Developing a Novel Pedestrian-Aware Car-following Strategy for Operating Connected and Autonomous Vehicles in Dense Crowds

Abstract

In contemporary urban landscapes, the integration of Connected and Autonomous Vehicles (CAVs) in real contexts is of paramount importance due to the potential of CAVs to transform transportation safety and efficiency. The coexistence of CAVs and pedestrians on the road in crowded environments with pedestrians presents intricate challenges in a real context that necessitates meticulous analysis. Such coexistence is expected in the near future for different purposes. Examples of the purposes include religious congregations such as Hajj, social crowds such as outdoor sports events, political crowds such as mass processions, etc. However, related research studies mostly ignore crowded environments with CAVs and pedestrians. This happens as the existing studies focus on only CAVs, only pedestrians, or both of them in sparse settings.

To address this gap in the literature, this research aims to comprehensively investigate crowded environments with both CAVs and pedestrians and analyze the behavioral dynamics as well as safety implications in crowded environments. Particularly, in this research, we first develop a pedestrian-aware car-following strategy that is applicable to all the existing car-following models in general. Next, we develop a novel pedestrian-aware car-following model, which enhances exploration coverage as well as realizes the pedestrians and vehicles in a new way. The newly proposed model is suitable for diversified crowd settings. To confirm this, we conduct experimental evaluation by performing advanced simulations in a traffic simulator platform named DhakaSim. Here, we modify and enhance the simulator to include different types of pedestrians. Besides, we implement our proposed car-following strategy with several existing car-following models. Additionally, we implement our newly proposed car-following model in the simulator. Afterward, we perform rigorous performance evaluation by comparing the performance of our proposed car-following model against that of other existing alternatives. Our performance evaluation encompasses diversified simulation settings and different real road networks from various parts of the world. Our simulation results demonstrate enhancing on-road experiences through our proposed approach, which ultimately contributes to a safer and more efficient mobility landscape in crowded environments.