ReLM: An Efficient Reinforcement Learning-Based Localization Algorithm for Mobile Wireless Sensor Networks

Internet of Things (IoT) based on mobile wireless sensor networks (MWSNs) is a promising paradigm for future networks. They are poised to revolutionize the entire technology landscape owing to the versatility of their applications, which range from smart agriculture to highly automated medical deliveries. However, it is very common that the sensor network topology is frequently evolving as the sensors are mobile and constantly moving. Then, their positions change over time. Therefore, the development of efficient localization schemes that support the topology changes and mobility of sensors remains a challenge. In this paper, we proposed a novel localization algorithm based on reinforcement learning (ReLM) to address the localization problem in MWSNs. Rather than using additional hardware, our proposed algorithm relies on local information and does not require any prior knowledge of the network to estimate the position of the sensors. It mainly provided a model that allows the mobile sensors to predict the best direction and position. We formulate the localization problem as an optimization problem that will be solved using reinforcement learning, which is based specifically on the Q-learning technique and the Boltzmann strategy. Hence, ReLM combines the localization of mobile nodes by the process of finding an adaptive policy to estimate their positions to minimize the localization error. Simulation results showed that our algorithm significantly outperformed the MCL algorithm and provided a lower position estimation error.