Project Database

This project is intended as a summer project only.

In this project, the student will work closely with other team members to develop data collection pipelines for experiments with an amphibious robot, and use these pipelines to collect and analyze experimental data. Specifically, the student will:

• Use a 4-camera system to track the 3D kinematics of the robot, ideally in real time. This includes calibrating camera intrinsics and extrinsics, labeling data and training a DeepLabCut network or applying marker-tracking libraries for offline tracking, and implementing real-time tracking nodes in ROS 2.
• Synchronize data collected from multiple sources, including cameras, force sensors, motors, and other onboard sensors.
• Visualize the collected data in MuJoCo viewer, RViz, Blender, or other 3D visualization tools.
• Assist with experimental data collection.

The student is expected to be familiar with programming in C/C++ and Python, ROS 2, and robot kinematics. Experience with Docker, Linux kernel development, communication protocols, and computer vision algorithms would be a plus.

Students interested in this project should send the following materials to the project assistant: (1) a resume, (2) a transcript showing relevant courses and grades, and (3) any additional materials that demonstrate their skills and project experience, such as videos, slides, code repositories, or previous project reports.

In nature, animals have many compliant structures that benefit their locomotion. For example, compliant foot/leg structures help adapt to uneven terrain or negotiate obstacles, flexible tails allow efficient undulatory swimming, and muscle-tendon structures help absorb shock and reduce energy loss. Similar compliant structures may benefit salamander-inspired robots as well.

In this study, the student will try simulating compliant structures (the feet of the robot) in Mujoco and optimizing the design. To bridge the sim-to-real gap, the student will first work with other lab members to perform experiments to measure the mechanical properties of a few simple compliant structures. Then, the student needs to simulate these experiments using the flexcomp plugin of Mujoco or theoretical solid mechanics models, and tune the simulation models to match the dynamical response in simulation with the experiments. Afterward, the student needs to optimize the design parameters of the compliant structures in simulation to improve the locomotion performance of the robot while maintaining a small sim-to-real gap. Finally, prototypes of the optimal design will be tested on the physical robot to verify the results.

The student is thus required to be familiar with Python programming, physics engines (preferably Mujoco), and optimization/learning algorithms. The student should also have basic mechanical design abilities to design mechanical structures and perform experiments. Students who have taken the Computational Motor Control course or have experience with data-driven design and solid mechanics would also be preferred.

The student who is interested in this project shall send the following materials to the assistants: (1) resume, (2) transcript showing relevant courses and grades, and (3) other materials that can demonstrate your skills and project experience (such as videos, slides, Git repositories, etc.).

We offer an exciting opportunity for a highly motivated graduate student in Mechanical Engineering to undertake a thesis project focusing on designing and constructing a robotic apparatus to test and validate the impact force of a dinosaur tail strike. This project spans approximately 6 months and requires a combination of mechanical design expertise, force plate measurements, innovation in biomimetic structures, and proficiency in data analysis.

Project Description

The thesis project revolves around designing, building, and controlling a life-sized robotic tail capable of replicating the striking force of a dinosaur’s club-shaped tail. The aim is to accurately measure impact force and velocity using a bone-like material reproduction sourced from fossils we have at the Palaeontological Institute and Museum of the University of Zurich. This endeavor will involve close collaboration with a multidisciplinary team and conducting experiments at our facilities at Empa Dübendorf by Zurich.

Responsibilities

• Utilize mechanical design skills (3D modeling) and motion control (microcontroller designing and programming) to create a functional life-sized Glyptodont's tail.
• Conduct tests to measure impact force and velocity, meticulously documenting experimental procedures and results.
• Employ data analysis techniques, including statistical tools or software, to interpret experimental findings.
• Demonstrate creativity in problem-solving, proposing enhancements to the biomimetic tail design where necessary.
• Collaborate effectively within a team, communicating ideas and contributing to the project's success.

Requirements

Background in mechanical designing with proficiency in 3D modeling.

Expertise in motion control, including microcontroller designing and programming.

Ability to collect, analyze, and interpret experimental data using statistical tools or software.

Strong problem-solving skills with a demonstrated ability to innovate in design and testing.

Excellent communication skills to collaborate within a team and articulate ideas effectively.

Expected Outcomes

Successful creation of a fully functional life-sized Glyptodont's tail within the thesis duration.

Execution of tests to accurately measure impact force and velocity.

Comprehensive documentation of experiments and results.

Recommendations for potential enhancements or modifications based on findings.

