基于风险建模的变曲率匝道自动驾驶轨迹规划方法

Trajectory Planning for Autonomous Vehicles on Ramp Scenarios with Gradual Curves Based on Risk Modeling

摘要:

变曲率匝道曲率不固定、平面布局多样，是自动驾驶面临的复杂道路场景。提出了一种考虑“车-路”耦合风险的换道算法，用以优化变曲率匝道下换道轨迹的安全性。算法以五次多项式为基础建立候选换道轨迹集，综合风险、效率和舒适性指标构建代价评估函数，实时执行代价最低的轨迹。算法在直接汇入式和平行布置式加速车道布局下进行了仿真测试，结果表明在保证效率的前提下，车辆间的冲突风险程度在加速车道直接汇入式和平行布置式两种布局下分别降低13.9%和12.6%，可以改进自动驾驶轨迹在复杂驾驶场景的安全性。

Abstract:

Ramp scenarios with gradual curves challenge the autonomous vehicles because of their irregular curvatures， diverse planar layouts， and multidimensional vehicle conflicts. This paper presents an algorithm based on the coupled "vehicle-road" risk to increase the safety of lane-changing on ramps with gradual curves. The quintic polynomial is adopted to construct a set of candidate lane-changing trajectories. The cost function is developed by risk， efficiency， and comfort indicators. Simulation tests under parallel continue curved and tapered continue curved ramps show that the proposed model improved safety performance by 13.9 % and 12.6 % under the two ramp configurations compared to earlier lane-changing algorithms based on collision detection and rule-based risk evaluation. These results showed the proposed algorithm can be applied to complicated driving scenarios to increase the trajectory safety of autonomous vehicles.