基于混合域优化控制的智能网联车辆运动规划模型

An Optimal Control Based Motion Planner in Mixed-domain

智能网联车辆具备提高交通安全与效率、降低能耗的巨大潜力。作为智能网联车辆决策控制的重要环节,运动规划对于智能网联车辆的循迹精度、控制效果具有显著影响。为了提高智能网联车辆控制精度,提出了一种智能网联车辆运动规划模型。该模型以追踪参考路径为目标,基于时空混合域的优化控制方法,避免了轨迹追踪过程中横向控制掺杂纵向误差的影响,提高了模型控制精度。通过考虑车辆动力学、转向传动系统动态和底层控制时延,该模型可规划车辆纵向运动指令(加速度)、横向运动指令(方向盘转角),并确保运动规划指令能够被车辆底层控制准确执行。最后,所提出的运动规划模型基于动态规划原理求解,提高了求解效率。对所提出的运动规划模型通过PreScan和Carsim联合仿真进行测试,结果显示该运动规划模型在直线换道、曲线巡航、U-turn多种场景下,循迹误差均小于9 cm,速度误差均小于0.7 km·h^(-1)。此外,该运动规划模型在实车测试中也展现了极佳的控制效果。实车测试结果表明:该运动规划模型在60 km·h^(-1)以下的直行、交叉口转弯、曲线巡航、换道等多种场景中,车辆循迹误差均小于12.5 cm,巡航速度误差小于3.32 km·h^(-1),其中,期望速度20 km·h^(-1)的直行及半径小于15 m的转弯场景下的循迹误差为11 cm,相较当前最前沿的实车测试研究,循迹精度提高了27%。此外,该运动规划模型在实车测试中的平均计算速度为5.15 ms,这说明该模型已具备实际落地应用能力。

Connected and Automated Vehicle(CAV)has the potential to improve transportation safety,mobility,and reduce fuel consumption.As a key building block of the decision-making of CAV,motion planning significantly affects tracking accuracy and control performance.In order to improve the control accuracy of CAV,a CAV motion planner was proposed in this study.With the objective of tracking a reference route,the proposed motion planner is based on time and space-mixed domains.This prevents the lateral control accuracy from being affected by the longitudinal control error and improved control accuracy too.This planner plans longitudinal motion command(acceleration)and lateral motion command(steering angle)with the consideration of vehicle dynamics,steering transmission dynamics,and control delay.This ensures that motion commands could be fulfilled accurately by vehicle local control.Finally,the proposed motion planner is solved through a dynamic programming-based method which improved solving efficiency.The proposed motion planner was tested by a PreScan and Carsim joint simulation.The results showed that the tracking and speed errors are less than 9 cm and 0.7 km·h^(-1),respectively in lane-change scenarios on straight roads,cruising scenarios on curves,and U-turn scenarios.In addition,the proposed motion planner performs well in the field test.Field test results showed that the tracking and speed errors are less than 12.5 cm and 3.32 km·h^(-1),respectively in all scenarios with desired speed under 60 km·h^(-1),including straight cruising,turning at intersections,curve-cruising,and lane-changing.In scenarios of straight cruising and turning with a radius smaller than 15 meters at 20 km·h^(-1)desired speed,the proposed motion planner improves 27%of tracking accuracy compared with that of a state-of-the-art planner in the field test.Moreover,the average computation time of the proposed motion planner is 5.15 milliseconds in the field test.It indicates that the proposed motion planner is with field implementation capability.