Albert-Ludwigs University Freiburg, Robot Learning Lab.
Last updated: Oct. 17, 2022
Mobile Manipulation and Autonomous Robotics
- Abhinav Valada, Assistant Professor, University Freiburg
- Tim Welschehold, Post-doctoral Researcher, University Freiburg
- Daniel Honerkamp, Researcher, University Freiburg
Mobile manipulation applications are ubiquitous across both industry and services and are a core component in visions such as robotic housekeepers. Progress in this area promises to unlock a much wider area of tasks for robotic automation. The challenges of mobile manipulation include the generation of whole-body motions for large continuous action spaces consisting of both base and arms, a combination of long-horizon task reasoning with short-term acting, mapping from abstract human-specified goals to executable goals, and the simultaneous sensing and exploration of the environment while fulfilling task goals. Learning algorithms struggle to directly learn to generate whole-body motions for mobile manipulators. Within our work we address this problem by decomposing and modularizing it and propose to use kinematic feasibility as a dense reward signal for a reinforcement learning agent that learns the navigation for a mobile base while the robot end-effector follows a task-specified trajectory generated by an independent arbitrary system. We develop a system capable of zero-shot generalization to unseen tasks in unseen environments on real-world robots and show its flexibility in complex, human-centered obstacle environments. Complementarily, we also work on the challenge of exploring indoor environments. To find task-relevant objects in unknown and partially observable environments, we introduce a unified model to balance short-term decision making and long-term exploration.
- Daniel Honerkamp, Tim Welschehold and Abhinav Valada, Learning Kinematic Feasibility for Mobile Manipulation through Deep Reinforcement Learning, EEE Robotics and Automation Letters (RA-L), 2021.
- Daniel Honerkamp, Tim Welschehold and Abhinav Valada, N2M2: Learning Navigation for Arbitrary Mobile Manipulation Motions in Unseen and Dynamic Environments, arXiv preprint arXiv:2206.08737, 2022.
- Fabian Schmalstieg, Daniel Honerkamp, Tim Welschehold and Abhinav Valada, Learning Long-Horizon Robot Exploration Strategies for Multi-Object Search in Continuous Action Spaces, Proceedings of the International Symposium on Robotics Research (ISRR), 2022.