Plan-A: A Cost Optimal Planner Based on SAT-Constrained Optimization

Planning as satisfiability, represented by SATPlan, can find plans with the shortest makespan for classical planning. However, it has been deemed a limitation of SAT-based approaches for delivering optimality regarding metrics other than the makespan. Our Plan-A solver uses the SAT-based framework to find plans with minimal total action costs. The idea is to optimally solve SAT instances based on certain optimality metrics. Bounding and pruning techniques are proposed and integrated with the standard DPLL algorithm to reduce the search cost.