Applying Genetic Algorithms to Hierarchical Task Network Planning

appears in PLANSIG 2000 We describe a new planner that uses a genetic algorithm and domain-specific knowledge bases to solve hierarchical task network (HTN) planning problems. The knowledge base describes decompositions of tasks into subtasks using different resources. For any given problem, the GA evolves a solution that specifies a recursive decomposition of the tasks and a set of resources to be used. The planner has been tested on two simple domains (logistics and disaster relief), and is now being applied to a more complex military ground operations domain