Learning Movement and Social Behaviors in Multi-Agent Settings
Seminar: EE: CS | November 8 | 3:30-4:30 p.m. | 540AB Cory Hall
Igor Mordatch, Senior Research Scientist, Open AI
In order to create truly autonomous robots, understand principles behind human cognition, or enable AI agents to successfully collaborate with humans, it is necessary to synthesize movement and social behaviors with the same wide variety, richness and complexity observed in humans and other animals. Moreover, these behaviors should emerge automatically from only a few core principles, and not be a result of extensive manual engineering or a mimicking of human patterns. In this talk, I will discuss how principles of optimal control, learning, and model-based reasoning can be used to generate and explain a broad range of human movement behaviors, from locomotion to dexterous manipulation, applicable both in simulation and on physical robot platforms. I will also discuss importance of multi-agent environments in development of social behaviors, from verbal and non-verbal communication, to cooperation and teaching.