Dual-network proportional variable speed limit control for mixed traffic flows with multi-level automated vehicles

Existing variable speed limit (VSL) research mainly focuses on mixed traffic flows composed of fully automated vehicles (Fully-AVs) and human-driven vehicles (HVs). However, semi-automated vehicles (Semi-AVs) have seen an increasing penetration rate in current traffic flows, and their automated driving (AD) disengagement strategy considerably impacts VSL. Research on optimal VSL control in mixed traffic flow with the coexistence of multi-level automated vehicles is limited. To address this gap, this study first analyzes the driving characteristics of vehicles at different automation levels under VSL scenarios. A dual-network Proportional Variable Speed Limit (PVSL) control method is then developed to mitigate frequent speed limit fluctuations and reduce the risk of AD disengagement. The PVSL framework integrates a global policy generator that determines the optimal speed limit and the proportion of vehicles to be regulated, and a local vehicle selector that assigns selection scores to each AV based on its microscopic vehicle-level features. These selection scores reflect the AVs’ priority for receiving speed limit control. Using both the selection scores and the speed limit ratio, the local vehicle selector chooses a top-ranked subset of AVs to receive PVSL commands. Finally, a traffic bottleneck scenario is simulated to test the performance of the PVSL controller. In the benchmark test, the PVSL reduces traffic delays by 53.1%, emissions by 21.4%, and fuel consumption by 14%. Furthermore, as the proportion of high-level AVs in mixed traffic flow increases and the AD disengagement rate decreases, traffic performance indicators, such as traffic delays, have shown a significant improvement.