An Extensional Junction Tree Approximate Inference Algorithm for Dynamic Influence Diagrams

We introduce a dynamic influence diagram (DID) to model a multi-agent system in dynamic environment, where the DID is an extension of a static influence diagram over time. Based on splitting junction tree and Boyen-Koller algorithm, an extensional junction tree approximate inference algorithm of DID is proposed in this paper, where clusters of junction tree are split by strategic relevance. Finally, in the setting of Robocup, we use DID to emulate three agents' pass-catch problem, and analyze the complexity and the error of extensional junction tree approximate inference algorithm

[1]  Hao Wang,et al.  Simplifying computation of dynamic influence diagrams , 2004, Proceedings of 2004 International Conference on Machine Learning and Cybernetics (IEEE Cat. No.04EX826).

[2]  Kevin P. Murphy,et al.  The Factored Frontier Algorithm for Approximate Inference in DBNs , 2001, UAI.

[3]  Anders L. Madsen,et al.  Solving Influence Diagrams using HUGIN, Shafer-Shenoy and Lazy Propagation , 2001, UAI.

[4]  Mark A. Paskin,et al.  Thin Junction Tree Filters for Simultaneous Localization and Mapping , 2002, IJCAI.

[5]  Fengzhan Tian,et al.  A DBN inference algorithm using junction tree , 2004, Fifth World Congress on Intelligent Control and Automation (IEEE Cat. No.04EX788).

[6]  Adnan Darwiche,et al.  Constant-space reasoning in dynamic Bayesian networks , 2001, Int. J. Approx. Reason..

[7]  Sridhar Mahadevan,et al.  Recent Advances in Hierarchical Reinforcement Learning , 2003, Discret. Event Dyn. Syst..

[8]  Xavier Boyen,et al.  Tractable Inference for Complex Stochastic Processes , 1998, UAI.

[9]  Anders L. Madsen,et al.  LAZY Propagation: A Junction Tree Inference Algorithm Based on Lazy Evaluation , 1999, Artif. Intell..