Dynamic Weighted Anytime Bounded Cost Search Algorithm

The best first algorithm like A* can find the optimal path if the time and memory is enough. However, it is usually difficult to do it in many cases. Moreover, there is always not enough time to support algorithm to find the optimal solution. On the other hand, users usually need an available solution rather than the optimal solution. Based on this, we propose Dynamic weighted Anytime Bounded Cost Search(DW-ABCS) algorithm. It combines the Bounded sub-optimal search(BSS) with the Bounded cost search(BCS) and improves the problem that BCS may not find a solution if the cost bound set is less than the optimal solution. Furthermore, in order to improve the efficiency of search, DW-ABCS algorithm dynamically modify the cost bound by considering the depth of nodes in the search process. In this way, the nodes considered become less and less and the complexity of algorithm will be lower and lower along the depth of search increasing, which improves the search efficiency significantly. In addition, the idea of anytime algorithm is used to find a better solution in every iteration. In this way, the algorithm can quickly find a sub-optimal path within suboptimal bound, and then reducing the sub-optimal bound to optimize it until the time out. If time permitted, the algorithm can converge to the optimal solution. Specially, we conducted a rigorous theoretical analysis. The various gird maps are used to verify the performance of DW-ABCS. Experimental results show that DW-ABCS algorithm can effectively identify the sub-optimal solution in different scenes.