Job
- Level
- Junior
- Job Field
- Software, Data
- Employment Type
- Full Time
- Contract Type
- Internship / school internship
- Location
- Böblingen
- Working Model
- Hybrid, Onsite
Job Summary
In this Master's thesis, you will focus on developing powerful robot learning systems using teleoperation and demonstration data, creating hand-retargeting pipelines and data collection tools for complex manipulation tasks.
Job Technologies
Your role in the team
- In our cross-functional team AI Research - Physical AI, we explore the latest technologies for AI-powered robotics architectures and modern AI stacks.
- A central research area is learning complex manipulation skills from human demonstrations.
- In this master's thesis, you investigate how teleoperation, demonstration data, and vision-language-action models can be used to develop powerful robot learning systems.
- Dexterous robotic hands enable the execution of complex manipulation tasks and are considered an important application area of modern robot learning methods.
- Current Learning-from-Demonstration and Vision-Language-Action approaches promise an efficient transfer of human skills to robotic systems.
- The aim of the work is the scientific investigation of the relationship between demonstration data, teleoperation procedures, and the performance of modern robot learning approaches.
- For this purpose, demonstration data are collected using a dexterous robotic hand, various learning methods are trained, and their generalization capabilities are systematically evaluated under different experimental conditions.
- The work aims to provide new scientific insights into the relationship between teleoperation, demonstration data quality, and the performance of modern robot learning methods.
- These challenges are coming your way:
- Literature review on dexterous teleoperation, hand retargeting, imitation learning, as well as modern vision-language-action models (e.g., ACT, Diffusion Policy, OpenVLA, or π0).
- Development or expansion of a teleoperation platform for controlling a five-finger robotic hand.
- Development of a hand retargeting pipeline to transfer human hand movements to the kinematics of a dexterous robotic hand while preserving the manipulation intent.
- Definition of representative manipulation tasks as well as the development of a synchronized data acquisition pipeline for integrating multi-camera image data, proprioceptive signals, and robot actions.
- Collection, preprocessing, and analysis of a demonstration dataset for Learning-from-Demonstration approaches.
- Training and benchmarking of imitation learning baselines as well as fine-tuning of modern vision-language-action models on the collected data.
- Planning and conducting controlled experiments to investigate the impact of demonstration quality, data volume, dataset diversity, and sensor technology on model performance.
- Scientific evaluation of the resulting strategies on a real robot platform regarding success rate, robustness, and generalization to unknown objects, object positions, and manipulation scenarios.
- Analysis of the results and derivation of scientifically grounded recommendations for efficient Learning-from-Demonstration systems.
This text has been machine translated. Show original
Our expectations of you
Education
- Ongoing master's degree in computer science, robotics, artificial intelligence, electrical engineering, or a comparable program.
Qualifications
- Interest in Robot Learning, Computer Vision, and Embodied AI.
- Independent, structured, and scientific working style.
- Proficient in spoken and written German and English.
- Commitment and team spirit.
Experience
- Good Python skills as well as initial experience with Machine Learning or Robotics.
- Ideally, practical experience with Deep Learning frameworks (e.g., PyTorch) and Linux.
This text has been machine translated. Show original
What we offer
- The activity can commence from October 2026.
This text has been machine translated. Show original
Benefits
More net
- 🏝️Summer and Christmas Bonus
- 🍰Employee Stock Option
- 💻Company Notebook for Private Use
- 🚙Company Car
- 📱Company Phone for Private Use
- 🛍Employee Discount
- 👴🏻Company Retirement Provision
- 👷♂️Additional Insurance
Health, Fitness & Fun
Work-Life-Integration
- 🐕Animals Welcome
- 🏠Home Office
- 🍼Day Care for Kids
- 🅿️Employee Parking Space
- 🚌Excellent Traffic Connections
- ⏰Flexible Working Hours
Food & Drink
Topics that you deal with on the job
Job Locations
This is your employer
Mercedes - Benz AG
The Mercedes-Benz brand of automobiles is a registered trademark of Daimler AG. In 2016, sales of new vehicles reached 2.08 million. With business units including Mercedes-Benz Cars, Daimler Trucks, Mercedes-Benz Vans, Daimler Buses and Daimler Mobility, the company ranks among the leading providers of premium cars and is the world's largest commercial vehicle manufacturer.
Description
- Founding year
- 1926
- Company Type
- Established Company
- Working Model
- Hybrid, Onsite
- Industry
- Vehicle Manufacturing, Supplier, Industry, Production
Employer reviews
by devworkplaces.com
Total
(1 Review)4.4
Workingconditions
4.8Engineering
4.0Career Growth
4.8Culture
4.2
Student for Master's thesis: Learning Dexterous Robot Manipulation
- Location
- Böblingen
- Working Model
- Hybrid, Onsite
- Diversity
- Open for all genders
