Case-Based Reflective Negotiation Model

摘要

The goal of our work was to create autonomous agents that worked in a domain of constrained resources and that used case-based negotiation to allocate resources in a satisfying manner. The domain of application was distributed sensor management for multi-target tracking. We developed two separate but similar formalisms for the implementation of the negotiation agents: first, a real-time architecture based on the Belief-Desire- Intention (BDI) agent framework and second, a near real-time architecture that relied on domain heuristics. Both architectures used Case-Based Reasoning (CBR) to select a negotiation strategy, and then negotiated using argumentation. The BDI architecture used a real-time Linux kernel to have time awareness. In addition to this main thrust of our work, we were also tasked to provide tools for analyzing the performance of Java code, real-time Linux services, a Communication Server to model the RF communication channels of the sensors, and for studying and improving the Challenge Problem code.