A Branch-and-Cut Approach for a Mixed Integer Linear Programming Compilation of Optimal Numeric Planning

Ryo Kuroiwa, Chris Beck

Keywords: numeric planning, mixed-integer linear programming

TL;DR: This paper proposes a novel mixed-integer linear programming compilation of numeric planning using a branch-and-cut algorithm.

Abstract: In this paper, we consider optimal numeric planning, focusing on numeric planning with simple conditions and on linear numeric planning. We propose a novel compilation of numeric planning to mixed-integer linear programming (MILP) and employ a branch-and-cut algorithm to lazily generate constraints. We empirically compare the proposed method with heuristic search algorithms and other model-based approaches including an existing MILP based method. Although the new method is not competitive with heuristic search algorithms, compared to the existing MILP based method, it finds the optimal solutions faster in some planning domains and solves two more instances in one domain.