Optimized traffic signal control system incorporating mixed traffic flow and adverse weather

As urban transportation networks have become increasingly complex and diverse, traditional signal control methods struggle to effectively manage dynamic traffic flows and adverse weather conditions. To tackle this issue, this paper introduces an improved traffic signal control system leveraging a Double Dueling Deep Q-Network algorithm, referred to as the Self-Adaptive Attention Double Dueling Deep Q-Network (SAA-D3QN) framework. This system not only considers mixed traffic flow and an adaptive attention mechanism but also introduces a novel reward function concept specifically designed for adverse weather conditions. Incorporating a mixed traffic flow model, the system can more precisely simulate the behavior of different vehicle types under various traffic conditions. The introduction of the adaptive attention mechanism enables the system to dynamically adjust its focus on critical areas when processing large amounts of traffic data, allowing for rapid identification and processing of key information. In addition, this paper conducts an in-depth analysis of traffic data under adverse weather conditions and propose a new reward function to enable the traffic signal system to adaptively adjust signal timing strategies under such circumstances. The experimental findings indicate that compared to traditional signal control methods, the SAA-D3QN traffic system significantly reduces average vehicle waiting time, enhances intersection throughput, and decreases traffic congestion.