Reflective Negotiating Agents for Real-Time Multisensor Target Tracking

Abstract

In this paper we describe a multiagent system in which agents negotiate to allocate resources and satisfy constraints in a real-time environment of multisensor target tracking. The agents attempt to optimize the use of their own consumable resources while adhering to the global goal, i.e., accurate and effective multisensor target tracking. Agents negotiate based on different strategies which are selected and instantiated using case-based reasoning (CBR). Agents are also fully reflective in that they are aware of all their resources including system-level ones such as CPU allocation, and this allows them to achieve real-time behavior. We focus our discussion on multisensor target racking, case-based negotiation, and real-time behavior, and present experimental results comparing our methodology to ones using either no negotiation or using a static negotiation protocol.