Despite the advances made in the last decade in automated planning, no planner out-\udperforms all the others in every known benchmark domain. This observation motivates\udthe idea of selecting different planning algorithms for different domains. Moreover, the\udplanners’ performances are affected by the structure of the search space, which depends\udon the encoding of the considered domain. In many domains, the performance of a plan-\udner can be improved by exploiting additional knowledge, for instance, in the form of\udmacro-operators or entanglements.\udIn this paper we propose ASAP, an automatic Algorithm Selection Approach for\udPlanning that: (i) for a given domain initially learns additional knowledge, in the form\udof macro-operators and entanglements, which is used for creating different encodings\udof the given planning domain and problems, and (ii) explores the 2 dimensional space\udof available algorithms, defined as encodings–planners couples, and then (iii) selects the\udmost promising algorithm for optimising either the runtimes or the quality of the solution\udplans.
展开▼
