MSND: Modified Standard Normal Deviate Incident Detection Algorithm for Connected Autonomous and Human-Driven Vehicles in Mixed Traffic

Abstract

Advances in IoT and IoV technology have made connected autonomous vehicles (CAVs) data sources. Using CAVs as data sources and in incident management algorithms can create faster, more reliable, and more effective algorithms. This paper proposes a modified standard normal deviation (MSND) incident detection algorithm that uses CAVs as data sources and considers multiple traffic parameters. MSND is utilized in conjunction with two other incident detection algorithms, Standard Normal Deviation (SNS) and California (CAL), in a method of incident management known as Variable Speed Limits (VSL). SUMO Traffic Simulation Software is used to evaluate the effectiveness of the proposed method. A 10.4-kilometer road network is developed. Numerous scenarios are simulated on this road network, with variables including traffic demand, autonomous vehicle penetration rate, incident location, incident length, and incident lane. On the effectiveness metrics of detection rate, false alarm rate, and mean time to detect, simulation results demonstrate that the proposed method outperforms the SND and California methods. In terms of detection rate, the MSND algorithm performs the best, with a 12.27% improvement over the SND algorithm and a 21.99% improvement over the California method. After integrating all incident detection algorithms with the VSL traffic management method and simulating each combination, it was determined that the MSND-VSL integration reduced average density in the critical region by 19.73 percent, followed by SND-VSL with a 13.94 percent reduction and CAL-VSL with a 9.9 percent reduction.