Divide-and-Evolve (DaE) is an original "memeticization" of Evolutionary Computation and Artificial Intelligence Planning. However, like any Evolutionary Algorithm, DaE has several parameters that need to be tuned, and the already excellent experimental results demonstrated by DaE on benchmarks from the International Planning Competition, at the level of those of standard AI planners, have been obtained with parameters that had been tuned once and for-all using the Racing method. This paper demonstrates that more specific parameter tuning (e.g. at the domain level or even at the instance level) can further improve DaE results, and discusses the trade-off between the gain in quality of the resulting plans and the overhead in terms of computational cost.

Abstract. Divide-and-Evolve (DaE) is an original "memeticization" of Evolutionary Computation and Artificial Intelligence Planning. DaE optimizes either the number of actions, or the total cost of actions, or the total makespan, by generating ordered sequences of intermediate goals via artificial evolution. The evolutionary part of DaE is based on the Evolving Objects (EO) library, and can theorically use any embedded planner. However, since the introduction of this approach only one embedded planner has been used: the temporal optimal planner CPT. In this paper, we built a new version of DaE based on time-based Atom Choice and we embarked another planner (the sub-optimal planner YAHSP) in order to test the technical robustness of the approach and to compare the impact of using an optimal planner versus using a sub-optimal planner for all kinds of planning problems.

Abstract. Divide-and-Evolve (DAE) is the first evolutionary planner that has entered the biennial International Planning Competition (IPC). Though the overall results were disappointing, a detailed investigation demonstrates that in spite of a harsh time constraint imposed by the competition rules, DAE was able to obtain the best quality results in a number of instances. Moreover, those results can be further improved by removing the time constraint, and correcting a problem due to completely random individuals. Room for further improvements are also explored.