Multirobot task assignment in active surveillance

Abstract

An architecture for the assignment of tasks in a multirobot system is presented. The tasks arrival times and order are not known, and no statistical distributions are assumed. Furthermore, the execution of tasks cannot be preempted and hard deadlines are imposed on both their assignment and conclusion times. The architecture design is inspired by economic markets and a performance measure, similar to market prices, is introduced. A theoretical analysis of the architecture operation is presented and the results are applied to the dimensioning of an active surveillance system for the UPC campus at Barcelona, Spain. © 2009 Springer Berlin Heidelberg.