Single-Agent Parallel Window Search: A Summary of Results

We show how node ordering can be combined with parallel window search to quickly find a nearly optimal solution to single-agent problems. First, we show how node ordering by maximum g among nodes with equal / = g + h values can improve the performance of IDA*. We then consider a window search where different processes perform IDA* simultaneously on the same problem but with different cost thresholds. Finally, we combine the two ideas to produce a parallel window search algorithm in which node ordering information is shared among the different processes. Parallel window search can be used to find a nearly optimal solution quickly, improve the solution until it is optimal, and then finally guarantee optimality, depending on the amount of time available. 1 Introduction and Overview Heuristic search is a fundamental problem-solving method in artificial intelligence. Common examples of single-agent search problems are the Eight Puzzle and its larger relative the Fifteen Puzzle. The Eight Puzzle consists of a 3x3 square frame containing 8 numbered square tiles and an empty position called the 'blank'. The legal operators slide any tile horizontally or vertically adjacent to the blank into the blank position. The task is to rearrange the tiles from some random initial configuration into a particular desired goal configuration. A real world Chris Ferguson helped derive the limitation of pure window search and produced the figure. Bob Felderman and Othar Hansson's discussions and draft reviews substantially improved the paper. Discussions with Andy Mayer also contributed. example of single-agent search is the traveling salesman problem of finding the shortest simply connected tour among a set of cities to be visited. An optimal-solution algorithm for single-agent heur-istic search is A* [Hart, 1968]. For each node n, the cost of a path from the initial state to node n, g(n), is added to the estimated cost of reaching a goal from node n, h(n), to arrive at an estimate of the cost of a path from the initial state to a goal state that passes through node n, f(n). A* works by always expanding next a node of minimum /(n) = g(n) + h(n) until a goal node is chosen for expansion. The solution length found by A* is guaranteed to be optimal (lowest cost) if the heuristic function never overestimates the cost of the cheapest path to the goal. In practice, however, A* is not practical because it requires an …