Toward an Enhanced Risk Assessment Sensitivity for Autonomous Vehicles with the Safety Potential Field Approach

Ensuring safety encodes its most important position in autonomous vehicle (AV) technology, with a growing awareness on the role of risk assessment for advancing AV safety. Traditional risk assessment metrics often neglect crucial lateral interactions, struggle to reflect risk evolution, and find it challenging to capture comprehensive risks from multiple objects. Although recent literature mitigated these shortcomings through the safety potential field approach, their methods still exhibit lack in sensitivity to relative motions and inaccurate aggregation of risks from various objects. This paper intro-duces a novel real-time risk assessment metric for AVs that harnesses the instantaneous increment of vehicular potential energy within the safety potential field, namely instantaneous increment of potential energy (IIP). This metric enhances the sensitivity of risk assessment for AVs by capturing relative motions and buying more response time by capturing the progress of risk formation. The performance of IIP are validated by multiple simulation-based case studies and showed superiority over conventional time-to-collision (TTC) type of metrics in detecting lateral risks. The results underscore the potential of proposed metric to predict hazards and guide AV actions, enhancing safety in AV driving environment.