Science, Electronics Engineering, Biomedical Engineering or related fields

• Independent worker with critical thinking and problem-solving skills

Prof Dr Carlo Menon, Dr. Weifeng Yang and Yuanlong Li will supervise the student and the research will be performed at ETH Zurich’s Biomedical and Mobile Health Technology research group (www.bmht.ethz.ch) in the Balgrist Campus in Zurich, Switzerland.

To apply, use the button below to tell us why you want to do this project, attach a CV with your current program of study, your grades and any other info you deem relevant. Please include length of time that your project or thesis will occupy (i.e. 2 months, 6 months, etc).

Transporting loads with drones is often constrained by traditional control systems that rely on predefined flight paths, GPS, or external motion capture systems. These methods limit a drone's adaptability and responsiveness, particularly in dynamic or cluttered environments. Vision-based control has the potential to revolutionize aerial transportation by enabling drones to perceive and respond to their surroundings in real-time. Imagine a drone that can swiftly navigate through complex environments and deliver payloads with precision using only onboard vision sensors. Applicants are expected to be proficient in Python, C++, and Git.

This project aims to develop a vision-based control system for drones capable of agile and efficient aerial transportation. The system will leverage real-time visual input to dynamically adapt to environmental conditions, navigate obstacles, and manage load variations with reinforcement or imitation learning.

Interested candidates should send their CV, transcripts (bachelor and master), and descriptions of relevant projects to Ismail Geles [geles (at) ifi (dot) uzh (dot) ch], Leonard Bauersfeld [bauersfeld (at) ifi (dot) uzh (dot) ch], Angel Romero [roagui (at) ifi (dot) uzh (dot) ch]

The Personalized Low Latency Interactive AI project addresses the limitations of current voice recognition systems, which often struggle with high latency, limited personalization, and poor performance in noisy settings.The project integrates robust speech-to-text conversion, Flash Attention-2 for low-latency processing, and advanced distilled large language models (LLMs) for real-time personality detection and response generation. The system ensures precise speaker identification and tailored interactions. User-specific memories are stored in a GDPR-compliant local database, enabling dynamic personalization based on preferences and personality traits. The project aims to revolutionize voice-based interactions in office and public settings, with a focus on scalability, accessibility, and sustainability.The Personalized Low Latency Interactive AI project aims to create a highly personalized, low-latency AI agent tailored for wheelchair users to support their mental well-being. Deployed on an Nvidia Orin 64GB development kit, the platform ensures seamless, real-time responses in noisy, multi-user environments, prioritizing accessibility and user engagement.

You will focus on integrating personality trait and preference estimation into the distilled LLM response system to create a personalized AI agent for wheelchair users.You will work on enhancing the system’s ability to learn user-specific memories through interactions, using Mediapipe for lip-reading and Insightface for person identification, while optimizing latency on the Nvidia Orin 64GB development kit.

• Implement personality trait and preference estimation using distilled LLMs.
• Integrate Lip-reading and Person identification for accurate memory retrieval.
• Optimize system latency for seamless, real-time responses on Nvidia Orin hardware.
• Test and validate the personalization pipeline for wheelchair users’ mental well-being.
• Develop and test memory-based personalization algorithms in Months 2–3.
• Integrate lip-reading and person identification modules on Nvidia Orin in Months 1–2.
• Optimize latency and validate system performance in Months 4–5.
• Prepare a thesis or report documenting findings in Month 6.

This role offers a unique opportunity to work with cutting-edge AI models, lip-reading, and person identification technologies while contributing to accessibility and social impact.

• Gain hands-on experience with distilled LLMs, Speech-to-Text and Person Identification Neural Networks.
• Develop expertise in low-latency AI systems and personalization algorithms.
• Collaborate with a multidisciplinary team of AI and accessibility experts.
• Produce a thesis or report with potential for academic publication.
• Access to state-of-the-art Nvidia Orin 64GB development kit.
• Opportunity to contribute to a high-impact project with real-world accessibility applications.
• Mentorship from researchers in AI and human-computer interaction.

We are looking for a driven candidate with a passion for AI, accessibility, and low-latency systems. The ideal candidate is a student pursuing a thesis or internship with strong technical skills and an interest in enhancing mental well-being through AI.

• Enrolled in a Master’s program in Computer Science, Engineering, or a related field.
• Experience with Python and familiarity with LLMs, signal processing, or computer vision.
• Knowledge of Mediapipe, Insightface, or Nvidia hardware is a plus.
• Strong problem-solving skills and commitment to accessibility-focused research.
• Available for a 6-month internship or thesis project.

For inquiries or to apply, please reach out with your CV and a brief cover letter. Applications will be reviewed on a rolling basis, and shortlisted candidates will be invited for an interview.

Email: shreyasvi.natraj@hest.ethz.ch, diego.paez@hest.ethz.ch Start Date: Flexible

Navigating in crowded environments is one of the most challenging robotic tasks due to the dynamic setting and the variety of human movements. This project seeks to overcome the limitations faced by planners in such dynamic environments while, in addition, being exposed to social norms that are crucial for safely (e.g., street crossing on a crosswalk, staying on the sidewalk). While recent advancements in reinforcement-learning-based local planners have demonstrated success in demanding scenarios, they require privileged information about human movements and cannot handle semantic constraints [1]. On the contrary, planners following semantic constraints remain limited to low update rates and static scenes [2]. To address these issues, we started to design a novel planner capable of safely navigating to goal positions while being exposed to different human movement patterns and semantic constraints, which has already achieved promising results. This project aims to advance the work and bring it to the real system (either ANYmal or Unitree B2W). Challenges of this project will include training a stable policy in RL using the IsaacLab framework[3], enhancing the variety of movement patterns, and performing the transfer from simulation to the real hardware.

• Literature review on crowd and social navigation
• Planner development using RL in Simulation
• Integration of ANYmal and real-world evaluation

• Extensive programming experience (Python) with large codebases
• Experience with deep learning projects (preferably with RL)
• Knowledge of planning, perception, and robot dynamics is a plus

Please send your CV and TOR to Fan Yang (fanyang1 @ethz.ch) and Pascal Roth (rothpa@ethz.ch)

Work packages

Literature research

Implement the teacher-student training baseline.

Implement the proposed active system identification method.

Systematic analysis and comparison between the methods, with visualization of the latent environment embedding.

Hardware deployment.

Requirements

Strong programming skills in Python

Experience in machine learning frameworks, especially reinforcement learning.

Publication

This project will mostly focus on simulated environments. Promising results will be submitted to machine learning conferences, where the method will be thoroughly evaluated and tested on different systems.

Related literature

https://www.science.org/doi/10.1126/scirobotics.aau5872

https://www.science.org/doi/10.1126/scirobotics.abc5986

https://www.science.org/doi/full/10.1126/scirobotics.abk2822

Please include your CV and transcript in the submission.

Chenhao Li

https://breadli428.github.io/

chenhli@ethz.ch

Work packages

Use video diffusion models to generate reference trajectories with semantic conditions.

Train a local navigation policy by learning from demonstrations using the generated trajectories.

Deployment on hardware.

Requirements

Strong programming skills in Python

Experience in reinforcement learning and imitation learning frameworks

Publication

This project will mostly focus on algorithm design and system integration. Promising results will be submitted to robotics or machine learning conferences where outstanding robotic performances are highlighted.

Related literature

Peng, Xue Bin, et al. "Deepmimic: Example-guided deep reinforcement learning of physics-based character skills." ACM Transactions On Graphics (TOG) 37.4 (2018): 1-14.

Li, C., Vlastelica, M., Blaes, S., Frey, J., Grimminger, F. and Martius, G., 2023, March. Learning agile skills via adversarial imitation of rough partial demonstrations. In Conference on Robot Learning (pp. 342-352). PMLR.

Li, Chenhao, et al. "FLD: Fourier latent dynamics for Structured Motion Representation and Learning."

Serifi, A., Grandia, R., Knoop, E., Gross, M. and Bächer, M., 2024, December. Vmp: Versatile motion priors for robustly tracking motion on physical characters. In Computer Graphics Forum (Vol. 43, No. 8, p. e15175).

Fu, Z., Zhao, Q., Wu, Q., Wetzstein, G. and Finn, C., 2024. Humanplus: Humanoid shadowing and imitation from humans. arXiv preprint arXiv:2406.10454.

He, T., Luo, Z., Xiao, W., Zhang, C., Kitani, K., Liu, C. and Shi, G., 2024, October. Learning human-to-humanoid real-time whole-body teleoperation. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (pp. 8944-8951). IEEE.

He, T., Luo, Z., He, X., Xiao, W., Zhang, C., Zhang, W., Kitani, K., Liu, C. and Shi, G., 2024. Omnih2o: Universal and dexterous human-to-humanoid whole-body teleoperation and learning. arXiv preprint arXiv:2406.08858.

Albaba, M., Li, C., Diomataris, M., Taheri, O., Krause, A. and Black, M., 2025. Nil: No-data imitation learning by leveraging pre-trained video diffusion models. arXiv preprint arXiv:2503.10626.

Please include your CV and transcript in the submission.

Chenhao Li

https://breadli428.github.io/

chenhli@ethz.ch

Work packages

Literature research

Understand the training pipeline of the paper Robotic World Model: A Neural Network Simulator for Robust Policy Optimization in Robotics.

Explore the possibility of using a first-order gradient in optimizing the policy.

Requirements

Strong programming skills in Python

Experience in machine learning frameworks, especially model-based reinforcement learning.

Publication

This project will mostly focus on simulated environments. Promising results will be submitted to machine learning conferences, where the method will be thoroughly evaluated and tested on different systems (e.g., simple Mujoco environments to complex systems such as quadrupeds and bipeds).

Related literature

Hafner, D., Lillicrap, T., Ba, J. and Norouzi, M., 2019. Dream to control: Learning behaviors by latent imagination. arXiv preprint arXiv:1912.01603.

Hafner, D., Lillicrap, T., Norouzi, M. and Ba, J., 2020. Mastering atari with discrete world models. arXiv preprint arXiv:2010.02193.

Hafner, D., Pasukonis, J., Ba, J. and Lillicrap, T., 2023. Mastering diverse domains through world models. arXiv preprint arXiv:2301.04104.

Li, C., Stanger-Jones, E., Heim, S. and Kim, S., 2024. FLD: Fourier Latent Dynamics for Structured Motion Representation and Learning. arXiv preprint arXiv:2402.13820.

Song, Y., Kim, S. and Scaramuzza, D., 2024. Learning Quadruped Locomotion Using Differentiable Simulation. arXiv preprint arXiv:2403.14864.

Li, C., Krause, A. and Hutter, M., 2025. Robotic World Model: A Neural Network Simulator for Robust Policy Optimization in Robotics. arXiv preprint arXiv:2501.10100.

