Anytime Heuristic Search

Abstract

We describe how to convert the heuristic search algorithm A* into an anytime algorithm that finds a sequence of improved solutions and eventually converges to an optimal solution. The approach we adopt uses weighted heuristic search to find an approximate solution quickly, and then continues the weighted search to find improved solutions as well as to improve a bound on the suboptimality of the current solution. When the time available to solve a search problem is limited or uncertain, this creates an anytime heuristic search algorithm that allows a flexible tradeoff between search time and solution quality. We analyze the properties of the resulting Anytime A* algorithm, and consider its performance in three domains; sliding-tile puzzles, STRIPS planning, and multiple sequence alignment. To illustrate the generality of this approach, we also describe how to transform the memoryefficient search algorithm Recursive Best-First Search (RBFS) into an anytime algorithm.

Heuristic Search When Time Matters

“…Anytime algorithms are a very general class and there are many anytime algorithms for heuristic search (Likhachev, Gordon, & Thrun, 2003;Hansen & Zhou, 2007;Richter, Thayer, & Ruml, 2010;van den Berg, Shah, Huang, & Goldberg, 2011;Thayer, Benton, & Helmert, 2012). In this paper we use Anytime Repairing A* (ARA*, Likhachev et al, 2003) since it tended to give the best performance over other approaches according to experiments done by Thayer and Ruml (2010).…”

“…Techniques like DAS and Bugsy, on the other hand, only use information that can be computed on-line. Hansen, Zilberstein, and Danilchenko (1997) show how heuristic search with inadmissible heuristics can be used to make anytime heuristic search algorithms. Like the techniques presented in this paper, they consider the problem of trading-off search effort for solution quality.…”

A Survey and Classification of A* Based Best-First Heuristic Search Algorithms

“…A common feature of some anytime algorithms is the ability to prune the open list after the computation of the first solution, providing the necessary upper bound. The Anytime Weighted A* (AWA*) [23] is an example of such algorithm: it reduces the size of the open list when it prunes some paths. Similarly to A*, AWA* maintains two lists of nodes: open and closed.…”

Coordinating Service Distribution through Cooperative "YuuZuu" Reservations

“…For arrangement of reservations, DREAM adopts heuristic anytime search used in general search problems [9] and constraint satisfaction problems [10] to regulate the time required for the search and to control the influence of reservation rearrangements on the remainder of its database. In the database, the DREAM reservation system has a set of reservations R = {R 1 , ..., R n }, which are divided into three types such as fixed if T (S s ) has no other spare candidate then 9: change T (S s ) to a fixed reservation 10: end if 11: end for 12: Figure 4 shows the algorithm used to make a picky reservation.…”