Available Theses

Open-Source Physical AI for Bimanual Contact-Rich Tasks

The aim of this work is to advance fundamental problems in physical AI, including coordinated bimanual control, representation and exploitation of contact interactions, and sim-to-real transfer. The project involves commissioning two OpenArm robotic arms in a leader–follower teleoperation setup, followed by the development and evaluation of learning-based bimanual control strategies. You will collect and generate manipulation data both through teleoperation and in simulated environments, and train coordinated bimanual control policies for contact-rich tasks using NVIDIA Isaac Lab or related frameworks. The project provides substantial freedom, allowing you to explore ideas that align with your personal interests. Foreseen applications include dexterous manipulation in manufacturing, household environments, or assistive robotics. Contributions to the OpenArm project are strongly encouraged.

Keywords

AI, Machine Learning, Robotics

Labels

Internship , Bachelor Thesis , Master Thesis

The Project

Your Task

You have...

Contact Details

More information

Open this project...

Published since: 2025-12-18 , Earliest start: 2026-01-01 , Latest end: 2026-12-31

Organization pd|z Product Development Group Zurich

Hosts Gimeno Lucas

Topics Engineering and Technology

Imitation learning in Electronics Disassembly

Disassemblage of Batteries is a challenging task with high variability, due to changing part geometries, corrosion, dirt, etc. The goal of this Thesis is to demonstrate an imitation-learning based disassembly of a PC Mainboard by removing all the cabling with a robot-arm.

Keywords

imitation learning, AI, Machine Learning, Control, Robotics, industry

Labels

Internship , Master Thesis

The Project

Your Task

You have...

About pd|z

Facts

Contact Details

More information

Open this project...

Published since: 2025-12-18 , Earliest start: 2025-12-31 , Latest end: 2026-09-30

Organization pd|z Product Development Group Zurich

Hosts Gimeno Lucas

Topics Engineering and Technology

Master Thesis in Learning and Control

We offer a variety of Master Thesis in the topics of Robotics, Learning and Control. Topics include: - imitation learning for cable disassembly - RL-based refinement of imitation learning policies - safe & stable RL and safety filters - encoding high level task primitives with learning policies - 3D and semantic perception of dynamic environments

Keywords

Robotics, Control, Imitation Learning, Reinforcement Learning, Perception, Computer Vision

Labels

Master Thesis

The Project

Your Task

You have...

About pd|z

Contact Details

More information

Open this project...

Published since: 2025-12-11 , Earliest start: 2025-12-31 , Latest end: 2026-09-01

Organization pd|z Product Development Group Zurich

Hosts Gimeno Lucas

Topics Information, Computing and Communication Sciences , Engineering and Technology

Designing a Parametric Software Toolchain for Self-Driving FDM 3D-Printing Laboratories

Self-driving laboratories have shown impressive capabilities in fields such as pharmacy, biology, chemistry, and increasingly in materials discovery, where automated systems execute thousands of experiments with minimal human intervention. Central to these systems is an autonomous design–make–test cycle, in which software and machines interact seamlessly—from parameter selection to fabrication, robotic sample handling, testing, and again parameter optimization. This thesis aims to develop a headless software toolchain that feeds structured and semantic input parameters (high-level geometric parameters) into corresponding parametric CAD models, exports them, configures an automated slicer from the command line, starts print jobs on a real FDM machine, collects test results from a separate testing system, and finally feeds all results into an optimization or decision model that adjusts the input parameters for the next iteration.

Keywords

Design automation, additive manufacturing, 3D-printing, FDM, Slicer, self-driving laboratory (SDL), Prusa XL, FreeCAD, Grasshopper, Closed-loop optimization

Labels

Semester Project , Bachelor Thesis , Master Thesis , ETH Zurich (ETHZ)

