Qualitative Simulation as a Temporally-extended Constraint Satisfaction Problem

qYaditionally, constraint satisfaction problems(CSPs) are characterized using a finite set of constraints expressed within a common,shared constraint language. Whenreasoning across time, however, it is possible to express both temporal and state-based constraints represented within multiple constraint languages. Qualitative simulation provides an instance of this class of CSPin which, traditionally, all solutions to the CSPare computed. In this paper, we formally describe this class of temporally-extended CSPsand situate qualitative simulation within this description. This is followed by a description of the DecSIMalgorithm whichis used to incrementally generate all possible solutions to a temporally-extended CSP. DecSIM combines problemdecomposition, a tree-clustering algorithm and ideas similar to directed arc-consistency to exploit structure and causality within a qualitative modelresulting in an exponential speed-up in simulation time whencompared to existing techniques.

• Type: Conference
• Year: 1998
• Where: AAAI
• Authors: Daniel J. Clancy, Benjamin Kuipers
• Keywords: AAAI 1998, Common, Constraint Languages, Intelligent Agents, Qualitative Simulation