Speeding up the convergence of online heuristic search and scaling up offline heuristic search

• Authors: David A. Furcy;Sven Koenig
• Venue: Speeding up the convergence of online heuristic search and scaling up offline heuristic search
• Year: 2004

Heuristic search algorithms are popular Artificial Intelligence methods for solving the shortest-path problem. This research contributes new heuristic search algorithms that are either faster or scale up to larger problems than existing algorithms. Our contributions apply to both online and offline tasks. For online tasks, existing real-time heuristic search algorithms learn better informed heuristic values and sometimes eventually converge to a shortest path by repeatedly executing the action leading to a successor state with a minimum cost-to-goal estimate. In contrast, we claim that real-time heuristic search converges faster to a shortest path when it always selects an action leading to a state with a minimum f-value (i.e., a minimum estimate of the cost of a shortest path from start to goal via the state), just like in the offline A* search algorithm. We support this claim by implementing this new non-trivial action-selection rule in FALCONS and by showing empirically that FALCONS significantly reduces the number of actions to convergence of a state-of-the-art real-time search algorithm. For offline tasks, we scale up two best-first search approaches. First, a greedy variant of A* called WA* is known (1) to consume less memory to find solutions of equal cost when it is diversified (i.e., when it performs expansions in parallel), as in KWA*; and (2) to solve larger problems when it is committed (i.e., when it chooses the state to expand next among a fixed-size subset of the set of generated but unexpanded states), as in MSC-WA*. We claim that WA* solves even larger problems when it is enhanced with both diversity and commitment. We support this claim with our MSC-KWA* algorithm. Second, it is known that breadth-first search solves larger problems when it prunes unpromising states, resulting in the beam search algorithm. We claim that beam search quickly solves even larger problems when it is enhanced with backtracking based on limited discrepancy search. We support this claim with our BULB algorithm. We demonstrate the improved scaling of MSC-KWA* and BULB empirically in three standard benchmark domains. Finally, we apply anytime variants of BULB to the multiple sequence alignment problem in biology.