Please include your CV and transcript in the submission.

Chenhao Li

https://breadli428.github.io/

chenhli@ethz.ch

Wearable sensors and electronic components are steadily entering everyday and biomedical settings. Modern advances in this field try to transfer these technologies from silicon embodiments onto fabric platforms. Such e-textiles would allow for non-invasive, unobtrusive, and easy usage of these wearable devices. And while sensors had plenty of success transferring to all-textile format, technologies capable of displaying information on fabric and garments are rare.

The present project aims to develop a technology capable of displaying various information on wearable all-textile platform. For this, a textile material capable of changing colors upon application of heat would be developed. A prototype of a textile display that would utilize these properties would be then assembled using the developed material.

Goals

• Study the thermochromic phenomena in textiles

• Develop thermochromic textile prototype

• Produce and test a prototype of wearable thermochromic display

• Write a scientific project report

Tasks

• Literature review (10%)

• Thermochromic textile development (40%)

• Wearable display development (40%)

• Data collection and analysis, reporting and presentation (10%)

Your Profile

• Background in Applied Physics, Chemical Engineering, Chemistry, Materials Science, Electronics Engineering, Biomedical Engineering or related fields

• Independent worker with critical thinking and problem-solving skills

Prof Dr Carlo Menon and Dr. Alexander Shokurov will supervise the student and the research will be performed at ETH Zurich’s Biomedical and Mobile Health Technology research group (www.bmht.ethz.ch) in the Balgrist Campus in Zurich, Switzerland.

To apply, use the button below to tell us why you want to do this project ("motivation"); attach a mini CV with your current program of study, your grades and any other info you deem relevant--maybe the name and e-mail of a postdoc or a professor willing to be your reference; and make any further comments ("additional remarks"). Please include length of time that your thesis will occupy (i.e. 6 months, etc), and the earliest date you can start.

One of the most important biomedical parameters that can be measured by wearable devices is human body movement. This movement is related limb motion, breathing, speech, heart rate etc. Accurate measurement of these movements requires precise strain and pressure sensors with high sensitivity.

Mechanophores are materials capable of changing physical properties (most often color) in response to local mechanical stimuli, like strain or stress. This is achieved by insertion of stress-responsive molecular units into the polymeric backbone.

In the present project, mechanophoric elastomeric materials would be synthesized to study if the mechanophoric action can enhance sensitivity of strain sensors for wearable applications. First, new approaches for facile upscaled production of mechanophoric elastomers will be developed through synthesis of small molecule functional cross-linkers, and then the developed mechanophore polymers will be applied as active material for electronic strain sensors.

We intend to carry out an exciting multidisciplinary study of the organic synthesis strategies, physico-chemical behavior of the new materials, and their practical applications in wearable sensors.

Goals

• Synthesize functional mechanophoric cross-linker small molecules;

• Incorporate synthesized mechanophores into elastomers and verify mechanophoric action upon linear strain;

• Validate the effect of mechanophoric units on strain sensing;

• Write a scientific project report;

Tasks

• Literature review (10%)

• Synthesis of functional cross-linkers and elastomers (40%)

• Validation of the effect of mechanophoric behavior on strain sensor properties (40%)

• Reporting and presentation (10%)

Your Profile

• Background in Chemistry, Chemical Engineering, Materials Science, or related fields

• Independent worker with critical thinking and problem-solving skills

Contact Details Prof Dr Carlo Menon and Dr. Alexander Shokurov will supervise the student and the research will be performed at ETH Zurich’s Biomedical and Mobile Health Technology research group (www.bmht.ethz.ch) in the Balgrist Campus in Zurich, Switzerland.

To apply, use the button below to tell us why you want to do this project ("motivation"); attach a CV with your current program of study, your grades and any other info you deem relevant.

Proper iodine intake is crucial for optimal health, impacting everything from metabolism to cognitive development in utero, as well as in both children and grown adults. However, assessing iodine levels can be challenging, particularly in point-of-care or remote settings. This project seeks to address this gap by developing a disposable sensor specifically designed to measure urinary iodine levels, providing rapid and accurate insights into patient nutrition.

The proposed project aims to utilize electrochemical detection and synthesis techniques involving carbon- and silver- based nanomaterials to produce and validate sensors for aqueous iodine. This work will include the design and fabrication of the sensor, calibration for iodine detection in water, synthetic urine, and then in real samples. In case of high success, portable electronics capable of driving the developed detection protocol will be also elaborated and validated.

Goals

• Develop the method for the graphene electrode formation and deposition of nanosilver

• Validate and test the sensor’s accuracy and reliability in measuring urinary iodine levels in synthetic samples of progressive complexity and real samples

• Develop electronic suite capable of driving the developed sensors

• Write a scientific project report

Tasks

• Literature review (10%)

• Optimize the deposition technique and fabrication process for the sensor element (50%)

• Validation of the sensor in synthetic and real life samples (20%)

• Development of signal read-out electronic components (10%)

• Data collection and analysis, reporting and presentation (10%)

Your Profile

• Background in Biomedical technology, Chemistry, Materials Science or related fields

• Prior experience with chemicals and standard chemical laboratory equipment

• Knowledge of electrochemistry and/or electrochemical sensing is highly desirable

• Independent worker with critical thinking and problem-solving skills

Prof Dr Carlo Menon and Dr. Alexander Shokurov will supervise the student and the research will be performed at ETH Zurich’s Biomedical and Mobile Health Technology research group (www.bmht.ethz.ch) in the Balgrist Campus in Zurich, Switzerland. When applying, please describe your motivation and your background in the cover letter, and attach your CV and all the transcripts of the previous studies.

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This project addresses a novel approach for simulating and controlling rock fracture processes using a combination of finite element analysis (FEM), supervised learning, and reinforcement learning (RL). The goal is to develop an intelligent system capable of predicting fracture patterns and actively guiding the fracturing process towards a desired rock geometry.

The pipeline begins with a fracture simulation based on FEM, which models the physical stresses and crack propagation given the original rock geometry [1]. The data generated from various of these simulations is then used to train a graph neural network (GNN) via supervised learning. This GNN learns to predict the fracture behavior and resulting rock shape based on initial conditions and applied forces, and generalizes across geometries and materials [2]. Concurrently, an RL agent is developed. This agent's objective is to determine the optimal actions (e.g., applied forces or drilling patterns) to achieve a desired rock geometry. The RL agent learns through interaction with a training environment which leverages the pre-trained GNN to rapidly simulate the outcomes of the agent's actions on the rock geometry.

By integrating FEM-based simulation with advanced machine learning techniques, this approach aims to create a powerful tool for understanding and controlling rock fracturing, with potential applications in fields such as mining, demolition and large scale construction robotics [3]. The GNN enables accurate and fast prediction of fracture mechanics on any geometries, while the RL agent provides intelligent, goal-oriented control over the fracturing process.

[1] Li, J., Dai, L., Wang, S., Liu, Y., Sun, Y., Wang, J. and Zhang, A., 2024. Theoretical and numerical analysis of the rock breaking process by impact hammer. Powder Technology, 448, p.120254.

[2] Mousavi, S., Wen, S., Lingsch, L., Herde, M., Raonić, B. and Mishra, S., 2025. RIGNO: A Graph-based framework for robust and accurate operator learning for PDEs on arbitrary domains. arXiv preprint arXiv:2501.19205.

[3] Ryan Luke Johns et al. ,A framework for robotic excavation and dry stone construction using on-site materials.Sci. Robot.8,eabp9758(2023).DOI:10.1126/scirobotics.abp9758

• GNN training for 2D and 3D rock shapes
• Online adaptation of rock parameters from observed fracture patterns
• RL agent identifying the optimal impact point, and performing the fracturing

• Programming experience in python, and with common learning frameworks
• Relevant knowledge in machine learning and deep learning
• Experience with RL, GNNs, and/or JAX is a plus

• Filippo Spinelli (fspinelli@ethz.ch)
• Sepehr Mousavi (smousav@ethz.ch)

Work packages

Understand why SAC underperforms in legged locomotion and establish baselines.

Make SAC work with massively parallel simulation frameworks.

Validate performance of improved SAC policies on real hardware.

Provide a rigorous, reproducible comparison between SAC and PPO.

Hardware validation encouraged

Requirements

Strong programming skills in Python

Experience in reinforcement learning frameworks, including SAC and PPO

Publication

This project will mostly focus on algorithm design and system integration. Promising results will be submitted to robotics or machine learning conferences.

Related literature

PPO

SAC

Please include your CV and transcript in the submission.

Chenhao Li

https://breadli428.github.io/

chenhli@ethz.ch

Our project focuses on overcoming current limitations in neuromuscular biohybrid robots, particularly scaffold design and the development of functional neuromuscular junctions (NMJs). You will apply expertise in biology, biomaterial synthesis, and biofabrication to tackle the key challenges of creating functional engineered tissues within the field of biohybrid robotics.

By incorporating moto-neurospheres into 3D co-culture systems with skeletal muscle tissues, we aim to establish global innervation and enable synchronized muscle contractions.

abadolato@ethz.ch rkk@ethz.ch

At the Rehabilitation Engineering Laboratory (Relab) at ETH Zurich, we have developed an innovative functional near-infrared spectroscopy (fNIRS) technology in collaboration with the neurotechnology startup Optohive. Optohive is building a cutting-edge platform for real-time neurological signal acquisition and analysis. We are offering a Masters thesis project for a motivated student to develop a complete signal processing pipeline tailored to neurological data, with the goal of detecting early biomarkers of cognitive or neurological conditions. This project blends neuroscience, signal processing, and artificial intelligence in a practical and high-impact context.

• A reproducible signal processing and machine learning pipeline (Python preferred)
• Comprehensive research documentation and result visualizations
• Integration of the pipeline into the Optohive platform (in collaboration with the development team)
• Final written report and scientific presentation

Feature Extraction - Time- and frequency-domain analysis (FFT, PSD) - Detection of task-evoked hemodynamic responses and connectivity patterns - Event-related potential (ERP) analysis - Extraction of statistical, entropy-based, and domain-specific features

AI & Machine Learning Integration - Classification and clustering using SVM, Random Forests, and deep learning models - Dimensionality reduction and feature selection (e.g., PCA, LDA) - Anomaly detection and discovery of temporal patterns in brain activity

Biomarker Discovery - Identifying features associated with specific cognitive or physiological states - Statistical validation and visualization of biomarkers - Benchmarking against existing approaches through literature review

• Background in biomedical signal processing, neuroscience, biomedical engineering or physics
• Strong programming skills in Python (NumPy, SciPy, MNE, scikit-learn, PyTorch/TensorFlow)
• Experience with EEG, fNIRS, or related neuroimaging modalities
• Interest in AI/ML applications in neuroscience and healthcare

For more information, please reach out to: marc.willhaus@hest.ethz.ch or dominik.wyser@hest.ethz.ch relab.ethz.ch | optohive.com

At the Rehabilitation Engineering Laboratory (Relab) at ETH Zurich, we are at the forefront of translating brain research into clinical applications. In collaboration with the neurotechnology startup Optohive, we have developed a novel functional near-infrared spectroscopy (fNIRS) system to record high-resolution neurological data. We are offering a Master’s thesis project focused on designing deep learning models for the analysis of brain time-series data. The objective is to uncover meaningful patterns and biomarkers from real-time neurophysiological recordings. You will work closely with our interdisciplinary team to shape the future of brain-computer interfaces through cutting-edge AI.

