Satisfiability planning with constraints on the number of actions

Markus Büttner

We investigate satisfiability planning with restrictions on the number of actions in a plan. Earlier work has considered encodings of sequential plans for which a plan with the minimal number of time steps also has the minimum number of actions, and parallel (partially ordered) plans in which the number of actions may be much higher than the number of time steps. For a given problem instance finding a parallel plan may be much faster than finding a corresponding sequential plan but there is also the possibility that the parallel plan contains unnecessary actions.

Our work attempts to combine the advantages of parallel and sequential plans by efficiently finding parallel plans with as few actions as possible. We propose techniques for encoding parallel plans with constraints on the number of actions. Then we give algorithms for finding a plan that is optimal with respect to a given number of steps and an anytime algorithm for successively finding better and better plans. We show that as long as guaranteed optimality is not required, our encodings for parallel plans are often much more efficient in finding plans of good quality than a sequential encoding.