Future mobility systems are expected to incorporate a broad range of transport modalities (passenger cars, truck platoons, etc.) at different automation levels (SAE Levels 3/4/5). During operation, automated vehicles will have to independently take safety-critical decisions (e.g., when to brake or change lanes) and estimate the impact of their behavior on the surrounding traffic, thus balancing individual and group safety. To achieve this, automated vehicles will require a quantitative metric of safety to guide their actions.
This article proposes one such metric, suitable for decision-making and autonomous navigation. The metric is meant to provide a quantification of the risk a vehicle incurs during operation by taking into account three main aspects of its operation: the probability of a hazard occurring (e.g., a rear-end collision), the potential impact of the driving conditions on the health of the vehicle’s passengers were the hazard to occur, and the capability of the vehicle to avoid the hazard. The article focuses on introducing the conceptual aspects of the metric first and then presents the initial results on estimating and collision probabilities. The other two aspects will be addressed elsewhere.