Resource coordination in wireless sensor networks by cooperative reinforcement learning

Abstract

Wireless Sensor Networks (WSN) typically operate in dynamic environments, hence we can not schedule the execution of tasks a priori. This must be done online in a way to minimize the resource consumption. We present a cooperative reinforcement learning approach to schedule the tasks in WSN. A WSN is composed of a large number of tiny sensing nodes capable of interacting with the environment, communicating wirelessly and perform limited processing. In every time step, the sensor nodes need to take decision about some tasks to perform. Our proposed algorithm helps sensor nodes to learn the usefulness of each task based on reinforcement learning. We present an object tracking application with online scheduling of tasks based on our proposed approach. Our simulation studies show a more efficient task scheduling than traditional resource management schemes such as static scheduling, random scheduling and independent reinforcement learning based scheduling of tasks.