Abstract | ||
---|---|---|
A common paradigm in classical planning is heuristic forward search. Forward search planners often rely on relatively simple best-first search algorithm, which remains fixed throughout the search process. In this paper, we introduce a novel search framework capable of alternating between several forward search approaches while solving a particular planning problem. Selection of the approach is performed using a trainable stochastic policy. This enables tailoring the search strategy to a particular distribution of planning problems and a selected performance metric, such as the IPC score or running time. We construct a strategy space using five search algorithms and a two-dimensional representation of the planneru0027s state. Strategies are then trained on randomly generated planning problems using policy gradient. Experimental results show that the learner is able to discover domain-specific search strategies, thus improving the planneru0027s performance with respect to the chosen metric. |
Year | Venue | Field |
---|---|---|
2018 | international conference on automated planning and scheduling | Heuristic,Search algorithm,Computer science,Performance metric,Planner,Artificial intelligence,Machine learning |
DocType | Volume | Citations |
Journal | abs/1810.09923 | 0 |
PageRank | References | Authors |
0.34 | 0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Pawel Gomoluch | 1 | 0 | 0.34 |
Dalal Alrajeh | 2 | 119 | 13.75 |
Alessandra Russo | 3 | 1022 | 80.10 |