Title of Presentation: Sensor Webs As Multiagent, Negotiating Systems
Primary (Corresponding) Author: Costas Tsatsoulis
Organization of Primary Author: University of Kansas
Abstract: Our work makes Sensor Webs intelligent, collaborative, and self-aware. The intelligence is added to each sensor, and allows the Sensor Web to be adaptive to the events in the environment it senses. The Sensor Web becomes a multi-agent system, where individual sensor agents can collaborate, form dynamic coalitions to handle tasks that a single sensor cannot do, and can negotiate in order to form the best set of coalitions that maximize the overall utlility of the Sensor Web.
Each sensor hosts an intelligent agent that adds to the sensor autonomous decision making. An event monitor snoops the sensor data collected by the sensor, and, if it identifies interesting events, it activates the task planner and the reasoning component. While the event monitor reacts to events, the task planner uses these events to predict future sensing requirements, needs, and parameters. The reasoning component decides whether the sensor can satisfy the new sensing needs arising from the event on its own, or it will require the assistance of other sensors. In the latter case, the agent contacts other sensor agents that it believes can assist it, and forms an initial coalition. When a sensor rejects a request to join a sensing coalition, the initiating sensor will engage in negotiation with it. Negotiation is a formal technique that allows autonomous agents to reach an agreement on resource sharing and collaborative tasks. Our sensor agents are self-aware and introspective, indicating that they have the ability to autonomously determine their current state, tasks, and capabilities, and to engage in negotiations. After the negotiation phase the coalition of sensors is finalized, and then proceeds with its sensing tasks.