Autonomous vehicle control method under pedestrian risk constraints on road segment

To address frequent autonomous vehicles (AVs) stops and operational inefficiency caused by conservative pedestrian yielding, this study proposes a collaborative AVs control method under pedestrian risk constraints. A dynamic pedestrian risk assessment model is developed, integrating lane position, relative distance/speed, pedestrian behavioral factors (e.g., waiting-time-induced impatience), and real-time risk perception. This model informs risk-threshold-based motion planning constraints, enabling coordinated AVs speed/position adjustments to proactively create pedestrian crossing windows. A 36,000-second SUMO simulation demonstrates significant improvements: In high-density scenarios (15-second pedestrian intervals), the method reduces average pedestrian risk to 0.21 (vs. 0/0.27 in conservative/Gupta approaches), boosts AVs throughput by 77% (11,857 crossings), and slashes delays by 98% (8.7s). At 30-second intervals, it achieves 0.16 risk (24% reduction), 72% higher efficiency (12,000 crossings), and 5.6s delays (99% improvement). Results validate superior performance in balancing safety and traffic flow optimization for urban mixed-traffic systems.