• A reproducible deep learning pipeline (Python preferred, using PyTorch or TensorFlow)
• Structured evaluation and comparative analysis of model architectures
• A biomarker candidate report with physiological interpretations
• Final thesis report and potential contribution to a scientific publication

Data Handling & Preparation - Sequence shaping and time-windowing for temporal data - Data augmentation techniques specific to neurological signals - Task-aligned labeling with cognitive events or external triggers Model Development - Design and train time-series models (CNNs, LSTMs, GRUs, Transformers) - Explore hybrid architectures combining convolutional and recurrent layers - Incorporate multimodal brain imaging data - Investigate self-supervised and contrastive learning for limited-label datasets Model Evaluation - Evaluate model performance using metrics like accuracy, F1-score, AUC - Visualize model insights through techniques like Grad-CAM, SHAP, or saliency maps - Analyze latent representations and decision boundaries Biomarker Discovery & Integration - Identify signal features learned by models that correlate with neurological biomarkers - Validate findings against known physiological indicators - Integrate trained models into Optohive’s signal processing platform

• Strong Python skills and experience with deep learning libraries (PyTorch/TensorFlow)
• Solid foundation in machine learning, biomedical signal processing, or applied AI
• Familiarity with neurophysiological data (EEG, fNIRS, EMG, etc.) or general time-series data
• Interest in neurotechnology and the development of brain-computer interfaces

For more information, please reach out to: marc.willhaus@hest.ethz.ch or dominik.wyser@hest.ethz.ch relab.ethz.ch | optohive.com

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

Humanoid robot hardware is beginning to match human capabilities in terms of agility and dynamic movement. This progress enables the use of human demonstrations as references for learning human-aligned locomotion skills. At the same time, Meta`s Project Aria allows to cheaply estimating the human body poses at low cost while only wearing glasses. Existing approaches use this demonstration data to train reinforcement learning (RL) agents in simulation with a formulation that incentivizes mimicking the reference motion. These policies either mimic the human demonstration directly or use motion retargeting mechanisms to transfer the style onto an RL agent trained in simulation. While these approaches have shown impressive results on flat ground, they often lack adaptability to complex environments—such as jumping over obstacles, climbing, or performing parkour-style movements that require environment-conditioned responses. This project addresses that limitation by collecting an aligned dataset of human motion in complex terrain, including climbing and parkour. The data is captured using Meta Project Aria glasses, which provide an affordable way to estimate human body poses. In addition, a precise 3D scan of the environment is generated to create an accurate mesh that can be transferred to the simulator. This setup enables training RL agents with a higher level of environmental awareness and adaptability.

What We Offer: - Active support and mentorship, including help on potential publications - Access to high-performance compute clusters - Weekly meetings and career guidance

Work Packages: - Implementing imitation learning RL training pipeli- ne on rough terrain - - Support the collection of imitation data - - Support creating of the real-2-sim pipeline

Who We're Looking For: - A highly motivated student or researcher - Experience with PyTorch or similar ML frameworks - (Optional) Background in computer vision

Please contact Chenhao Li, Jonas Frey and Xi Wang via Mail. chenhao.li@ai.ethz.ch (CC. jonfrey@ethz.ch, xi.wang@inf.ethz.ch)

• Subject: Application: RL Learning Progress - Firstname Lastname
• Content: Please provide 3 sentences why you would like to do the project within the mail.
• Appendix: Bachelor transcript. Current Master transcript. CV. If you did not obtain your Bachelor`s at ETH please provide a relative grade ranking.

Neurological disorders often result in upper limb impairments, which affect quality of life and increase dependency. The VPIT provides valuable data through robotic testing of 3D position and grip force during a reaching task. This thesis will use AI techniques to analyze this data, aiming to uncover novel metrics that provide insights into motor impairments and offer better diagnosis and rehabilitation approaches.

Explore time series data from the VPIT using explainable AI methods to identify and interpret new movement and force metrics that describe upper limb impairments in individuals with neurological disorders.

• Prepare and clean the VPIT time series data (position and grip force) for analysis, ensuring it is ready for machine learning modeling.
• Apply machine learning models (e.g., LSTMs, CNNs, Random Forests) to the raw time series data and use explainable AI techniques (e.g., SHAP, LIME, attention mechanisms) to uncover important movement and force patterns that relate to upper limb impairments.
• Analyze the results to identify key temporal features that correlate with neurological impairments, and provide insights into their potential role in diagnosis and rehabilitation.

• Strong background in machine learning and time series analysis, with proficiency in Python.
• Interest in applying AI techniques to clinical data, particularly in neurological research.
• Motivated, independent, and creative problem-solving approach.
• Strong communication skills in English.

If you are interested or have any further questions, please contact: Nadine Domnik (nadine.domnik@hest.ethz.ch). Please include your CV and transcript of records in your application.

The VPIT consists of a haptic device (Geomagic Touch) connected to a virtual environment, used to assess sensorimotor function of the upper limb, while the Inverse3 is a newer haptic device designed for 3D interaction and navigation. This project will compare the two devices by evaluating their hardware performance (precision, response times, and capabilities) and software APIs (ease of use, flexibility, and compatibility). Both devices will be tested to assess their ability to measure movement and grip force.

Evaluate the potential of the Inverse3 haptic device as a replacement or complement to the VPIT for upper limb assessments by comparing the hardware features, performance, and software capabilities.

• Compare the hardware features (precision, response time) and usability of the VPIT and Inverse3 devices during upper limb tasks.
• Analyze the software API of both devices, focusing on ease of integration and flexibility in clinical settings.
• Compare the movement and force data generated by both devices and assess their ability to measure relevant upper limb functions.
• Determine whether the Inverse3 can be a valid alternative to VPIT for clinical use, providing recommendations for improvements or applications.

• Familiar with hardware (testing and evaluation), preferably with haptic devices.
• Proficiency in programming and working with software APIs (Python preferred).
• Interest in rehabilitation technologies and clinical research.
• Strong analytical and communication skills.

If you are interested or have any further questions, please contact: Nadine Domnik (nadine.domnik@hest.ethz.ch). Please include your CV and transcript of records in your application.

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

Stroke impacts over 14 million people annually, creating a massive global demand for effective rehabilitation solutions and placing a significant burden on healthcare systems. The Rehabilitation Engineering Laboratory (RELab) at ETH Zurich is pioneering a game-changing closed-loop neurorehabilitation device that integrates cutting-edge motion tracking with non-invasive brain stimulation to accelerate motor recovery in stroke patients. As this revolutionary technology approaches the commercialization phase, a compelling financial model and a strategic business plan are essential to capture investor interest, secure funding, and dominate the growing neurorehabilitation market.

This project offers an unpaid internship or thesis opportunity for a driven student to spearhead the development of a robust financial model and a high-impact business plan for this innovative medical device. Working closely with a dynamic team of engineers, neuroscientists, and business strategists, the student will gain unparalleled experience in financial forecasting, market analysis, and startup strategy within the high-stakes medical device sector. This role is tailored for ambitious, business-minded individuals eager to blend entrepreneurial acumen with biomedical innovation to launch a transformative healthcare solution

The goal of this project is to craft a sophisticated financial model and a persuasive business plan to drive the successful commercialization of a breakthrough neurorehabilitation device. This involves projecting financial viability, defining a scalable business model, and developing a market entry strategy that positions the device as a leader in the stroke rehabilitation market. The project will focus on quantifying market opportunities, engaging key stakeholders (e.g., healthcare providers, insurers, investors), and building a roadmap to achieve rapid market penetration and sustainable growth.

• Perform in-depth market research to assess the global neurorehabilitation market, identifying growth trends, key competitors, and untapped opportunities in stroke rehabilitation.
• Build a comprehensive financial model, including R&D and manufacturing cost projections, pricing strategies, revenue forecasts, and ROI scenarios to attract investors.
• Develop a strategic business plan, encompassing executive summary, competitive analysis, go-to-market strategy, operational roadmap, and funding requirements.
• Investigate reimbursement models and payer dynamics (e.g., private insurers, Medicare, European healthcare systems) to optimize pricing and market access.
• Conduct a SWOT analysis to highlight the device’s competitive edge and address potential market risks.
• Create investor-focused deliverables, such as pitch decks, financial summaries, and one-pagers, to secure funding from venture capitalists and strategic partners.
• Ensure all financial and business planning materials are well-organized, professional, and ready for stakeholder presentations.
• (For thesis students) Produce a thesis analyzing the financial and strategic framework for commercializing the device, with actionable recommendations for market leadership and scalability.

We are seeking a highly motivated BSc or MSc student with a background in Business Administration, Finance, Entrepreneurship, Economics, or a related field. Candidates with an interest in Biomedical Engineering or Health Sciences are also welcome if they have a strong business orientation. Ideal candidates should have: - Passion for entrepreneurship, medical device innovation, and disrupting the healthcare industry. - Experience or strong interest in financial modelling, business strategy, or startup development. - Proficiency in Excel for financial analysis; familiarity with Python, R, or business intelligence tools is a plus. - Understanding of the healthcare or medical device sector is advantageous but not required. - Sharp analytical skills and a knack for translating data into actionable business insights. - Exceptional communication skills in English to craft persuasive business documents and pitches; German is a plus for European market analysis. - A proactive, results-driven mindset and the ability to thrive in a fast-paced, interdisciplinary startup environment.

This project offers a rare chance to hone high-demand skills in financial strategy, startup planning, and medical device commercialization while shaping the future of healthcare.

To apply, please submit your application via the button below. Include a concise motivation statement outlining why you’re excited about this project and how your business skills will drive its success. Attach a mini CV highlighting your current program of study, grades, and any relevant experience or skills. Contact: Dr. Paulius Viskaitis: paulius.viskaitis@hest.ethz.ch Dr. Dane Donegan: dane.donegan@hest.ethz.ch

Project Description:

Navigation can be viewed similarly to next-word prediction in language models—a seemingly simple task that requires a deep understanding of the environment. In this project, we aim to advance reinforcement learning (RL) for navigation by investigating whether an agent can autonomously select increasingly challenging tasks that drive continuous improvement in its navigation behavior.

