Student assistant - Reinforcement Learning for Dynamic In-Hand Dexterous Manipulation (Stuttgart, DE, 70569) bei Fraunhofer-Gesellschaft

Advertisement for the field of study such as: computer science, simulation, robotics or comparable.

Robust and repeatable dynamic in-hand manipulation is an essential skill for enabling generalized robotic automation solutions. The focus of this project is to develop a sim-to-real pipeline for training robotic hands for in-hand manipulation of objects. Key parts of this project will include fine-tuning simulation models of robot hands to accurately capture the physics of their real counterparts, using deep reinforcement learning to train policies for in-hand manipulation of various objects in simulation (i.e. Isaac Sim), improving contact and force modeling within the simulation and deploying the learned policies on the real robot. 

The place of work would be in Heilbronn.

What you will do

• Literature Review: Survey reinforcement learning, in-hand manipulation, object grasping and multi-modal (vision, proprioception, and tactile) sensor fusion
• Framework Development: Cell setup in sim and real: build a digital twin in Isaac Sim/Lab for the robot cell; mirror the real setup with matching kinematics, collision geometries and dynamics
• Algorithm Implementation: Develop and deploy reinforcement learning algorithms to train in-hand manipulation policies for a specific robot hand (object reorientation, grasping, etc.). Propose and implement strategy for domain randomization that improves policy transfer to the real robot. Conceptualize and implement solutions for quick deployment of robust policies across a wide range of automation challenges
• Simulation and Testing: Use scripted test suites in Isaac Sim with domain randomization (poses, friction, lighting). Log success rate, cycle time, final grasp stability and feasibility. Evaluate performance on varied objects. Identify failure conditions in sim and real and propose strategies to improve performance

What you bring to the table

• Enrolled student at a German University/Hochschule
• High motivation and initiative
• Very good English language skills
• Programming skills in Python, C++ and C.
• Familiarity with PyTorch
• Experience using IsaacLab, Mujoco or similar physics engines
• Experience in reinforcement learning is a plus
• Independent, responsible and structured working style

What you can expect

• Pleasant working atmosphere
• Interesting and industry-relevant tasks
• A dynamic, interdisciplinary team
• Insights into the future topic of automation
• Good technical equipment

We value and promote the diversity of our employees' skills and therefore welcome all applications - regardless of age, gender, nationality, ethnic and social origin, religion, ideology, disability, sexual orientation and identity. Severely disabled persons are given preference in the event of equal suitability. Remuneration according to the general works agreement for employing assistant staff.

With its focus on developing key technologies that are vital for the future and enabling the commercial utilization of this work by business and industry, Fraunhofer plays a central role in the innovation process. As a pioneer and catalyst for groundbreaking developments and scientific excellence, Fraunhofer helps shape society now and in the future. 

Interested? Apply online now. We look forward to getting to know you!
 

Ms. Jennifer Leppich

Recruiting

+49 711 970-1415

[email protected] 

Fraunhofer Institute for Manufacturing Engineering and Automation IPA 

www.ipa.fraunhofer.de 

Requisition Number: 81788                Application Deadline: