Vehicular Fog Resource Allocation Approach for VANETs Based on Deep Adaptive Reinforcement Learning Combined With Heuristic Information

Abstract

Intelligent Transport Systems (ITS) are gradually progressing to practical application because of the rapid growth in network and information technology. Currently, the low-latency ITS requirements are hard to achieve in the conventional cloud-based Internet of Vehicles (IoV) infrastructure. In the context of IoV, Vehicular Fog Computing (VFC) has become recognized as an inventive and viable architecture that can effectively decrease the time required for the computation of diverse vehicular application activities. Vehicles receive rapid task execution services from VFC. The benefits of fog computing and vehicular cloud computing are combined in a novel concept called fog-based Vehicular Ad Hoc Networks (VANETs). These networks depend on a movable power source, so they have specific limitations. Cost-effective routing and load distribution in VANETs provide additional difficulties. In this work, a novel method is developed in vehicular applications to solve the difficulty of allocating limited fog resources and minimizing the service latency by using parked vehicles. Here, the improved heuristic algorithm called Revised Fitness-based Binary Battle Royale Optimizer (RF-BinBRO) is proposed to solve the problems of vehicular networks effectively. Additionally, the combination of Deep Adaptive Reinforcement Learning (DARL) and the improved BinBRO algorithm effectively analyzes resource allocation, vehicle parking, and movement status. Here, the parameters are tuned using the RF-BinBRO to achieve better transportation performance. To assess the performance of the proposed algorithm, simulations are carried out. The results defined that the developed VFC resource allocation model attains maximum service satisfaction compared to the traditional methods for resource allocation.

Keywords

• Vehicular fog resource allocation
• vehicular ad hoc networks
• revised fitness-based binary battle royale optimizer
• deep adaptive reinforcement learning
• reward assessment
• service satisfaction