Development and Training of Multiagent Systems
Multiagent systems are distributed systems in which multiple independent computational entities (agents) interact, but none of them has a global view of the system, nor knows exactly what the others are doing. These kinds of scenarios are very common: teams of robots, vehicular or airport traffic, multiagent simulations of social behavior, trading floor agents, opponents in video games, dynamic social networks, swarms of animated orcs and zombies in movies, and so on.
It is very difficult to predict the dynamics of a multiagent system, and thus challenging to optimize or design agents for such a system. This is because of the number and heterogeneity of the agents involved, the complexity of the interactions among them, and their tendency to step on each others' toes due to their incomplete knowledge of the system as a whole, and lack of centralized control. I will discuss a selection of research in various areas of multiagent systems, including multiagent stochastic optimization, robot swarm behaviors, and real-time training of multiagent teams. These areas are very different from one another, but all share common multiagent systems difficulties which make them challenging.
BIOGRAPHY
Sean Luke is Associate Professor in the Department of Computer Science and Associate Director of the Center for Social Complexity at George Mason University. Prof. Luke received his Ph.D. in Computer Science at the University of Maryland, College Park. His research interests include multiagent systems, multiagent learning, robotics, simulation, models of social behavior, and stochastic optimization and metaheuristics. He has approximately 100 publications.