A reinforcement learning strategy for task scheduling of WSNs with mobile nodes

In this paper we describe a Markov Decision Process (MDP) based technique called Q-Learning which has been adapted for scheduling of tasks for wireless sensor networks (WSNs) with mobile nodes. The limited energy resources of WSN nodes have determined researchers to focus their attention at energy efficient algorithms which address issues of optimum communication, synchronization and scheduling. When considering scheduling however, most of the research existent is focused on MANETs with the purpose of scheduling packet transmission to reduce traffic load and avoid collisions. The task of WSN nodes depends very much on the applications they are intended for and mostly refer to monitoring the variation of one or more environmental parameters. Transmission is used only under certain conditions defined by the application and thus task scheduling should not be limited to communication optimization and collision avoidance. For this reason we present a reinforcement learning technique which can be used for WSNs with mobile nodes to schedule tasks in order to obtain efficient energy consumption while maintaining a high quality of service.