Risk-aware unsignalized intersection management in unstructured mixed-traffic environment: a real-time hierarchical safety evaluation method

The mixed operation of connected automated vehicles and connected manual vehicles has reshaped overall transportation management. Managing unsignalized intersections in unstructured mixed-traffic environment presents inherent safety challenges, arising from a series of factors such as human-driving uncertainties, irregular road boundaries, and uneven terrain. To this end, this study proposes a risk-aware management method that determines the right-of-way assignment strategy from a set of candidates, thereby optimizing passing orders for enhanced operational safety and efficiency. Rather than directly solving the mixed-integer optimization inherent in right-of-way assignment, the proposed method adopts a simulation–evaluation paradigm to determine a near-optimal strategy. Central to this method is a hierarchical architecture designed to assess the safety performance of each candidate right-of-way strategy. Specifically, the upper layer generates simulated trajectories under candidate strategies and the lower layer constructs a spatiotemporal risk field along these trajectories to quantify hazards arising from the unstructured road environment and vehicles, capturing potential conflicts and human-driving uncertainties. This risk-aware management method yields both instantaneous and long-term risk metrics. By employing simulated trajectories as an intermediary, the abstract right-of-way assignment task can be transformed into quantifiable safety assessment of candidate strategies. Following the determination of the near-optimal safety strategy, this study further provides efficient vehicle passing trajectories and a real-time enhancement method. Numerical experiments help validate the effectiveness of the proposed method.