PSO ‐ FRL : A Hybrid Optimization and Adaptation Approach for Task Scheduling in Edge Enabled Sensor Networks

Modern Edge Computing (EC) platforms are designed to manage a large amount of sensor data from different sensor networks deployed on the edge. However, the network parameters of the Sensor Networks (SNs) may not necessarily be aligned with the design. Principles of legacy networks. Most of the existing works are designed to address either the resource efficiency or the performance aspects alone. In this article, we propose a hybrid approach, namely, Particle Swarm Optimization‐based Fuzzy Reinforcement Learning (PSO‐FRL) method that leverages the optimization capabilities of the PSO method and the adaptability features of the Reinforcement Learning (RL) method along with the fairness features of Fuzzy Logic. The proposed hybrid method efficiently designs the task scheduling and resource allocation of Edge Computing‐based SNs, which is a complex optimization problem. Using the proposed hybrid method, the sensor networks resource efficiency is improved and the task scheduling is optimized that lead to the improvements in throughput, Quality of Service (QoS), spectral efficiency and network lifetime, by achieving an average task scheduling efficiency of 98% over the PSO alone and RL alone. Hence, the proposed design is a suitable trade‐off between short‐term performance gains and long‐term adaptability. It is also scalable and is expected to support the large number of next generation Internet‐of‐Things (IoT)‐enabled Sensor Networks for use cases in smart city, industrial automation, and environmental monitoring, and so forth.