The core idea is to use learning progress as a signal for task difficulty. Tasks where the agent is improving are considered valuable, as they are neither too easy (where the agent already performs well) nor too hard (where little to no learning occurs). This naturally focuses learning on the "sweet spot" of challenge, avoiding both uninformative, trivial experiences and tasks beyond the agent’s current capability.

By adopting this approach, we can train more sophisticated navigation agents in procedurally generated environments with randomly assigned goals—without relying on hand-crafted heuristics to define a curriculum of terrain difficulty.

What We Offer:

• Active support and mentorship, including help on potential publications
• Collaboration with leading international institutions (Stanford, UC Berkeley)
• Access to high-performance compute clusters
• Weekly meetings and career guidance

Related Work:

• Rudin, Nikita, David Hoeller, Philipp Reist, and Marco Hutter. "Learning to walk in minutes using massively parallel deep reinforcement learning." In Conference on Robot Learning, pp. 91-100. PMLR, 2022.
• Lee, Joonho, Marko Bjelonic, Alexander Reske, Lorenz Wellhausen, Takahiro Miki, and Marco Hutter. "Learning robust autonomous navigation and locomotion for wheeled-legged robots." Science Robotics 9, no. 89 (2024): eadi9641.
• Zhang, Chong, Jin Jin, Jonas Frey, Nikita Rudin, Matías Mattamala, Cesar Cadena, and Marco Hutter. "Resilient legged local navigation: Learning to traverse with compromised perception end-to-end." In 2024 IEEE International Conference on Robotics and Automation (ICRA), pp. 34-41. IEEE, 2024.
• Wang, Xin, Yudong Chen, and Wenwu Zhu. "A survey on curriculum learning." IEEE transactions on pattern analysis and machine intelligence 44, no. 9 (2021): 4555-4576.
• Portelas, Rémy, Cédric Colas, Katja Hofmann, and Pierre-Yves Oudeyer. "Teacher algorithms for curriculum learning of deep rl in continuously parameterized environments." In Conference on Robot Learning, pp. 835-853. PMLR, 2020.
• Li, Chenhao, Elijah Stanger-Jones, Steve Heim, and Sangbae Kim. "Fld: Fourier latent dynamics for structured motion representation and learning." arXiv preprint arXiv:2402.13820 (2024).

• Highly motivated
• Experience with PyTorch or similar ML frameworks
• (Optional) Background in RL

Please contact Chenhao Li and Jonas Frey via Mail. chenhao.li@ai.ethz.ch (CC. jonfrey@ethz.ch)

• Subject: Application: RL Learning Progress - Firstname Lastname

• Content: Please provide 3 sentences why you would like to do the project within the mail.

• Appendix: Bachelor transcript. Current Master transcript. CV. If you did not obtain your Bachelor`s at ETH please provide a relative grade ranking.

Training sub-optimal policies is relatively straightforward and provides a solid foundation for reinforcement learning (RL) agents. However, these policies cannot improve online in the real world, such as when racing drones with RL. Current methods fall short in enabling drones to adapt and optimize their performance during deployment. Imagine a drone equipped with an initial sub-optimal policy that can navigate a race course but not with maximum efficiency. As the drone races, it learns to optimize its maneuvers in real-time, becoming faster and more agile with each lap. Applicants are expected to be proficient in Python, C++, and Git.

This project aims to explore online fine-tuning in the real world of sub-optimal policies using RL, allowing racing drones to improve continuously through real-world interactions.

Interested candidates should send their CV, transcripts (bachelor and master), and descriptions of relevant projects to Ismail Geles [geles (at) ifi (dot) uzh (dot) ch], Elie Aljalbout [aljalbout (at) ifi (dot) uzh (dot) ch]

Drone racing demands split-second decisions and precise maneuvers. However, training drones for such races relies heavily on crafted reward functions. These methods require significant human effort in design choices and limit the flexibility of learned behaviors. Inverse Reinforcement Learning (IRL) offers a promising alternative. IRL allows an AI agent to learn a reward function by observing expert demonstrations. Imagine an AI agent analyzing recordings of champion drone pilots navigating challenging race courses. Through IRL, the agent can infer the implicit factors that contribute to success in drone racing, such as speed and agility. Applicants are expected to be proficient in Python, C++, and Git.

We want to explore the application of Inverse Reinforcement Learning (IRL) for training RL agents performing drone races or FPV freestyle to develop methods that extract valuable knowledge from the actions and implicit understanding of expert pilots. This knowledge will then be translated into a robust reward function suitable for autonomous drone flights.

Interested candidates should send their CV, transcripts (bachelor and master), and descriptions of relevant projects to: Ismail Geles [geles (at) ifi (dot) uzh (dot) ch], Leonard Bauersfeld [bauersfeld (at) ifi (dot) uzh (dot) ch]

Event cameras offer remarkable advantages, including ultra-high temporal resolution in the microsecond range, immunity to motion blur, and the ability to capture high-speed phenomena (https://youtu.be/AsRKQRWHbVs). These features make event cameras invaluable for applications like autonomous driving. However, efficiently processing the sparse event streams while maintaining low latency remains a difficult challenge. Previous research has focused on developing sparse update frameworks for event-based neural networks to reduce computational complexity, i.e., FLOPs. This project takes the next step by directly lowering the processing runtime to unlock the full potential of event cameras for real-time applications.

The focus of the project is to reduce runtime using common hardware (GPUs), which have been highly optimized for parallelization. The project will explore drastically new processing paradigms, which can potentially be transferred to standard frames. This ambitious project requires a strong sense of curiosity, self-motivation, and a principled approach to tackling research challenges. You should have solid Python programming skills and experience with at least one deep learning framework. If you’re excited about exploring cutting-edge techniques to push the boundaries, please feel free to contact us.

Key Requirement

• Background in Deep Learning: Proficiency in Python and familiarity with state-of-the-art deep learning frameworks.

• Problem-Solving Skills: Ability to approach research problems in a principled way.

Nico Messikommer [nmessi (at) ifi (dot) uzh (dot) ch], Nikola Zubic [zubic (at) ifi (dot) uzh (dot) ch], Prof. Davide Scaramuzza [sdavide (at) ifi (dot) uzh (dot) ch]

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

VLMs excel in reasoning and dynamic code generation, making them ideal for tasks like excavation sequencing, obstacle management, and multi-robot coordination. Applications include dynamic trenching, rock field clearing, and safety monitoring. The goal is to deploy VLM-based systems on autonomous excavators to enhance efficiency and adaptability.

Develop simulated and real scenarios for VLM-driven planning.

Integrate VLMs into excavation control systems for triggering tasks and code generation.

Benchmark performance in complex planning scenarios.

lterenzi@ethz.ch

The evolution of large language models (LLMs) has opened new avenues in automation and planning. Notably, Claude 3.7 Sonnet introduces hybrid reasoning, enabling both rapid responses and detailed, step-by-step problem-solving \cite{anthropic2025}. Such capabilities position these models as potential candidates for tasks requiring intricate planning, such as excavation.Possible final deployment in the real world with our excavator robot. Excavation planning is a challenging problem requiring spatial reasoning, decision-making under constraints, and long-horizon planning. Recent advances in AI have led to agents that can master complex games like Go and navigate automation-heavy environments. This project aims to determine whether these AI systems can efficiently plan excavation tasks and how they compare to reinforcement learning-based approaches.

Terra is a flexible, JAX-accelerated grid-world environment designed for training AI agents in earthworks planning. It allows for high-level motion and excavation planning, formulated as a reinforcement learning (RL) problem. Terra's multi-GPU capabilities enable rapid training, achieving intelligent excavation planning in minutes on high-end hardware.

First, we will test the zero-shot capabilities of state-of-the-art LLMs and agents in excavation planning. By providing structured prompts and game renderings, we will evaluate whether these models can reason effectively about excavation tasks.

\item Design a pipeline that enables modern AI agents and large language models (LLMs) to play the excavation planning game in \textbf{Terra}.

\item Evaluate whether models like Claude 3.7 Sonnet or GPT-4 can solve excavation tasks zero-shot or require fine-tuning.

\item Train AI models with reinforcement learning using Terra’s multi-GPU acceleration.

\item Deploy the trained models onto a real-world robotic excavator for autonomous excavation.

• general programming experience with python
• experience training neural network
• bonus: experience with large language models

lterenzi@ethz.ch

Transcatheter procedures avoid the trauma and risks of open-heart surgery by delivering devices that are intended to replicate surgical repair and replacement. We are creating novel devices and tools for both valve repair and replacement. These projects require innovative design and creative problem-solving skills along with expertise in prototyping and experimental evaluation.

This Master thesis is conducted at the Harvard Medical School, Boston.

Applicants should inquire by email with Prof. Pierre Dupont (Pierre.Dupont@childrens.harvard.edu) with a description of their qualifications, background, and availability.

We are developing robotic catheters for heart valve repair and for treatment of arrythmias. Robotics offers the advantage of reducing the learning curve for complex beating-heart procedures and provides a platform for introducing automation. Important components of these projects can include: (1) developing and implementing autonomous control strategies, (2) integration of therapeutic devices, and (3) testing in anatomical and animal models.

This Master thesis is conducted at the Harvard Medical School, Boston.

Applicants should inquire by email with Prof. Pierre Dupont (Pierre.Dupont@childrens.harvard.edu) with a description of their qualifications, background, and availability.

Our project focuses on optimizing the ultrasound-assisted bioprinting process by leveraging real-time feedback and image processing. Ultrasound propagates through different materials, creating high and low-pressure nodes that can form various patterns within a printed structure. However, the process is challenged by ultrasound reflection and scattering.

To address these challenges, we have developed a transparent nozzle that allows real-time observation using multiple cameras. This setup enables the integration of advanced image processing techniques, such as template matching, to accurately print specific patterns. Additionally, the system is fully automated, with a function generator for wave creation and cooling elements.

The core of the project involves analyzing the printing process and acoustic cell patterning using computer vision. By incorporating real-time feedback from sensors, we can optimize parameters on the fly, ensuring precise and consistent pattern formation. This approach is tailored for bioapplications, where maintaining specific conditions is crucial for success

Utilize the developed transparent nozzle equipped with multiple cameras for real-time observation of the ultrasound-assisted bioprinting process.

Integrate advanced image processing techniques, such as template matching, to accurately print specific patterns.

Implement real-time feedback from sensors to dynamically optimize parameters, ensuring precise and consistent pattern formation.

