Multi-agent pursuit-evasion algorithm based on contract net interaction protocol

Abstract

The motivation is the deployment of large numbers of inexpensive robots in hostile environments to pursue evasive targets. Based on the interaction protocols of contract net and subscribe-publish, a distributive pursuit-evasion algorithm is proposed for multi-agents to pursue multiple evasive agents. The pursuit agents engaged in teams adopt a tail-chase strategy while the rest pursuit agents use a hug-a-tree policy. The evading agents move towards directions with fewer pursuit agents in their neighborhood. All agents communicate according to the subscribe-publish protocol and the pursuit agents coordinate their acts through the contract net protocol. The result of simulation under JADE platform shows that the pursuit agents can dynamically form several teams to catch respective evaders efficiently without centralized leadership or hierarchical coordination. Distributive in nature, it allows a large-scale agent fleet to perform complex tasks in a coordinated way. © Springer-Verlag Berlin Heidelberg 2005.