A reinforcement learning-based sleep scheduling algorithm for cooperative computing in event-driven wireless sensor networks

Emergency event monitoring is an important application of wireless sensor networks (WSNs). In the traditional cloud-assisted WSNs, the monitored data needs to be sent back to the cloud for processing. The round-trip of huge and burst data causes long delay and high energy consumption. With the increase of the computing capability of sensor nodes, local processing of events can be accomplished through cooperative computing between sensor nodes. However, if a node continues to undertake cooperative computing tasks, the energy consumption of nodes will be unbalanced and the network lifetime will be shortened. Sleep scheduling is an effective method to achieve energy balance, which can activate idle nodes alternately to participate in cooperative computing. In this paper, in order to make WSN more efficient to complete the local processing of events, we build a new system model based on node cooperative computing, and propose a multi-node Q learning-based cooperative computing node selection algorithm to obtain the sleep scheduling strategy. Simulation results show that compared with the classical algorithms, the proposed algorithm can complete more event processing, improve the reliability, and prolong the lifetime of WSNs.