Our mission is to enable embodied AI agents to perform long-horizon tasks in home environments while seamlessly interacting with humans
“The vision of PEARL entails that the generalization power of machine learning methods is necessary to enhance classical model-based approaches in robotics, e.g., path planning, and to endow robots with intelligent behavior finally.”
Future artificial intelligence (AI) assistive robots will need to autonomously accomplish household tasks and naturally interact with humans. Research in AI and robotics has accomplished significant results in mobile robot navigation and in fixed base manipulation, but is still limited to specific installations and specialized to single tasks. We believe it is time to build on these significant progresses to finally achieve the vision of embodied AI systems that support humans in everyday tasks. Mobile manipulator robots are the most promising candidates for this role, as they combine the advantages of mobile platforms and manipulator arms. Current approaches in mobile manipulation propose solutions for semi-structured settings and repetitive scenarios, but lack intelligent behavior for generalizing to various problems, and for addressing the uncertainty of human-inhabited environments.
The embodiment hypothesis suggests that intelligent behavior may be acquired by the continual purposeful interaction of an agent with an environment and the induced sensorimotor experience. Motivated by this paradigm, we propose to advance beyond the current ad-hoc solutions by conducting research in our group, that is dedicated to developing methods for robot learning of mobile manipulation for intelligent assistance.
PEARL aims to introduce novel methods at the intersection of robotics and machine learning for mobile manipulation by enhancing model-based methods with the adaptation properties acquired by exploration and learning. This synergy will enable robots to solve long-horizon household tasks while interacting with humans and their environment, taking a step towards our vision for embodied AI assistants. We identify the following properties to be critical for acquiring intelligent mobile manipulation assistants:
- Robust mobile manipulation skills for household tasks, e.g., fetch, carry
- Fluent human-robot interaction, e.g., through safe object handovers
- Adaptive planning for learning long-horizon tasks, like fetch-carry-handover
Interactive Robot Perception & Learning
Powered by WordPress