Abstract:

beyond the abilities of individuals. Artificial intelligence endeavors to solve complex tasks in many domains, such as planning in manufacturing, autonomic computing, collaborative electrical power management, autonomous traffic control, and disaster rescue. As such tasks are beyond the capabilities of a single intelligent agent, collaborative intelligence of multi-agent systems will be indispensable to address these challenges.  

Bio:

My research aims to enable effective, autonomous multi-agent collaboration. In this talk, I present a scalable multi-agent learning framework that allows a network of agents to concurrently learn to effectively coordinate and collaborate in a dynamic, uncertain environment. This framework exploits both interaction locality and non-local information, effectively scaling to thousands of intelligent agents. To address fairness in multi-agent systems, I define egalitarian solution concepts for multi-agent decision making under uncertainty and develop corresponding methods for computing optimal fair policies.