Improved Pattern Selection for PDB Heuristics in Classical Planning (Extended Abstract)

Sascha Scherrer, Florian Pommerening, Martin Wehrle · Proceedings of the International Symposium on Combinatorial Search · 2021

The iPDB approach selects patterns by a local search in the space of pattern collections. This search often gets stuck in local optima, which limits the quality of the resulting heuristic. In this research abstract, we report on current progress to tackle this problem. We investigate variable neighborhood search with encouraging experimental results.

Read the paper · More papers on PaperTik