Developing an Innovative Platform for Modeling Active Road User Interactions and Safety: Integration of Computer Vision, Agent-based, and Machine Learning Models

促进步行和骑自行车等积极交通方式在交通工程中受到越来越多的重视，因为它们是交通可持续发展的重要贡献者，对公众健康和环境有几个好处。然而，经常出行的人是最脆弱的道路使用者，因为他们因道路碰撞而死亡或受重伤的风险更高。迫切需要对积极的道路用户互动行为有一个坚实的了解，并采用创新的技术来提高其安全性。因此，本研究计划提出了一个创新的平台，用于分析和建模活跃的道路用户交互，并提高其安全性。拟议的平台涉及三个相互关联的重点领域。重点领域I涉及获取各种活跃的道路用户数据集(主要是在感兴趣的地点收集的视频数据和GPS数据)，并应用包括机器学习技术和计算机视觉在内的创新分析工具来分析收集的数据。该分析将调查道路使用者相互之间的活跃互动、与机动车辆的互动(冲突)以及道路使用者的违规行为(乱穿马路和时间违法)，并探讨这些行为对道路安全的影响。 在焦点区域II，将开发一种新的基于智能体的微观仿真模型，用于建模详细的道路用户交互。该模型以人工智能原理为基础，采用基于智能体的建模方法来模拟主动的道路用户交互。该模型将为解决与道路使用者安全相关的各种应用和评估现有交通设施的服务水平提供强大的工具。拟议的模型将使不同的设计备选方案能够在对现有条件做出任何改变之前进行分析，并调查可能难以使用实际数据探索的情景。重点领域三涉及通过在宏观和微观两个层面上对碰撞和交通冲突数据进行先进的统计建模，制定一个全面的框架来分析现行道路使用者的安全。此外，研究还将建立道路使用者碰撞和冲突之间的统计关系。 拟议的研究将创建强大的分析工具，为所有年龄和能力的活跃道路使用者设计安全的交通设施提供支持。拟议的平台将帮助交通工程师和规划者促进和适应积极的交通活动，这有助于公众健康、安全和减少温室气体排放。它还将使我们能够在早期规划阶段考虑道路使用者的安全。拟议平台的采用预计将直接通过改善道路安全或间接通过改善空气质量和增加人口的体力活动水平来拯救数千人的生命。

Promoting active modes of transportation, such as walking and biking, is receiving more emphasis in transportation engineering as they represent vital contributors to transportation sustainability and have several benefits to public health and the environment. Nevertheless, active travellers are the most vulnerable road users, as they are at a higher risk of being killed or severely injured due to road collisions. There is an urgent need to develop a solid understanding of active road user interaction behaviour and adopt innovative techniques to enhance their safety. As such, an innovative platform for analyzing and modeling active road user interactions and enhancing their safety is proposed in this research program. The proposed platform involves three interrelated focus areas. Focus Area I involves the acquisition of a variety of active road user datasets (mainly, video data collected at locations of interest and GPS data) and applying innovative analytical tools, including machine learning techniques and computer vision, to analyze the collected data. The analysis will investigate active road user interactions with each other, interactions with motorized vehicles (conflicts), and road user violation behaviour (jaywalking and temporal violations), and explore the implications of these behaviours on road safety. In Focus Area II, a novel agent-based micro-simulation model for modeling detailed road user interactions will be developed. The model relies on artificial intelligence principles and utilizes the agent-based modeling approach to simulate active road user interactions. The model will provide a powerful tool for addressing a variety of applications related to road user safety and assessing the level of service of active transportation facilities. The proposed model will enable the analysis of different design alternatives before making any changes to existing conditions and investigating scenarios that might be difficult to explore using actual data. Focus Area III involves developing a comprehensive framework for analyzing active road user safety through the advanced statistical modeling of collision and traffic conflict data on both the macro and the micro levels. In addition, the research will establish the statistical relationship between road user collisions and conflicts. The proposed research will create powerful analytical tools to support the design of safe transportation facilities for active road users of all ages and abilities. The proposed platform will aid transportation engineers and planners to promote and accommodate active transportation activities, which contributes to public health, safety, and reducing GHG emission. It will also enable to consider road user safety in the early planning stages. The adoption of the proposed platform is expected to save thousands of lives, either directly by improving road safety or indirectly by improving air quality and increasing the physical activity level of the population.