If you are a Master's student passionate about pushing the boundaries of robotics, biomimicry, and mechanical engineering and are looking for an engaging thesis project, we encourage you to apply. Please submit your resume/CV along with a cover letter detailing your relevant experience and why you are excited about this exceptional thesis opportunity to Auke Ijspeert as well as Ardian Jusufi.

INTRODUCTION

Egocentric wearable sensing is becoming a key enabler for embodied AI and robotics. Meta’s Project Aria research glasses provide rich, multimodal, first-person observations, including RGB video, scene cameras, IMUs, microphones, eye-tracking, and more, in a socially acceptable, all-day wearable form factor. They are specifically designed to advance egocentric AI, robotics, and contextual perception research.

In collaboration with Meta, we aim to tightly couple Aria research glasses with our existing manipulation platforms at EPFL. This project will:

1. Integrate the Aria Research Kit, including Aria Gen 2, with our ROS2-based robot platforms, including ViperX 300S and WidowX-250 arms on a mobile base. This includes calibration and time-synchronisation with RGB-D cameras and robot state.
2. Design and execute egocentric data collection in household-like environments, combining Aria, RealSense cameras, robot joint trajectories, and language annotations.
3. Develop egocentric data-driven policies for robotic manipulation through learning from demonstration. The robot can be a robot arm or a humanoid robot such as Reaman.
4. Explore one or more robotics applications powered by Aria signals, such as intention-aware teleoperation, egocentric demonstrations for policy learning, or vision-language-action fine-tuning for assistance tasks.
5. Perform systematic platform testing, validation, and documentation to deliver a reusable research pipeline for future projects. Excellent programming skills in Python are a plus.

IMPORTANCE

We have well-documented tutorials on using the robots, teleoperation interfaces for data collection, using the HPC cluster, and a complete pipeline for training robot policies. The Aria Research Kit and tools, including recording, calibration, dataset tooling, and SDK, will be integrated into this ecosystem. This allows the student to focus on the research questions rather than low-level setup. We already have a good base and results from the ongoing project.

WHAT MAKES ARIA SPECIAL FOR ROBOTICS?

Project Aria glasses are multi-sensor research smart glasses. They include multiple cameras with wide field of view, IMUs, microphones, eye gaze, and a Machine Perception Service that provides SLAM poses, hand poses, and related perception outputs. They are explicitly positioned by Meta as a research kit for contextual AI and robotics, especially for using egocentric sensing to build embodied agents that understand and act in the world.

Compared to a normal RGB-D camera, Aria provides:

• Egocentric view: what the human or robot sees while acting.
• Calibrated head pose and trajectory through SLAM in MPS.
• Hand and gaze information, depending on which parts of the system are used.
• A portable, wearable, and socially acceptable form factor.

WHAT WE HAVE

1. Ready-and-easy-to-use robot platforms: ViperX 300S and WidowX-250 arms, configured with 4 RealSense D405 cameras, various grippers, and a mobile robot platform.
2. Egocentric sensing hardware: Meta Project Aria research glasses, including Aria Gen 2, with access to the Aria Research Kit and tooling for data recording and processing.
3. Computing resources: two desktop PCs with NVIDIA GPUs 5090 and 4090.

CANDIDATES

Interested students can apply by emailing sichao.liu@epfl.ch or lixuan.tang@epfl.ch.

Please attach your transcript and a short description of your past/current experience on related topics, such as robotics, computer vision, machine learning, AR, or egocentric perception.

The position is open until we have final candidates. Otherwise, the position will be closed.

RECOMMENDED READING

1. Kareer, Simar, et al. “EgoMimic: Scaling Imitation Learning via Egocentric Video.” 2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2025.
2. Liu, Vincent, et al. “EgoZero: Robot Learning from Smart Glasses.” arXiv preprint arXiv:2505.20290, 2025.
3. Punamiya, Ryan, et al. “EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World.” arXiv preprint arXiv:2604.07607, 2026.
4. Saroha, Abhishek, et al. “EgoFlow: Gradient-Guided Flow Matching for Egocentric 6DoF Object Motion Generation.” arXiv preprint arXiv:2604.01421, 2026.
5. Zheng, Ruijie, et al. “EgoScale: Scaling Dexterous Manipulation with Diverse Egocentric Human Data.” arXiv preprint arXiv:2602.16710, 2026.
6. Aria Project: https://www.projectaria.com/resources/
7. Aria GitHub: https://github.com/facebookresearch/projectaria_tools