Group Reinforcement Learning-assisted Resource Allocation Scheme for Vehicular Edge Networks to Reduce Task Wait Times

Smart transportation systems (STS) and vehicular communication networks (VCNs) rely on mobile edge computing (MEC) devices to provide resource support at the ridge of the vehicles. Heterogeneous vehicular tasks rely on unconditional resource availability to improve service dissemination. Based on different task overloading and application task demands, the need for accurate task management becomes mandatory. This article introduces a Group Reinforcement Learning (GRL)-assisted Queued Resource Allocation (QRA) scheme for improving the STS and VCN communication efficacy. In the proposed scheme, the resources are queued based on their availability for dynamic allocation and offloading. Based on the vehicle's active tasks, the queued resources are allocated; the increase in allocation time performs task offloading to prevent wait time/ task failures. Individual reinforcement agents for the following pairs: queued resources, task allocation, and empty queue, task offloading are generated to improve the efficacy. This efficacy is improved by converting offloading agents towards task allocation using reinforced search. The search for available resources based on low waiting times is used to fill the queue such that the wait interval is reduced. Based on agent groups, the maximum allocation and task completion are performed such that the need for additional wait time is reduced. The proposed scheme improves the task completion ratio by 10.31% and resource allocation rate by 13.48%, reducing the wait time by 11.55% for the maximum number of vehicles considered.