Abstract: In the mixed traffic stream of interacting human-driven and automated vehicles (AVs), compliance with traffic rules is crucial for the safe operation of AVs. Existing rules, however, are designed for human drivers, and the vagueness of some of these rules poses challenges for AVs. Insufficient attention has been given to how these rules are interpreted by AVs, leaving it unclear which rules AVs will struggle to understand and how to improve them. This ambiguity has led to a lack of consideration of quantified traffic rules in AV trajectory planning methods. This study therefore reviewed traffic rules from six countries/regions: the State of California and New York in the United States, Germany, China, Queensland in Australia, and the United Kingdom. Based on PEGASUS scenario layers, a coding table was proposed to digitize intersection traffic rules, incorporating formalization and quantification to determine which specific rules needed attention. Focusing on the left-turn traffic rules at signalized intersections, problematic rules were improved and the applicable digitized rules were combined into a trajectory planning method. Responsibility-Sensitive Safety was embedded into the sampling trajectory planning model as a safe distance constraint based on dynamic occupancy grid map. Using real traffic data, the validity of the model was verified by simulation of the left-turn scenario’s five maneuvers. The results show that there are many unquantified rules in existing traffic rules, particularly governing different maneuvers. Additionally, the findings suggest that more safety constraints do not necessarily result in better traffic rule performance, especially for static parameters. Instead, dynamic safety distances can effectively improve efficiency and ensure safety. Furthermore, the proposed rule-based trajectory planning method ensures efficiency and comfort while significantly improving safety. Based on each maneuver tested during simulation, the method changed serious conflict to non-conflict and reduced the average collision probability by 25.51%-86.15%.
Shi Ruolin, Xuesong Wang*. Digitizing traffic rules to guide automated vehicle trajectory planning. Expert Systems with Applications , 2025: 126661.