Tour Path Scheduling Using Optimized Deep Reinforcement Learning for IoT Mobile Data Collectors

ABSTRACT In Internet of Things (IoT)–based wireless sensor network (WSN), mobile data collectors (MDCs), which move over various geographic regions to transport data from sensors to access points, are thought to be a more effective way than the traditional data collection techniques employing static sinks. The direct transfer of data from all sensors to the base station would be inefficient given the energy constraints on the sensor node. This is a result of the data redundancy brought about by nearby sensors' relatively strong correlation. Moreover, base stations are unable to handle the enormous volumes of data produced by a larger sensor network. In order to integrate data and generate meaningful information at sensors or intermediate nodes, specific networks are therefore needed. In this paper, tour path scheduling using optimized deep reinforcement learning (DRL) for MDCs in IoT‐WSN. The DRL algorithm schedules the visiting pattern of the MDCs based on the type of IoT sensors and their data generation rates. To accelerate the convergence speed of DRL, sunflower optimization (SFO) algorithm is used. Then, optimum tour paths are determined using Capuchin Search Algorithm (CapSA) based on path stability and data collection latency. Simulation results have shown that DRL–SFO–CapSA minimizes the data collection delay and packet drop while maximizing the packet delivery ratio and residual energy.