Computation Offloading and Resource Allocation in IoT-Based Mobile Edge Computing Systems

With the rise in popularity of artificial intelligence (AI) and internet of things (IoT) technologies, advanced AI technologies have been widely applied to support delay/time-sensitive tasks of IoT-based user equipment (UE) in IoT systems, which allows IoT-based UEs to offload their tasks to a remote fog, edge or cloud computing server. To reduce the consumption of delays (which may include transmission delays, queueing delays, and processing delays) while efficiently allocating the computation resource at a remote server, an efficient offloading decision solution needs to be proposed. In this paper, an IoT-based network system consisting of two layers will be proposed, where The bottom layer is the IoT-based UE layer, which includes multiple IoT-based UEs, and the top layer is the mobile edge computing (MEC) layer, which includes an edge node embedded with the base station. We propose a double Q-Learning-based offloading decision and computation resource allocation optimisation algorithm (DQOCA), which aims to jointly optimise the offloading decisions among all IoT-based UEs and optimise computation resource at the MEC server to reduce the maximum delay consumption among all IoT-based UEs. Simulation findings show that, in comparison to benchmarks (i.e., local processing and edge processing schemes), our proposed approach greatly minimises the maximum delay consumption.