A Real-Time Planning System in Dynamic Nondeterministic Environments
Xiao Chen · Chinese Journal of Computers · 2005
Planning is a main research field in artificial intelligence, which has a broad application background. In recent years, the focus has shifted to the problem of planning in dynamic nondeterministic environments. This paper presents a novel continual planning system, POMDPRS, which combines POMDP and PRS to gain stronger abilities of adapting to dynamic nondeterministic environments. In this system authors substitute the probabilistic distribution belief model and the maximum utility principle of POMDP for the first-order logic belief representation and plan selection mechanism of PRS; on the other hand, adopt the continual planning mechanism of PRS, which interleaves plan generation and execution, to make the POMDP decision in dynamic environments more efficient under some conditions.