Enhancing safety assessment for fully autonomous vehicles: a role for agent-based modelling

The advent of fully autonomous cars-level 5 automation, according to the Society of Automotive Engineers (SAE)-represents a paradigm shift in the entire mobility landscape. This article explores the critical aspect of safety evaluation for these vehicles, essential for their societal integration. In fact, ensuring passengers, pedestrians, and road users’ safety is crucial for adopting this technology. A major challenge for this safety evaluation is selecting suitable modelling tools to assess the complex dynamics of autonomous systems. This research rigorously evaluated various modelling approaches based on accuracy, scalability, and flexibility, finding Agent-Based Modelling (ABM) to be the most effective. Through ABM it is possible to simulate individual agents representing vehicles, pedestrians, and infrastructure, allowing realistic scenario simulations and safety performance evaluations. Analysing this tool, the aim is to highlight the various dynamics of autonomous driving systems and their implications for the future of transportation safety. Insights from this study inform policy, industry practices, and future research on autonomous vehicle safety, contributing to the ongoing discourse on their safety and reliability.