Vadim Bulitko: “Real-time Learning and Search in Game-like Environments”
June 20, 2006 | USC ICTSpeaker: Vadim Bulutko
Host: Michael van Lent
The pursuit of moving targets in real-time environments such as computer games and robotics presents several challenges to situated agents.
A priori unknown state spaces and the need to interleave acting and planning limits applicability of traditional learning, heuristic search, and adversarial game-tree search methods. In this talk we demonstrate how even simple opponent modeling techniques can be used to boost efficiency of moving target search methods. We then introduce automated techniques for reducing state space size by building hierarchical abstraction of maps. The talk will be concluded with an outlook on marrying opponent modeling and state abstraction techniques.