Apply these optimized techniques to achieve reliable and efficient bioprinting for bio applications.

Experience with coding in Python is necessary, and knowledge of machine learning, control, 3D printing, and fluid dynamics is desirable.

Please send your CV and transcript of records to Prajwal Agrawal at pprajwal@ethz.ch, Mahmoud Medany at mmedany@ethz.ch, and Professor Daniel Ahmed at dahmed@ethz.ch.

Autonomous Driving made tremendous progress in recent years through innovations in learning-based methods [1]. An emergent enabler are Birds-Eye-View methods that allow vehicles to understand and reason about their surroundings in real time. In this project, we aim to transfer this research to autonomous mobile robots in real-world, human-inhabited environments. While rules of navigation and traversability are well defined for autonomous driving, one exciting aspect of this project is on finding analogous representations for everyday human environments. We would like to explore these methods on a range of robots including Spot, Anymal, the Ultra Mobility Vehicle and Humanoids.

References [1] Liao, B., Chen, S., Zhang, Y., Jiang, B., Zhang, Q., Liu, W., ... & Wang, X. (2024). Maptrv2: An end-to-end framework for online vectorized hd map construction. International Journal of Computer Vision, 1-23. [2] Kim, Y., Lee, J. H., Lee, C., Mun, J., Youm, D., Park, J., & Hwangbo, J. (2024). Learning Semantic Traversability with Egocentric Video and Automated Annotation Strategy. arXiv preprint arXiv:2406.02989. [3] https://rpl-cs-ucl.github.io/STEPP/

This project is hosted at The AI institute in collaboration with RSL.

• Research latest BEV methods
• Develop BEV-inspired methods for mobile robotic semantic traversability.
• Deployment on real robots

• Excellent knowledge of C++, Python
• Familiarity with learning framework, e.g. pytorch
• Experience with ROS2 is a plus

Email Abel (agawel@theaiinstitute.com) and Laurent (lkneip@theaiinstitute.com). Please include your CV and up-​to-date transcript.

Human environments often adhere to implicit and explicit semantic structures that are easily understood by humans. For Autonomous mobile robots to act in these environments, we aim to investigate how to represent this understanding of the environment. One technology that gained popularity in recent years are scene graphs. These allow robots to spatially deconstruct the world in a graph of multiple levels of abstraction, where nodes represent places, rooms, objects, etc and edges the relationships between them. In this project, we aim to elevate semantic scene graphs to a level to perform semantically-guided navigation and interaction. We would like to explore these methods on a range of robots including Spot, Anymal, the Ultra Mobility Vehicle and Humanoids.

References

[1] Hughes, N., Chang, Y., & Carlone, L. (2022). Hydra: A real-time spatial perception system for 3D scene graph construction and optimization. arXiv preprint arXiv:2201.13360. [2] Honerkamp, D., Büchner, M., Despinoy, F., Welschehold, T., & Valada, A. (2024). Language-grounded dynamic scene graphs for interactive object search with mobile manipulation. IEEE Robotics and Automation Letters. [3] Gu, Q., Kuwajerwala, A., Morin, S., Jatavallabhula, K. M., Sen, B., Agarwal, A., ... & Paull, L. (2024, May). Conceptgraphs: Open-vocabulary 3d scene graphs for perception and planning. In 2024 IEEE International Conference on Robotics and Automation (ICRA) (pp. 5021-5028). IEEE.

This thesis will be hosted at The AI Institute in collaboration with RSL.

• Familiarization with latest scene graph frameworks
• Build new scene graph representations that seamlessly integrate with robot navigation
• Deployment on real robots

• Excellent knowledge of C++, Python
• Familiarity with learning framework, e.g. pytorch
• Experience with ROS2 is a plus

Email Abel (agawel@theaiinstitute.com) and Alex (aliniger@theaiinstitute.com). Please include your CV and up-​to-date transcript.

The Autonomous River Cleanup (ARC) is a student-led initiative supported by the Robotic Systems Lab with the goal of removing riverine waste. By joining ARC, you will contribute to ARC's most recent project, SARA. Project SARA is the next iteration of our bridge-mounted camera system to detect waste in rivers and measure pollution. What is new is that we will use a smartphone as the core component of the monitoring system. Today's smartphones are compact, relatively inexpensive, and equipped with a good camera and high computing power. One of the main challenges is the sizing of the external battery and solar panel, and determining the optimal runtime of the system throughout the day. Since it is impossible to run the system 24/7 without increasing the system's complexity, the goal is to find the most suitable time period for monitoring while keeping the system dimensions to a reasonable size.

During your time at ARC, you will do the following:

• Model the power dynamics of the system based on the system’s power consumption and available solar radiation data

• Determine the ideal battery capacity and solar panel size while ensuring a reasonable trade-off between the system’s mass and runtime for monitoring

• If time allows, validate your simulation with experimentation on the real hardware

Ideally, you already have the following skills or are eager to learn them:

• Basic knowledge of modeling system dynamics and power electronics

• Familiarity with programming (ideally Python) for simulating your model

• Structured and methodical way of working

We look forward to hearing from you! For the application, please specify why you are interested in the project (motivational statement) and include your CV and Transcript of Records. Please reach us via our arc@ethz.ch.

Supervisors:

• Emre Elbir, eelbir@ethz.ch

• Jonas Stolle, jstolle@ethz.ch

The Autonomous River Cleanup (ARC) is a student-led initiative supported by the Robotic Systems Lab with the goal of removing riverine waste. By joining ARC, you will contribute to ARC's most recent project, SARA. Project SARA is the next iteration of our bridge-mounted camera system to detect waste in rivers and measure pollution. What is new is that we will use a smartphone as the core component of the monitoring system. Today's smartphones are compact, relatively inexpensive, and equipped with a good camera and sufficient computing power. Usually, the baseline model works well on the location it has been trained on but loses performance once deployed at a new location. Therefore, we need to investigate methods to make our algorithms more robust to location changes and require only minimal retraining since obtaining new location data is tedious. Even though modern smartphones have high computational power, they are still limited compared to other hardware intended for machine learning applications. Thus, we have to rely on lightweight models.

During your time at ARC, you will do the following:

• Familiarize yourself with previous work at ARC and in the literature on domain adaptation techniques

• Determine suitable lightweight detection and classification models for a smartphone application

• Analyze different domain adaptation techniques requiring minimal new data for retraining

Ideally, you already have the following skills or are eager to learn them:

• Familiarity with Python, ROS, and version control

• Solid knowledge of computer vision and machine learning

• Structured and methodical working style

We look forward to hearing from you! For the application, please specify why you are interested in the project (motivational statement) and include your CV and Transcript of Records. Please reach us via our arc@ethz.ch.

Supervisors:

• Emre Elbir, eelbir@ethz.ch

• Jonas Stolle, jstolle@ethz.ch

The Autonomous River Cleanup (ARC) is a student-led initiative supported by the Robotic Systems Lab with the goal of removing riverine waste. By joining ARC, you will contribute to ARC's current developments, where we aim to improve our Mobile Autonomous Recycling Container - MARC. MARC is our robotic sorting platform for autonomous waste sorting. It consists of two robot arms which pick up waste items from a moving conveyor belt. During previous work, we realized that the location of the robot base and the number of degrees of freedom (DoF) is not ideal in this confined space. Therefore, we want to find a suitable location for the robot’s base in order to better access the workspace and determine the optimal number of DoF for fast and efficient pick and place movements.

During your time at ARC, you will do the following:

• Familiarize yourself with our current simulation environment to optimize our robot configuration

• Model and simulate different robot arm types in confined spaces for fast pick and place movements

• Determine the optimal number of degree of freedom and robot arm location

Ideally, you already have the following skills or are eager to learn them:

• Familiarity with Python, ROS, and version control

• Basic knowledge about path planning algorithms and robot dynamics

• Experience in working with an existing code framework

We look forward to hearing from you! For the application, please specify why you are interested in the project (motivational statement) and include your CV and Transcript of Records. Please reach us via our arc@ethz.ch.

Supervisors:

• Emre Elbir, eelbir@ethz.ch

• Jonas Stolle, jstolle@ethz.ch

The Autonomous River Cleanup (ARC) is a student-led initiative supported by the Robotic Systems Lab with the goal of removing riverine waste. By joining ARC, you will contribute to ARC's most recent project, SARA. Project SARA is the next iteration of our bridge-mounted camera system to detect waste in rivers and measure pollution. What is new is that we will use a smartphone as the core component of the monitoring system. Today's smartphones are compact, relatively inexpensive, and equipped with a good camera and high computing power. Since we plan to run it throughout the year, the system needs to be reasonably protected against environmental factors. In summer, we face the issue that the smartphone heats itself, and in winter, the smartphone gets too cold, which, in both cases, leads to a shutdown. This thesis aims to assess the thermal influence on the system and derive suitable protective solutions.

During your time at ARC, you will do the following:

• Analyze thermal and environmental factors acting on the system throughout the year

• Determine suitable solutions ensuring adequate heat transfer and environmental protection

• Experimentally test your proposed solutions

Ideally, you already have the following skills or are eager to learn them:

• Basic knowledge of thermodynamics and fluid dynamics

• Familiarity with Python and CAD software

• Interested in working with hardware and physical testing

We look forward to hearing from you! For the application, please specify why you are interested in the project (motivational statement) and include your CV and Transcript of Records. Please reach us via our arc@ethz.ch.

Supervisors:

• Emre Elbir, eelbir@ethz.ch

• Jonas Stolle, jstolle@ethz.ch

This master's project (or thesis) examines the application of reinforcement learning and numerical optimal control to achieve high-performance, agile flight of a flexible quadrotor in confined environments. While high-fidelity models exist for many robotic platforms, their computational demands often limit their use in real-time control scenarios. This project aims to identify and utilize a model of a flexible quadrotor that strikes a balance between accuracy and efficiency.

The approach combines model predictive control with precise numerical integration over a short initial horizon and a simplified, lower-fidelity model for longer-term planning. This hybrid strategy enables long prediction horizons, which are crucial for executing agile maneuvers, such as flying through narrow gaps or navigating tight indoor spaces.

A reinforcement learning policy will be developed using the lab’s high-performance simulators to complement and enhance the control strategy. The project offers the opportunity to work at the intersection of learning, planning, and control, with a strong emphasis on deploying high-speed, intelligent robotics in challenging real-world scenarios. It suits students interested in advanced control, dynamics, reinforcement learning, and robotics.

Applicants should have proficiency in model predictive control, numerical optimization, and reinforcement learning, as well as experience in programming with Python and C++. Initial exposure to NMPC solvers such as acados is expected, as well as familiarity with simulation software and real-time data processing. A solid understanding of drone dynamics and control systems, combined with a background in signal processing and nonlinear dynamic systems.