Description

Autonomous experimentation is becoming a key methodology across the sciences—but FDM-based self-driving laboratories remain largely unexplored, especially when working with novel materials (e.g., flexible polymers, conductive filaments, composites) for which simulation models are inaccurate or do not exist. Such cases require real physical experiments to understand and optimize mechanical, electrical, or functional properties. To enable this, we need a scripted, fully automated, headless toolchain that can: 1. accept structured input parameters (JSON/CSV), 1. feed them into a parametric 3D model using CAD tools (FreeCAD, Grasshopper, etc.) from the CLI or inside Docker, 1. run a slicer non-interactively to produce G-code (again with parameters from the input), 1. launch print jobs on an FDM system such as the Prusa XL (with multiple toolheads), 1. interface with an external testing system that measures functional properties, 1. and return those results to an optimization model. Self-driving research in additive manufacturing is emerging, and initial demonstrations in the literature show that large-scale autonomous design–make–test workflows can reveal structures and material behaviors that are impossible to predict purely via simulation (see https://www.nature.com/articles/s41467-024-48534-4). Design automation tools have been a focus of our group for some time with examples such as "MIKE - Bicycle Mount Generator - Bike Connector Tool" (see: https://www.printables.com/model/1390943-mike-bicycle-mount-generator-bike-connector-tool). **Flexible Thesis Focus** (to be defined with the student) Depending on your interests and background, the thesis can emphasize: - Design optimization and ML (using an existing dataset), - Developing new parametric models and running your own experiment series (e.g., with conductive filament), - Building robust CAD/slicer automation and headless tool pipelines (reproducibility, Docker, open-source), - Integrating multi-toolhead workflows (e.g., printing + testing with specialized toolheads). **Your Qualification** - Experienced with CAD and parametric design (FreeCAD or Rhino Grasshopper are helpful). - Understanding of FDM printing, slicing (and G-code). - Programming: mainly Python, maybe C embedded firmware. - Interest in automation, tooling, and reproducible workflows. - Able to use systematic approaches to problems. - Motivated and independent. **Location and organization** You will have access to our group's facilities, including a nice shared desk, 3D printers, and an excellent coffee machine. There will be 2+ other ongoing theses within the same domain, and we would welcome exchange and support between students.

Goal

About pd|z

Contact Details

More information

Open this project...

Published since: 2025-12-11 , Earliest start: 2026-02-01 , Latest end: 2026-06-30

Organization pd|z Product Development Group Zurich

Hosts Stein Tom

Topics Engineering and Technology

Multimodal Deep Learning for Gaze-Driven Step Recognition in Orthopedic Training

This project investigates how combining egocentric video with eye-tracking data can improve automated step segmentation in a simulated orthopedic procedure. Using pretrained visual feature extractors and temporal models such as TCNs or Transformers, we aim to predict fine-grained surgical actions and evaluate the added value of gaze information. The goal is to enhance workflow understanding, interpretability, and objective feedback in surgical training.

Keywords

Eye-tracking, Egocentric video, Step segmentation, Orthopedics, Machine Learning, Transformers, Convolutional Neural Networks

Labels

Semester Project , Master Thesis

PLEASE LOG IN TO SEE DESCRIPTION

More information

Open this project...

Published since: 2025-12-11 , Earliest start: 2026-01-05 , Latest end: 2026-12-31

Organization pd|z Product Development Group Zurich

Hosts Hogenkamp Maarten

Topics Medical and Health Sciences , Information, Computing and Communication Sciences , Engineering and Technology

Automated Design and Additive Manufacturing of Bending Jigs for Induction Coil Fabrication

Induction heating is used in soldering applications where fast, efficient and localized heating is necessary. Examples include the repair of photovoltaic modules and soldering of smart textiles. Heat is introduced in the solder joint through electromagnetic induction, where high frequency excitation of a coil is used to induce current in the solder pad. Coil for these applications are often bent from copper tube, to allow for internal water cooling. Coils are often application-specific and therefore manually produced by bending. This process depends on a skilled fabricator, and results are often not reproducible. The idea of this thesis is to use automatically generated bending jigs that can be additively manufactured. A design automation workflow should be developed that generates the necessary jigs based on a given coil geometry without manual intervention. The jigs should then be evaluated by using them to manufacture various coils.

Keywords

Design automation, additive manufacturing, 3D-printing, induction heating, hands-on

Labels

Semester Project , Bachelor Thesis , Master Thesis , ETH Zurich (ETHZ)

About Us

Description

Your Qualification

Contact Details

More information

Open this project...

Published since: 2025-12-01

Organization pd|z Product Development Group Zurich

Hosts Ritter Simon

Topics Engineering and Technology