An Evolutionary Approach to Capacitated Resource Distribution by a Multiple-agent Team

Mudassar Hussain¹, Bahram Kimiaghalam¹, Abdollah Homaifar¹, Albert Esterline¹, and Bijan Sayyarodsari²

¹NASA Autonomous Control and Information Technology Center,
Department of Electrical Engineering,
North Carolina A&T State University,
Greensboro, NC 27411
muddie27@yahoo.com,
{bahram,homaifar,esterlin}@ncat.edu

²Pavilion Technologies,
11100 Metric Blvd., #700
Austin, TX 78758
bijan@pav.com

Abstract. A hybrid implementation of an evolutionary metaheuristic 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.

LNCS 2723, p. 657 ff.

Full article in PDF

lncs@springer.de
© Springer-Verlag Berlin Heidelberg 2003