An evolutionary approach to capacitated resource distribution by a multiple-agent team

Abstract

A hybrid implementation of an evolutionary metahueristic scheme with local optimization has been applied to a constrained problem of routing and scheduling a team of robotic agents to perform a resource distribution task in a possibly dynamic environment. In this paper a central planner is responsible for planning routes and schedules for the entire team of cooperating robots. The potential computational complexity of such a centralized solution is addressed by an innovative genetic, approach that transforms the task of multiple route design into a special manifestation of the traveling salesperson problem. The key advantage of this approach is that globally optimal or near optimal solutions can be produced in a timeframe amenable for real-time implementation. The algorithm was tested on a set of standard problems with encouraging results. © Springer-Verlag Berlin Heidelberg 2003.