Learning-Based Coordination of Large Heterogeneous Distributed Sensor Networks

摘要

The coordination of large heterogeneous distributed sensor networks is a critical task, and has applications in both smart grid and advanced energy system management. As the number of sensors in a network reaches thousands or more, coordination of these sensors becomes an increasingly complex problem not well suited for traditional control techniques. In this paper, we introduce an artificial-intelligence based multiagent framework for coordinating thousands of sensors in a heterogeneous distributed sensor network with up to 10,000 devices. We test our algorithm on an abstract distributed sensor network in a time-extended multi-target defect combination problem. Our results demonstrate the proposed algorithm can successfully schedule sensor sleep/sense schedules while ensuring network quality of service is maintained. Our algorithm results in up to 76% less average network error and up to 22% greater source coverage than traditional learning based techniques, and is robust to sensor noise and failures within the network.