This project can be taken as a student project or a master's thesis. Student projects can focus on reinforcement learning or model predictive control, while master's theses are required to compare both.

Interested candidates should send their CV, transcripts (bachelor and master), and descriptions of relevant projects to Rudolf Reiter (rreiter AT ifi DOT uzh DOT ch), Angel Romero (roagui AT ifi DOT uzh DOT ch), and Leonard Bauersfeld (bauersfeld AT ifi DOT uzh DOT ch).

This master's project focuses on enabling real-time, vision-based control for quadrotors by distilling large, complex world models into lightweight versions suitable for deployment on resource-constrained platforms. The goal is to achieve fast, efficient inference from camera inputs, supporting agile indoor navigation in previously unseen environments.

Large-scale vision models capable of generating and understanding complex scenes are typically too computationally intensive for onboard use. This project addresses that challenge by applying model distillation techniques to transfer knowledge from a pre-trained, high-capacity model to a smaller, faster one. The distilled models will be deployed on quadrotors to evaluate real-world performance, focusing on latency, energy consumption, and navigation success. Beyond standard RGB input, the project will also investigate using additional visual modalities like depth and semantic segmentation to enhance control capabilities.

The work will follow a structured timeline, starting with a literature review and dataset setup, moving through distillation and model optimization, and ending with deployment and testing. This project is an excellent fit for students interested in robotics, computer vision, and efficient deep learning, and it offers the chance to contribute to the future of responsive, autonomous robotic systems.

Applicant Requirements:

• Proficiency in reinforcement learning, robotics, and computer vision

• Strong programming skills with Python

• First experience with large neural network world models

• Knowledge of simulation software and real-time data processing

• Understanding of drone dynamics and control systems

Investigate model distillation techniques and their application to vision-based world models for deployment on navigation tasks.

Interested candidates should send their CV, transcripts (bachelor and master), and descriptions of relevant projects to Rudolf Reiter (rreiter AT ifi DOT uzh DOT ch) and Daniel Zhai (dzhai (at) ifi (dot) uzh (dot) ch).

This master's project offers an exciting opportunity to work on real-world vision-based drone flight without relying on simulators. The goal is to develop learning algorithms that enable quadrotors to fly autonomously using visual input, learned directly from real-world experience. By avoiding simulation, this approach opens up new possibilities for the future of robotics.

A significant focus of the project is achieving high sample efficiency and designing a robust safety framework that enables effective exploration by leveraging the latest research results. The project will begin with state-based learning as an intermediate step, progressing toward complete vision-based learning. It builds on recent research advances and a well-established drone navigation and control software stack. The lab provides access to multiple vision-capable quadrotors ready for immediate use.

This project is ideal for motivated master’s students interested in robotics, learning systems, and real-world deployment. It offers a rare chance to contribute to a high-impact area at the intersection of machine learning, control, and computer vision, with strong potential for further academic or industrial opportunities.

Applicants should have proficiency in computer vision, reinforcement learning, and robotics, as well as strong programming skills in Python and C++. Initial experience with large neural network world models is expected, as well as familiarity with simulation software and real-time data processing. A solid understanding of drone dynamics and control systems is also essential.

The goal is to investigate how the latest reinforcement learning advances push the limits of learning real-world tasks such as agile vision-based flight.

Interested candidates should send their CV, transcripts (bachelor and master), and descriptions of relevant projects to Rudolf Reiter (rreiter AT ifi DOT uzh DOT ch) and Angel Romero (roagui AT ifi DOT uzh DOT ch)

Drone dynamics can change significantly during flight due to variations in load, battery levels, and environmental factors such as wind conditions. These dynamic changes can adversely affect the drone's performance and stability, making it crucial to develop adaptive control strategies. This research project aims to develop and evaluate a meta model-based reinforcement learning (RL) framework to address these variable dynamics. By integrating dynamic models that account for these variations and employing meta-learning techniques, the proposed method seeks to enhance the adaptability and performance of drones in dynamic environments. The project will involve learning dynamic models for the drone, implementing a meta model-based RL framework, and evaluating its performance in both simulated and real-world scenarios, aiming for improved stability, efficiency, and task performance compared to existing RL approaches and traditional control methods. Successful completion of this project will contribute to the advancement of autonomous drone technology, offering robust and efficient solutions for various applications. Applicants are expected to be proficient in Python, C++, and Git.

Develop methods for meta (model-based) RL to handle variable drone dynamics.

Interested candidates should send their CV, transcripts (bachelor and master), and descriptions of relevant projects to: Ismail Geles [geles (at) ifi (dot) uzh (dot) ch], Elie Aljalbout [aljalbout (at) ifi (dot) uzh (dot) ch], Angel Romero [roagui (at) ifi (dot) uzh (dot) ch]

Optohive is developing a wearable brain imaging system that provides affordable and accessible insights into brain function, offering a quantitative approach to diagnosing and treating neurological disorders.

Our mission is to advance understanding of the human brain and translate this knowledge into innovative diagnostic and therapeutic solutions, addressing the growing prevalence of neurological conditions.

We are currently seeking a Mechanical Engineer to help drive the continued development and refinement of our hardware.

In this internship, you will advance our innovative system by working on industrialization and new developments. This position comes with a strong commercial focus, aligning your technical excellence with market needs.

You will join our interdisciplinary team, taking on key mechanical engineering responsibilities to further develop and refine our optohive hardware: - Design and develop innovative wearables for Optohive. - Prepare components for production, including injection molding processes. - Enhance usability and aesthetics for an improved user experience. - Develop and automate manufacturing workflows for scalability. - Conduct CAD simulations to validate designs and performance. - Set up product requirements.

• MSc. or BSc. in Mechanical Engineering.
• Prior experience in CAD-Design, ideally Solidworks.
• Interest in Working in a dynamic Startup-environment.

dominik@optohive.com

Imagine tiny robots, barely visible to the eye, that can navigate complex environments and perform intricate tasks—all without the bulky brains that bigger robots need! Unlike their larger counterparts, micro- and nanomachines can't pack in heavy computational gear. Instead, they rely on their ingenious designs and smart materials to sense, control, and adapt to their surroundings. In this exciting project, we’re pioneering a microfluidic approach to craft intelligent microrobots with several responsive abilities. The breakthroughs from this research will not only answer key questions in robotics but also propel the use of intelligent micromachines in high-impact areas like sophisticated biomedical devices. Dive into this project with us, and be at the forefront of developing the smart, tiny robots of tomorrow!

The following experience or skills would be ideal but not necessary:

• Knowledge in biomedical engineering.

• Prior experience in chemistry lab.

• Experience or knowledge in microfluidic devices.

References

M. Hu et al. "Self‐Reporting Multiple Microscopic Stresses Through Tunable Microcapsule Arrays." Adv. Mater. 37.3 (2025): 2410945

B. J. Nelson & S. Pané “Delivering drugs with microrobots.” Science 382.6675 (2023): 1120-1122.

• Manipulation of droplet-generation microfluidic systems. (~ 1 month)

• Develop microfabrication process to produce microcapsules from different responsive polymers. (~ 3 months)

• Investigate and test the fabricated intelligent microrobots under different environmental cues. (~ 2 months)

Curious? Please contact minghu@ethz.ch (Dr. Minghan Hu, SNSF Ambizione group leader).

HALVE actuators are soft actuators powered by low voltages, designed for safe and scalable robotic applications. They use dielectric polymers and ultra-thin electrodes fabricated through vapor deposition and blade coating, sealed with a customized CNC technology.

We offer a wide range of topics for thesis and semester projects. Based on your interests, you can focus on material development, actuator design, experimental testing, or application-oriented integration. Projects will be highly hands-on, with access to cleanroom tools, mechanical and electrical testing equipment, and high-voltage electronics.

By the end of your project, you will have developed and tested your own HALVE actuator prototypes—and potentially helped define a new standard in soft actuation.

• Literature review on soft electrostatic actuators and materials
• Design of actuator geometry and electrode/dielectric layer stack
• Fabrication of thin functional films
• Experimental testing (force vs. strain, durability, etc.)
• Performance benchmarking and design iteration
• Application demo with integrated soft robotic system

You ideally have:

• Strong interest in soft robotics, materials, or actuator systems

• Ability to work independently and collaboratively

• Experience in prototyping, fabrication, or electronics is a plus

• Familiarity with CAD, Python, or basic circuit design

• Knowledge of polymer materials and/or electrostatics

• Open to students from Mechanical, Electrical, Materials Engineering, or related fields

The project aims to create neuromuscular constructs with biomimetic innervation. It seeks to enhance tissue regeneration using localized delivery of pro-regenerative molecules. Another goal is to demonstrate the constructs' use in drug testing and bio-hybrid robotics. Scaffolds will be fabricated using electrospun fibers made from polycaprolactone, gelatin, and polypyrrole (Figure 1). Carboxyl-functionalized graphene will enable drug release; ferulic acid will promote nerve regeneration. Aligned fiber sheets will be formed and assembled into blocks or conduits. These will be seeded with iPSC-derived muscle cells and motor neurons. A nerve network will be formed by inserting conduits into muscle blocks. Constructs will be tested for actuation response under electrical stimulation and drug modulation. Multi-unit configurations will be built and analyzed to demonstrate robotic neural control.

Work Packages - Literature review on muscle-based soft actuators - Design of scaffolds for neurotized muscle tissue - 3D biofabrication - Test regenerative responses to the drug principle - Characterization, actuation, and control of the realized tissue

Requirements - High motivation and problem-solving ability - Knowledge of cell culture, biofabrication, tissue engineering, and/or fluorescence microscopy - Must-have: capable of working with a high degree of independence

This project aims at (i) generating functional neuromuscular constructs with biomimetic innervation design; (ii) proving enhanced tissue regeneration (via localized delivery of pro-regenerative molecules); (iii) demonstrating suitability for functional studies (drug testing and bio-hybrid robotics).

Dr. Miriam Filippi, mfilippi@ethz.ch, Soft Robotics Lab, Institute of Robotics and Intelligent Systems, D-MAVT Prof. Robert Katzschmann, rkk@ethz.ch, Soft Robotics Lab, Institute of Robotics and Intelligent Systems, D-MAVT

We build humanoid robotic hands that are dexterous, robust and easy to fabricate. This universal gripper could enable the automation of human-like tasks that are too complex for conventional grippers.

We offer a variety of topics for master’s/bachelor’s thesis or semester projects for students in mechanical engineering, electrical engineering and robotics. Together we will define a focus on specific features for you to independently explore, develop, design and build.

By the end of your project you should have put together a fully functional robotic hand implementing your novel features and systematically analyze its performance experimentally.

