Interactive-deepening fuzzy heuristic search algorithms and their learning investigation

This paper presents a fuzzy, heuristic search algorithm FIDA* that applies iterative-deepening methods and dynamic modification of the heuristic function. It then improves the FIDA* algorithm by incorporating methods to propagate the modified heuristic function, &htilde;(n). It is pointed out that the algorithm FIDA* can be applied to learn over repeated uses on problems with the same state space and goal node. This paper presents a learning algorithm, LFIDA*, that overcomes difficulties that often occur in designing &htilde;(n).