Simple and Efficient Bi-Objective Search Algorithms via Fast Dominance Checks
Carlos HernĂĄndez,William Yeoh,Jorge A. Baier,Han Zhang,Luis Suazo,Sven Koenig,Oren Salzman
DOI: https://doi.org/10.1016/j.artint.2022.103807
IF: 14.4
2022-10-15
Artificial Intelligence
Abstract:Many interesting search problems can be formulated as bi-objective search problems, that is, search problems where two kinds of costs have to be minimized, for example, travel distance and time for transportation problems. Instead of looking for a single optimal path, we compute a Pareto-optimal frontier in bi-objective search, which is a set of paths in which no two paths dominate each other. Bi-objective search algorithms perform dominance checks each time a new path is discovered. Thus, the efficiency of these checks is key to performance. In this article, we propose algorithms for two kinds of bi-objective search problems. First, we consider the problem of computing the Pareto-optimal frontier of the paths that connect a given start state with a given goal state. We propose Bi-Objective A* (BOA*), a heuristic search algorithm based on A*, for this problem. Second, we consider the problem of computing one Pareto-optimal frontier for each state s of the search graph, which contains the paths that connect a given start state with s . We propose Bi-Objective Dijkstra (BOD), which is based on BOA*, for this problem. A common feature of BOA* and BOD is that all dominance checks are performed in constant time, unlike the dominance checks of previous algorithms. We show in our experimental evaluation that both BOA* and BOD are substantially faster than state-of-the-art bi-objective search algorithms.
computer science, artificial intelligence