Collaboration abounds in nature, enabling groups to accomplish tasks beyond capabilities of any individual. Collaborative artificial intelligence is essential for intelligent agents, including physical robots and software agents, to solve complex real-world problems, by enabling them to work effectively with each other and humans.
In this talk, I will first introduce a hierarchical planning method for enabling large-scale human-robot collaboration. This planning method allows human-robot teams to work together safely and efficiently under temporal and spatial constraints and under uncertainty. I will then present a coordinated, distributed reinforcement learning framework for enabling a network of autonomous agents, such as multi-robot systems and sensor networks, to learn to efficiently collaborate and coordinate. This framework allows agents with limited communication to effectively learn when, with whom, and how to coordinate in a dynamic, uncertain environment. Finally, I will discuss future research directions, including collaborative learning to facilitate skill reuse and evolution, as well as collaborative planning for enabling fluid human-robot collaboration.
Chongjie Zhang is a Postdoctoral Associate in the Computer Science and Artificial Intelligence Lab at Massachusetts Institute of Technology. He received his Ph.D. in computer science from the University of Massachusetts at Amherst. His research interests broadly encompass artificial intelligence, reinforcement learning, human-robot interaction, decision-making under uncertainty, and game theory.