Proceedings

Exploiting Structure in Policy Construction

Craig Boutilier, Richard Dearden, Moisés Goldszmidt

Download BibTeX Cite

Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence. Learning. Pages 1104-1112.

Abstract

Markov decision processes (MDPs) have recently been applied to the problem of modeling decisiontheoretic planning. While traditional methods for solving MDPs are often practical for small states spaces, their effectiveness for large AI planning problems is questionable. We present an algorithm, called structured policy iteration (SPI), that constructs optimal policies without explicit enumeration of the state space. The algorithm retains the fundamental computational steps of the commonly used modified policy iteration algorithm, but exploits the variable and propositional independencies reflected in a temporal Bayesian network representation of MDPs. The principles behind SPI can be applied to any structured representation of stochastic actions, policies and value functions, and the algorithm itself can be used in conjunction with recent approximation methods.

Craig Boutilier, Richard Dearden, and Moisés Goldszmidt. 1995. Exploiting Structure in Policy Construction. In Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence, IJCAI-95. IJCAI, 1104–1112. https://ijcai.org/proceedings/1995/143
Welcome Future Conferences AI Journal Division IJCAI Proceedings About IJCAI IJCAI Awards