Behavior Abstraction for Tractable Simulation

ion for Tractable Simulation Daniel J. Clancy and Benjamin Kuipers Department of Computer Sciences University of Texas at Austin Austin, Texas 78712 clancy@cs.utexas.edu and kuipers@cs.utexas.edu Abstract Most qualitative simulation techniques perform simulation at a single level of detail highlighting a fixed set of distinctions. This can lead to intractable branching within the behavioral description. The complexity of the simulation can be reduced by eliminating uninteresting distinctions. Behavior abstraction provides a hierarchy of behavioral descriptions allowing the modeler to select the appropriate level of description highlighting the relevant distinctions. Two abstraction techniques are presented. Behavior aggregation eliminates occurrence branching by providing a hybrid between a behavior tree representation and a history based description. Chatter box abstraction uses attainable envisionment to eliminate intractable branching due to chatter within a behavior tree simulat..