Behavioural models explaining and predicting transparent negotiations between AVs and human road users : Deliverable 2.2 in the EC ITN project SHAPE-IT

Report, 2023

One key enabler for the development of automated vehicles that can coexist and negotiate with human road users in a safe, transparent, and human-acceptable manner is the development of mathematical models of human behaviour in the relevant interactive scenarios. Such models are needed as components both in real-time AV algorithms and in simulation tools for virtual AV testing. However, many important questions about how to model human road user interactions remain unanswered, and five of the Early Stage Researchers (ESRs) in the SHAPE-IT project have targeted such research questions. This report provides an overall introduction to the area of human-AV interaction modelling and summarises the ESRs’ research and findings in this area, including links to the ESRs’ peer-reviewed papers and preprints (providing full details). The SHAPE-IT modelling research by these ESRs has spanned a broad spectrum of modelling approaches and modelling use cases and has generated results such as: More computationally effective and transferable algorithms for real-time prediction of pedestrian movement; novel insights about human driver communication and behaviour during lane changes; a general model of human AV passengers’ subjective perception of traffic risk; a demonstration that models of vehicle-pedestrian interactions based on behavioural game theory outperform conventional game theory models; and novel insights about how cyclists’ head, eye, and pedalling behaviour predict cyclist-vehicle interactions. The research by these five ESRs has increased our understanding of, and capability to predict and simulate, how humans interact in traffic, with direct relevance for development of safe and human-acceptable AVs.