Asymmetric Planning for Incremental Real-Time Heuristic Search

Real-world applications typically permit only constant time intervals between subsequent moves of an agent interacting with an environment. Hence there is a need to efficiently distribute the limited computation time and also other resources available for planning ahead and for evaluating alternative future action sequences. In this paper we study a particular instance of this general problem, namely, the choice of local search space in real-time heuristic search. For larger local search spaces, the well-known LRTA* algorithm finds shorter solutions, but requires a lot of memory, especially in huge domains such as sliding tile puzzles. We suggest a novel algorithm CLRTA* that uses large, but asymmetric local search space, placed mostly within the previously explored area. In our experiments on sequences of sliding tile puzzles and mazes, CLRTA* produced solutions not much longer than the standard symmetric LRTA* with the same number of states examined per decision, while using much less memory.