Time-constrained heuristic search for practical route finding

In this paper, we describe a heuristic search algorithm that finds a provably optimal solution subject to a specified time interval. While this algorithm is amenable to previous works on an approximate A* search with a bounded error (∈), it allows us to therminate the search to retain the specified time interval by changing the value of ∈ during the search. Our basic search strategy is that node expansion around the starting node is pessimistic (exact search), and we accomplish the approximate exploration of nodes around the goal by increasing ∈. This strategy is suitable for real-time route finding in automobile navigation systems. We conducted our experiments to clarify the practical features of the algorithm, using a digital map of a commercial automobile navigation system. The resulting advantage is that the estimation of the finishing time of the search is quite accurate, and optimal solutions are produced by making full use of the permissible search time.