Jeffrey R. N. Forbes
BS Stanford University, Computer Science, 1993
of California, Berkeley, Computer Science 2000.
Major area: Artificial intelligence Advisor: Stuart Russell
of Computer Science Duke University
home page: faculty at http://www.cs.duke.edu/~forbes/
My research focuses on artificial intelligence and the design of intelligent agent architectures. My work draws from machine learning, adaptive control, simulation environments, and empirical AI techniques. Directed by my work on controlling autonomous vehicles in the Bayesian Automated Taxi project, I have developed a reinforcement learning control system. There exist future applications with a variety of complex systems.
Consider the problem of controlling an agent in a domain with complex real-time dynamics. In most systems, there is rarely a known optimal trajectory from which we can either directly derive a control policy or simply generalize from given examples. Instead, the goal is to maximize general performance according to a given set of factors. Markov decision processes (MDPs) and optimal control provide a framework to properly formulate this objective and have become the basis of much of the work in intelligent control and planning. Reinforcement learning (RL) is one method whereby the agent successively improves its control strategy through experience and feedback (reward) from the system. RL techniques have shown some promise in solving complex control problems. However, RL algorithms often do not scale up well to nonuniform problems with large or infinite state and action spaces. My research develops an architecture for extending RL to complex control domains by effectively maintaining a value function approximation and exploiting the structure of the environment using domain models and hierarchy. The system lends itself to applications in a variety of robotic and simulated domains.
J. Forbes, T. Huang, K. Kanazawa, and S. Russell. The BATmobile: Towards a Bayesian Automated Taxi. In Proceedings of the International Joint Conference on Artificial Intelligence, 1995.
M. Wellman, C. Liu, D. Oynadath, S. Russell, J. Forbes, T.
Huang, and K. Kanazawa. Decision-theoretic reasoning for traffic
monitoring and vehicle control. In Proceedings of the Intelligent
Vehicles '95 Symposium, 1995.
|This website was created by and is maintained
Dr. Scott Williams, Professor of Mathematics
State University of New York at Buffalo
visitors since opening 5/25/97