Efficient Wireless Sensor Network for Radiation Detection in Nuclear Sites

Abstract

Due to the severe damages of nuclear accidents, there is still an urgent need to develop efficient radiation detection wireless sensor networks (RDWSNs) that precisely monitor irregular radioactivity. It should take actions that mitigate the severe costs of accidental radiation leakage, especially around nuclear sites that are the main sources of electric power and a lot of health and industrial applications. Recently, leveraging machine learning (ML) algorithms to RDWSNs is a promising solution due to its several pros, such as online learning and self-decision making. This paper addresses a novel and efficient ML-based RDWSNs that utilizes millimeter waves (mmWaves) to cope with future networks requirements. Specifically, we leverage an online learning multi-armed bandit (MAB) algorithm called Thomson sampling (TS) to a 5G enabled RDWSN to efficiently forward the measured radiation levels of the distributed radiation sensors within the monitoring area. The utilized sensor nodes are lightweight smart radiation sensors that are mounted on mobile devices and measure radiations using software applications installed in these mobiles. Moreover, a battery aware TS (BA-TS) algorithm is proposed to forward the sensed radiation levels to fusion decision center efficiently. BA-TS reflects the remaining battery of each mobile device to prolong the network lifetime. Simulation results ensure the efficiency of the proposed BA-TS algorithm regards throughput and network lifetime over TS and exhaustive search method.