CAREER: Rapid and Continuous Adaptation of Service Robots via Coupled Learning and Planning

About This Grant

Service robots can be of great help in hospitals, caring for older adults, and for everyday home tasks. However, today's robots need a large amount of retraining whenever they face new situations. This is like having to go back to school for every new job. This Faculty Early Career Development (CAREER) project research will help robots learn like humans do - learning on the job. The new system will enable home robots to learn new skills from human teachers quickly and on their own. It will also allow them to adapt fast to new tasks through smart planning. To get young students excited about robotics and artificial intelligence (AI), the team will develop teaching programs for K-12 students from different backgrounds. The project will also create new college classes that combine AI and robotics to prepare students for future jobs. Robot learning on its own is challenging in real world scenarios - robots must learn individual skills as parts of longer, more complex tasks. For example, learning to cut vegetables requires first learning to grasp knives well. This research will develop a new robot system where intelligent planning helps guide the robot to learn. The system helps the robot find missing skills and design situations that help the robot to practice these skills. The system also uses these learned skills to help the robot plan for more complex tasks. The system has three main components: (1) learning new skills with a symbolic planner as guidance; (2) efficient learning from occasional human guidance when needed; and (3) fast adaptation to new tasks through intelligently combining learned skills. The team will test the system in simulated environments and on real robots used in home care research centers. The focus will be on daily household tasks that need long-term adaptation. This research will speed up the adoption of capable robots in healthcare, manufacturing, and home assistance areas. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.