Task assignment with dynamic token generation

The problem of assigning tasks to a group of agents acting in a dynamic environment is a fundamental issue for a MAS and is relevant to several real world applications. Several techniques have been studied to address this problem, however when the system needs to scale up with size, communication quickly becomes an important issue to address; moreover, in several applications tasks to be assigned are dynamically evolving and perceived by agents during mission execution. In this paper we present a distributed task assignment approach that ensure very low communication overhead and is able to manage dynamic task creation. The basic idea of our approach is to use tokens to represent tasks to be executed, each team member creates, executes and propagates tokens based on its current knowledge of the situation. We test and evaluate our approach by means of experiments using the RoboCup Rescue simulator.© Springer-Verlag Berlin Heidelberg 2005.