• Literature review of existing work
• Modeling and calculation of concepts, structures, mechanisms, etc.
• CAD design, PCB design, prototype fabrication and testing
• System integration, modeling, control
• Improve and iterate on hand design features
• Practical experiments solving various manipulation challenges with the robotic hand

• Enrolled Bachelor’s or Master’s student in Robotics, Mechatronics, Electrical, Mechanical Engineering or a related engineering discipline.
• Proficiency in at least several of the following technical topics: C++, ROS, ROS2, PCB design, actuation & controls, sensing, CAD design, FEM simulations, 3D printing, silicone casting
• First-hand experience with prototyping, actuators and sensors
• Excellent academic track record
• High motivation and problem-solving ability
• Capable of both working independently and collaborating in a team
• Fluent English speaker

Stefan Weirich: stefan.weirich@mimicrobotics.com

Event-based cameras offer significant benefits in difficult robotic scenarios characterized by high-dynamic range and rapid motion. These are precisely the challenges faced by spacecraft during landings on celestial bodies like Mars or the Moon, where sudden light changes, fast dynamics relative to the surface, and the need for quick reaction times can overwhelm vision-based navigation systems relying on standard cameras. In this work, we aim to design novel spacecraft navigation methods for the descent and landing phases, exploiting the power efficiency and sparsity of event cameras. Particular effort will be dedicated to developing a lightweight frontend, utilizing asynchronous convolutional and graph neural networks to effectively harness the sparsity of event data, ensuring efficient and reliable processing during these critical phases. The project is in collaboration with European Space Agency at the European Space Research and Technology Centre (ESTEC) in Noordwijk (NL).

Investigate the use of asynchronous neural networks (either regular or spiking) for building an efficient frontend system capable of processing event-based data in real-time. Experiments will be conducted both pre-recorded dataset as well as on data collected during the project. We look for students with strong programming (Pyhton/Matlab) and computer vision backgrounds. Additionally, knowledge in machine learning frameworks (pytorch, tensorflow) is required.

Interested candidates should send their CV, transcripts (bachelor and master) to: Roberto Pellerito (rpellerito@ifi.uzh.ch), Marco Cannici (cannici@ifi.uzh.ch) and Davide Scaramuzza (sdavide@ifi.uzh.ch).

Traditional facial motion capture systems often rely on marker-based methods or multi-camera rigs to track facial movements. However, these approaches can be limited in capturing fine details such as subtle wrinkles and micro-expressions. Recent advancements in learning-based techniques have enabled high-fidelity facial tracking using monocular RGB images, but the temporal resolution is constrained by the frame rate of conventional cameras. Event-based cameras offer a promising alternative, providing superior temporal resolution without the need for costly and bulky high-speed RGB cameras. This project aims to leverage the advantages of event-based cameras to achieve unprecedented quality in tracking subtle facial movements.

Develop a facial motion capture system that utilizes event-based cameras to accurately track fine facial movements, including micro-expressions and subtle wrinkles. The system should overcome the limitations of traditional methods by providing higher temporal resolution and capturing intricate facial details.

Interested candidates should send their CV, transcripts (bachelor and master) to: Roberto Pellerito (rpellerito@ifi.uzh.ch), Nico Messikommer (nmessi@ifi.uzh.ch) and Davide Scaramuzza (sdavide@ifi.uzh.ch).

Project background

In recent years, the field of microrobotics has garnered significant attention, particularly for its potential applications in biomedical engineering, such as targeted drug delivery, minimally invasive surgery, and precise medical diagnostics. Traditional microrobot fabrication techniques predominantly rely on top-down methods, such as 3D printing and lithography. While effective, these methods often involve complex, time-consuming processes and face limitations in achieving high precision at the microscale.

Project details

Our approach diverges from these conventional methods by employing a bottom-up fabrication technique, leveraging the principles of self-assembly and droplet manipulation. Specifically, we focus on the innovative use of ferrofluid droplets and magnetic fields to sculpt microrobots with customized shapes. This method allows for greater flexibility and precision in designing microrobots, enabling the creation of complex geometries that would be challenging to achieve with top-down techniques.

The following experience or skills would be ideal but not necessary:

• Know-how in nanoparticles synthesis & self-assembly.

• Prior experience in chemistry lab.

• Prior experience or knowledge in magnetic control systems.

References

M. Hu et al. "Shaping the assembly of superparamagnetic nanoparticles." ACS Nano 13.3 (2019): 3015-3022.

B. J. Nelson & S. Pané “Delivering drugs with microrobots.” Science 382.6675 (2023): 1120-1122.

• Build up a droplet printing fabrication platform towards microrobots fabrications. (~ 1 month)

• Optimize the fabrication process to produce microrobots with tailored structures. (~ 3 months)

• Investigate how different microrobot shapes influence their movement under physiological conditions. (~ 2 months)

Please contact minghu@ethz.ch (Dr. Minghan Hu, SNSF Ambizione group leader).

This project explores and extends the novel "deep state-space models" framework by leveraging their transfer function representations. In contrast to time-domain parameterizations (e.g., S4 layers), transfer function parameterization enables direct computation of the model’s corresponding convolutional kernel via a single Fast Fourier Transform. This is state-free, and in theory, it would maintain constant memory and computational overhead regardless of the state size, therefore offering substantial speed and scalability advantages over existing approaches. Building on these promising theoretical results, this project aims to derive better scaling laws for neuromorphic systems by studying and deploying state-free inference in diverse long-sequence and event-based vision applications.

Implement the transfer function-based state-space model, then comprehensively benchmark its training speed, memory usage, and performance on neuromorphic and event-based vision tasks. Investigate how state-free inference behaves as model size and sequence length grow, deriving empirical or theoretical scaling relationships. Compare this approach with other state-of-the-art methods (e.g., S4, Transformer-based models) in terms of speed, memory footprint, and model accuracy or task performance. Prerequisites include familiarity with basics of LTI systems and linear ODEs, and Python programming language.

Interested candidates should send their CV, transcripts (bachelor and master) to Nikola Zubic (zubic@ifi.uzh.ch), Marco Cannici (cannici@ifi.uzh.ch) and Davide Scaramuzza (sdavide@ifi.uzh.ch).

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions:

• Click link "Open this project..." below.
• Log in to SiROP using your university login or create an account to see the details.

If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This master thesis project will focus on the implementation of an additional degree of freedom for insertion, and the following evaluation of the endoscope. Your task will be to develop a test strategy, using a sensorized brain phantom and perform the evaluation experiments.

Test telemanipulation and contact forces of robotic endoscope for neurosurgery

Sara Lisa Margherita Ettori (PhD candidate): saralisamargherita.ettori@unibas.ch

Co-supervised by Jing Yuan Luo (Mujoco)

• Literature research
• Investigate the bias and variance properties of the PPO value function
• Design and implement a novel algorithm that achieves variance reduction through monte carlo sampling via massive environment parallelism
• Re-implement existing SOTA algorithms as benchmarks
• Bonus: provide theoretical insights to justify your proposed monte carlo method

• Background in Learning
• Excellent knowledge of Python

vklemm@ethz.ch

Besides RPG, this project will be co-supervised by Aurelio Sulser (from Alg. & Opt. group at ETH) and prof. Siddhartha Mishra.

This project explores a novel approach to graph embeddings using electrical flow computations. By leveraging the efficiency of solving systems of linear equations and some properties of electrical flows, we aim to develop a new method for creating low-dimensional representations of graphs. These embeddings have the potential to capture unique structural and dynamic properties of networks. The project will investigate how these electrical flow-based embeddings can be utilized in various downstream tasks such as node classification, link prediction, graph classification and event-based vision tasks.

The primary goal of this project is to design, implement, and evaluate a graph embedding technique based on electrical flow computations. The student will develop algorithms to compute these embeddings efficiently, compare them with existing graph embedding methods, and apply them to real-world network datasets. The project will also explore the effectiveness of these embeddings in downstream machine learning tasks. Applicants should have a strong background in graph theory, linear algebra, and machine learning, as well as proficiency in Python and ideally experience with graph processing libraries like NetworkX or graph-tool.

Interested candidates should send their CV, transcripts (bachelor and master) to Nikola Zubic (zubic@ifi.uzh.ch), Aurelio Sulser (asulser@student.ethz.ch) and Davide Scaramuzza (sdavide@ifi.uzh.ch).

Recent advancements in machine learning have highlighted the potential of Long Sequence Modeling as a powerful approach for handling complex temporal dependencies, positioning it as a compelling alternative to traditional Transformer-based models. In the context of drone racing, where split-second decision-making and precise control are of greatest importance, Long Sequence Modeling can offer significant improvements. These models are adept at capturing intricate state dynamics and handling continuous-time parameters, providing the flexibility to adapt to varying time steps essential for high-speed navigation and obstacle avoidance. This project aims to bridge this gap by investigating the application of Long Sequence Modeling techniques in RL to develop advanced autonomous drone racing systems. The ultimate goal is to improve autonomous drones' performance, reliability, and adaptability in competitive racing scenarios.

Develop a Reinforcement Learning framework based on Long Sequence Modeling tailored for drone racing. Simulate the framework to evaluate its performance in controlled environments. Conduct a comprehensive analysis of the framework’s effectiveness in handling long sequences and dynamic racing scenarios. Ideally, the optimized model should be deployed in real-world drone racing settings to validate its practical applicability and performance.

Interested candidates should send their CV, transcripts (bachelor and master) to Nikola Zubic (zubic@ifi.uzh.ch), Angel Romero Aguilar (roagui@ifi.uzh.ch) and Davide Scaramuzza (sdavide@ifi.uzh.ch).

Processing the sparse and asynchronous data from event-based cameras presents significant challenges. Transformer-based models have achieved remarkable results in sequence modeling tasks, including event-based vision, due to their powerful representation capabilities. Despite their success, their high computational complexity and memory demands make them impractical for deployment on resource-constrained devices typical in real-world applications. Recent advancements in efficient sequence modeling architectures offer promising alternatives that provide competitive performance with significantly reduced computational overhead. Recognizing that Transformers already demonstrate strong performance on event-based vision tasks, we aim to leverage their strengths while addressing efficiency concerns.

Study knowledge transfer techniques to transfer knowledge from complex Transformer models to simpler, more efficient models. Test the developed models on benchmark event-based vision tasks such as object recognition, optical flow estimation, and SLAM.

Interested candidates should send their CV, transcripts (bachelor and master) to Nikola Zubic (zubic@ifi.uzh.ch), Giovanni Cioffi (cioffi@ifi.uzh.ch) and Davide Scaramuzza (sdavide@ifi.uzh.ch).

Work packages

Literature research

Global motion reconstruction from videos.

Learning from reconstructed motion demonstrations with reinforcement learning on a humanoid robot.

