Making overtaking cyclists safer: Driver intention models in threat assessment and decision-making of advanced driver assistance system

Licentiate thesis, 2021

Introduction: The number of cyclist fatalities makes up 3% of all fatalities globally and 7.8% in the European Union. Cars overtaking cyclists on rural roads are complex situations. Miscommunication and misunderstandings between road users may lead to crashes and severe injuries, particularly to cyclists, due to lack of protection. When making a car overtaking a cyclist safer, it is important to understand the interaction between road users and use in the development of an Advanced Driver Assistance System (ADAS). Methods: First, a literature review was carried out on driver and interaction modeling. A Unified Modeling Language (UML) framework was introduced to operationalize the interaction definition to be used in the development of ADAS. Second, the threat assessment and decision-making algorithm were developed that included the driver intention model. The counterfactual simulation was carried out on artificial crash data and field data to understand the intention-based ADAS's performance and crash avoidance compared to a conventional system. The method focused on cars overtaking cyclists when an oncoming vehicle was present. Results: An operationalized definition of interaction was proposed to highlight the interaction between road users. The framework proposed uses UML diagrams to include interaction in the existing driver modeling approaches. The intention-based ADAS results showed that using the intention model, earlier warning or emergency braking intervention can be activated to avoid a potential rear-end collision with a cyclist without increasing more false activations than a conventional system. Conclusion: The approach used to integrate the driver intention model in developing an intention-based ADAS can improve the system's effectiveness without compromising its acceptance. The intention-based ADAS has implications towards reducing worldwide road fatalities and in achieving sustainable development goals and car assessment programs