Receiver Beacon Transmission Interval Design Using Machine Learning for Multi-Stage Wireless Sensor Networks

Abstract

U-Bus Air is a protocol that can transmit packets by multi-hop communication, and is utilized in tall buildings and apartments. While U-Bus Air is a time asynchronous wireless sensor network, packet collisions cause degradation of communication quality as the number of terminals increases. A previous research has been proposed, an appropriate transmission cycle control algorithm based on the estimated collision probability by using collected topology information at a Fusion Center (FC). By using this algorithm, the communication performance is improved by feedback from FC to each terminal. However, its communication overhead becomes large due to topology information and feedback packets, and communication efficiency is degraded. In order to solve this problem, we propose a method to autonomously control receiver beacon transmission interval by leveraging decentralized Q-Learning. In this method, each terminal transmits information of a packet holding time at the same time as responding to the beacon. Finally, each terminal utilizes Q-Learning to derive an appropriate transmission interval. In computer simulation, we show an improvement of packet delivery rate (PDR), throughput and power consumption compared with the previous method.