A Multi-Agent Temporal Constraint Satisfaction System Based on Allen's Interval Algebra and Probabilities

We present a multi-agent distributed temporal scheduling system based on probabilistic Interval Algebra networks (i.e., PIA networks). The overall goal is to make each agent’s network consistent and optimal. Each agent is independent and has sole control over its network. But, it must communicate and coordinate with other agents when modifying or updating shared edges. We present an algorithm which allows the agents to collaboratively solve and recommend a temporal schedule. At the agent level, this schedule is optimal under the given local constraints. Although the global solution may not be optimal, we try to generate near optimal ones. Note that our distributed system is not centrally controlled. We use our system to solve scheduling problems from a university problem domain. We also describe the implementation of our proposed system which uses Java, JXTA, and JGraph.