August 1, 2025

A Deep Reinforcement Learning Approach for Dynamic Resource Allocation in VANETs With Human–Centric Interaction Interfaces

Key Points

The proposed method achieves a 92.8% packet delivery ratio, significantly enhancing resource allocation efficiency and reducing congestion.
In comparison with existing models, this approach demonstrates a latency of 12.5 ms and a throughput of 8.7 Mbps, highlighting better performance metrics.
Implementing a double deep Q-network alongside an adaptive artificial fish swarm algorithm optimizes decision-making in resource allocation for VANETs.
This approach suggests a promising direction for future communication protocols in the smart transportation system domain, enhancing overall efficiency.

Abstract

ABSTRACT Vehicular ad hoc networks (VANETs) play an important role in smart transportation systems (STS) by providing various multimedia and safety services to drivers, passengers/customers, and pedestrians. As the existing wireless communication protocols (WCPs) face difficulties in scalability and efficiency, there is a critical need for the growth of next‐generation communication protocols in VANETs. This research proposes a novel approach, adaptive artificial fish swarm algorithm driven double deep Q‐network (AAFSA‐DDQNet) for dynamic resource allocation in VANETs, and it also focuses on human–centric interaction interfaces (HCIIs). The main aim of this research is to enhance resource allocation while ensuring efficient data transmission in VANET environments. The proposed method integrates AAFSA and DDQNet to address the challenges of decreasing data collision and enhancing backoff distribution in the network. The AAFSA is used for enhancing global search capability, while DDQNet is employed for making optimal decisions regarding resource allocation. The control and service channel intervals are adjusted to improve network performance. The approach is implemented using MATLAB. The proposed model is compared with Optimized Reinforcement Learning with Adaptive Coati Optimization (ORL‐ACO) and DQRNN, demonstrating superior performance: 92.8% packet delivery ratio (PDR), 12.5 ms latency, 8.7 Mbps throughput, and a 7.2% collision rate, significantly outperforming existing models in terms of bandwidth utilization, computational efficiency, and transmission reliability. Finally, the proposed AAFSA‐DDQNet‐based approach offers a promising solution for dynamic resource allocation in VANETs, leading to enhanced communication efficiency and reduced congestion.

Ask AI

Helpful

Bookmark

View Full Paper