Requirements

Strong programming skills in Python

Experience in computer vision and reinforcement learning

Publication

This project will mostly focus on algorithm design and system integration. Promising results will be submitted to machine learning / computer vision / robotics conferences.

Related literature

Yuan, Y., Iqbal, U., Molchanov, P., Kitani, K. and Kautz, J., 2022. Glamr: Global occlusion-aware human mesh recovery with dynamic cameras. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (pp. 11038-11049).

YUAN, Y. and Makoviychuk, V., 2023. Learning physically simulated tennis skills from broadcast videos.

Shin, S., Kim, J., Halilaj, E. and Black, M.J., 2024. Wham: Reconstructing world-grounded humans with accurate 3d motion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 2070-2080).

Peng, X.B., Abbeel, P., Levine, S. and Van de Panne, M., 2018. Deepmimic: Example-guided deep reinforcement learning of physics-based character skills. ACM Transactions On Graphics (TOG), 37(4), pp.1-14.

The objective of this project is to develop a robust system for extracting human motions from video footage and transferring these motions to a humanoid robot using learning from demonstration techniques. The system will be designed to handle the noisy data typically associated with video-based motion extraction and ensure that the humanoid robot can replicate the extracted motions with high fidelity while respecting physical rules.

Proposed Methodology

Video Data Collection and Motion Extraction:

• Collect video footage of soccer player celebrations and other dynamic human activities.

• Starting from existing monocular human pose/motion estimation algorithms to extract 3D motion data from the videos.

• Incorporate physics-based corrections similar to those employed in WHAM to address issues like jitter, foot sliding, and ground penetration in the extracted motion data.

Motion Learning:

• Applying existing learning from demonstration algorithms in a simulated environment to replicate kinematic motions reconstructed from the videos while respecting physical rules using reinforcement learning.

Implementation on Humanoid Robot:

• This is encouraged since we have our robot lying there waiting for you.

Please include your CV and transcript in the submission.

Manuel Kaufmann

https://ait.ethz.ch/people/kamanuel

kamanuel@inf.ethz.ch

Chenhao Li

https://breadli428.github.io/

chenhli@ethz.ch

Work packages

Literature research

Utilize simulation platforms (e.g., Isaac Lab) for initial policy development and training.

Explore model-free RL approaches, potentially incorporating curriculum learning to gradually increase task complexity.

Investigate perception models for object detection and trajectory forecasting, possibly leveraging lightweight deep learning architectures for real-time processing.

Implement and test learned behaviors on a physical humanoid robot, addressing the challenges of sim-to-real transfer through domain randomization or fine-tuning.

Requirements

Solid foundation in robotics, control theory, and machine learning.

Experience with reinforcement learning frameworks (e.g., PyTorch, TensorFlow, or RLlib).

Familiarity with robot simulation environments (e.g., MuJoCo, Gazebo) and real-world robot control.

Strong programming skills (Python, C++) and experience with sensor data processing.

Publication

This project will mostly focus on algorithm design and system integration. Promising results will be submitted to machine learning / robotics conferences.

Perception & Prediction

• Develop a real-time perception pipeline capable of detecting and tracking incoming projectiles. Utilize camera data or external motion capture systems to predict ball trajectories accurately under varying speeds and angles.

Reactive Motion Planning

• Design algorithms that plan evasive maneuvers (e.g., side-steps, ducks, or rotational movements) within milliseconds of detecting an incoming threat, ensuring the robot’s center of mass remains stable throughout.

Learning-Based Control

• Apply reinforcement learning or imitation learning to optimize dodge behaviors, balancing between minimal energy expenditure and maximum evasive success. Investigate policy architectures that enable rapid reactions while handling noisy observations and sensor delays.

Robustness & Evaluation

• Test the system under diverse scenarios, including multi-ball environments and varying throw speeds. Evaluate the robot’s success rate, energy efficiency, and post-dodge recovery capabilities.

Implementation on Humanoid Robot:

• This is encouraged since we have our robot lying there waiting for you.

Please include your CV and transcript in the submission.

Chenhao Li

https://breadli428.github.io/

chenhli@ethz.ch

Work packages

Literature research

Human motion capture and retargeting

Skill space development

Hardware validation encouraged upon availability

Requirements

Strong programming skills in Python

Experience in reinforcement learning and imitation learning frameworks

Publication

This project will mostly focus on algorithm design and system integration. Promising results will be submitted to robotics or machine learning conferences where outstanding robotic performances are highlighted.

Related literature

Peng, Xue Bin, et al. "Deepmimic: Example-guided deep reinforcement learning of physics-based character skills." ACM Transactions On Graphics (TOG) 37.4 (2018): 1-14.

Starke, Sebastian, et al. "Deepphase: Periodic autoencoders for learning motion phase manifolds." ACM Transactions on Graphics (TOG) 41.4 (2022): 1-13.

Li, Chenhao, et al. "FLD: Fourier latent dynamics for Structured Motion Representation and Learning."

Serifi, A., Grandia, R., Knoop, E., Gross, M. and Bächer, M., 2024, December. Vmp: Versatile motion priors for robustly tracking motion on physical characters. In Computer Graphics Forum (Vol. 43, No. 8, p. e15175).

Fu, Z., Zhao, Q., Wu, Q., Wetzstein, G. and Finn, C., 2024. Humanplus: Humanoid shadowing and imitation from humans. arXiv preprint arXiv:2406.10454.

He, T., Luo, Z., Xiao, W., Zhang, C., Kitani, K., Liu, C. and Shi, G., 2024, October. Learning human-to-humanoid real-time whole-body teleoperation. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (pp. 8944-8951). IEEE.

He, T., Luo, Z., He, X., Xiao, W., Zhang, C., Zhang, W., Kitani, K., Liu, C. and Shi, G., 2024. Omnih2o: Universal and dexterous human-to-humanoid whole-body teleoperation and learning. arXiv preprint arXiv:2406.08858.

Please include your CV and transcript in the submission.

Chenhao Li

https://breadli428.github.io/

chenhli@ethz.ch

Acoustic standing waves have been widely used to control and manipulate particles in liquid environments, but their potential in gas and air-based systems remains underexplored. By tuning acoustic parameters, it is possible to create virtual tweezers and particle streams, enabling selective manipulation.

This project aims to:

Investigate the effect of acoustic standing waves on particle transport in air. Characterize particle selectivity based on size and weight. Develop a compact driver circuit for airborne acoustic manipulation.

The student will engage in both theoretical modeling and experimental validation, collaborating with Honeywell engineers on electronic circuit design, hardware prototyping, and test setups. A successful outcome could lead to an extended R&D phase or a potential PhD project to further develop and implement these findings.

The project includes the following tasks:

Developing simulation models to analyze the effect of acoustic standing waves on particle transport in air. Characterizing particle selectivity for different sizes, weights, and flow parameters. Building a proof-of-concept (POC) demonstrator showcasing the ability to manipulate at least one type of particle. Estimating the driving circuit power envelope and cost for practical implementation.

Please send your CV and transcript of records to: Mahmoud Medany – mmedany@ethz.ch and Prof. Daniel Ahmed - dahmed@ethz.ch

Work packages

Design a Loosely Guided RL Framework that integrates simple reference trajectories into the training loop.

Evaluate Exploration Efficiency by comparing baseline RL methods with the guided approach.

Demonstrate Complex Parkour Behaviors such as climbing, jumping, and dynamic traversal using the guided RL policy.

Hardware validation encouraged

Requirements

Strong programming skills in Python

Experience in reinforcement learning and imitation learning frameworks

Publication

This project will mostly focus on algorithm design and system integration. Promising results will be submitted to robotics or machine learning conferences where outstanding robotic performances are highlighted.

Related literature

Peng, Xue Bin, et al. "Deepmimic: Example-guided deep reinforcement learning of physics-based character skills." ACM Transactions On Graphics (TOG) 37.4 (2018): 1-14.

Li, C., Vlastelica, M., Blaes, S., Frey, J., Grimminger, F. and Martius, G., 2023, March. Learning agile skills via adversarial imitation of rough partial demonstrations. In Conference on Robot Learning (pp. 342-352). PMLR.

Li, Chenhao, et al. "FLD: Fourier latent dynamics for Structured Motion Representation and Learning."

Serifi, A., Grandia, R., Knoop, E., Gross, M. and Bächer, M., 2024, December. Vmp: Versatile motion priors for robustly tracking motion on physical characters. In Computer Graphics Forum (Vol. 43, No. 8, p. e15175).

Fu, Z., Zhao, Q., Wu, Q., Wetzstein, G. and Finn, C., 2024. Humanplus: Humanoid shadowing and imitation from humans. arXiv preprint arXiv:2406.10454.

He, T., Luo, Z., Xiao, W., Zhang, C., Kitani, K., Liu, C. and Shi, G., 2024, October. Learning human-to-humanoid real-time whole-body teleoperation. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (pp. 8944-8951). IEEE.

He, T., Luo, Z., He, X., Xiao, W., Zhang, C., Zhang, W., Kitani, K., Liu, C. and Shi, G., 2024. Omnih2o: Universal and dexterous human-to-humanoid whole-body teleoperation and learning. arXiv preprint arXiv:2406.08858.

Please include your CV and transcript in the submission.

Chenhao Li

https://breadli428.github.io/

chenhli@ethz.ch

piezo elements have been used for various particle manipulation applications, from humidifiers to large-scale cooling solutions. However, low temperatures and environmental factors can complicate the vaporization process, affecting performance and efficiency. Additionally, piezo elements generate heat, which influences the size and distribution of atomized particles.

This project aims to:

Investigate the effect of piezoelectric elements on liquid and gel atomization under different conditions (viscosity, temperature, environmental factors). Develop a proof-of-concept (POC) demonstrator to validate the feasibility of controlled atomization.

Optimize the driving circuit with a focus on low power consumption and efficiency.

Evaluate system parameters, ensuring repeatability and calibration of particle size and flux.

The project will combine theoretical modeling, circuit design, and experimental validation, providing new insights into piezo-based atomization technologies.

The student will engage in both theoretical modeling and experimental validation, collaborating with Honeywell engineers on electronic circuit design, hardware prototyping, and test setups. A successful outcome could lead to an extended R&D phase or a potential PhD project to further develop and implement these findings.

Deploying a proof-of-concept (TRL3) demonstrator to showcase the physical principles of piezo-driven atomization.

Optimizing the driving circuit to minimize power consumption while maintaining effective atomization.

Evaluating the atomization process, focusing on repeatability, particle size control, and flux calibration.

Please send your CV and transcript of records to: Mahmoud Medany – mmedany@ethz.ch and Prof. Daniel Ahmed - dahmed@ethz.ch