Fair channel allocation in IEEE 802.11p for high throughput and low-latency

Abstract

The emergence of Intelligent Transportation Systems (ITS) necessitates reliable and efficient connectivity among vehicles. The IEEE 802.11p standard with one control channel and six service channels operating in 75-MHz spectrum at 5.9 GHz is responsible for fair channel access for V2X communication. Nevertheless, issues like latency and network congestion continue to exist, requiring dynamic channel allocation techniques. To ensure proper operation of Vehicular Ad hoc NETworks (VANETs), Medium Access Control (MAC) plays a vital role. This paper proposes an efficient channel allocation algorithm for the MAC layer of VANET to overcome the stringent demand of throughput and delay. The channel allocation policy adapts to the dynamic vehicular environment. The paper focuses on time slot allocation for fair channel access by avoiding transmission collision between two RSUs. Then the channel allocation is designed as a knapsack problem, where the packets are given priority to access the channel based on their weight factor. A Deep Reinforcement Learning (DRL) Asynchronous Advantage Actor Critic (A3C) algorithm is used to solve the knapsack problem. By utilizing the A3C algorithm, optimal policy is achieved that learns the environment for channel allocation, enabling real-time adaptations to varying network conditions and vehicular mobility patterns. The algorithm handles high-dimensional state and action spaces, allowing for improved decision-making based on current channel utilization and packet prioritization. The proposed framework presents a high-throughput, low-latency channel allocation model that effectively addresses the stringent demands of both safety and non-safety packets, ensuring timely transmission of critical messages. Extensive simulation results prove the efficacy of the proposed algorithm High Throughput Low Latency- Actor-Critic MAC (HTLL-ACMAC) over existing algorithms. The performance evaluation demonstrates that the proposed model reduces the delay by approximately 13%, and maximizes the network throughput by approximately 38% compared to baseline models.