Task planning with OMT: An application to production logistics

Task planning is a well-studied problem for which interesting applications exist in production logistics. Planning for such domains requires to take into account not only feasible plans, but also optimality targets, e.g., minimize time, costs or energy consumption. Although there exist several algorithms to compute optimal solutions with formal guarantees, heuristic approaches are typically preferred in practical applications, trading certified solutions for a reduced computational cost. Reverting this trend represents a standing challenge within the domain of task planning at large. In this paper we discuss our experience using Optimization Modulo Theories to synthesize optimal plans for multi-robot teams handling production processes within the RoboCup Logistics League. Besides presenting our results, we discuss challenges and possible directions for future development of OMT planning.