Learning to Do HTN Planning

We describe the HDL algorithm, which learns HTN domain representations by examining plan traces produced by an expert problem-solver. Prior work on learning HTN methods required everything to be given in advance except for the methods’ preconditions, and the learner would learn the preconditions. In contrast, HDL has no prior information about the methods. In our experiments, in most cases HDL converged fully with no more than about 200 plan traces. Furthermore, even when HDL was only halfway to convergence, it usually was able to produce HTN methods that were sufficient to solve more than 3/4 of the planning problems in the test set.