Distributed Algorithm for Incrementally Solving the Decoupled Multi-agent Simple Temporal Problem

Abstract

—Applying temporal constraint in planning is a well-known problem, which keeps a plan is flexible until a specific schedule is generated. In this area, Decoupled Multi-agent simple temporal problem (DMaSTP) is suitably applied for planning of a multi-agents system. However, in scheduling problem, new events or temporal constraints are added regularly and force scheduler to check the consistency of exist MaSTP and retighten exist constraints. In this paper, we study a distributed scheduling algorithm for incrementally solving a DMaSTP. We have strongly considered the problem of adding a set of new constraints into a tightening consistent DMaSTP that tightened by also a distributed algorithm or set as empty. The algorithm checks whether the new adding constraints threaten the consistence of DMaSTP or not and decouple such new adding constraints when necessary and retighten DMaSTP. We have proposed a distributed algorithm, called DI-DMaSTP that solves the above problem and theoretically prove its correctness and outperformance, besides we have experienced with the variant datasets.