针对高速自动驾驶车辆实时高精度的运动控制问题,提出一种上层为基于路径点Cost的路径点筛选器与基于横纵向轮胎力分析的速度规划器、下层为基于线性时变动力学模型预测的路径跟踪控制器与速度控制器的两层架构,并引入最小均方(Least Mean Square,LMS)自适应状态估计器提升系统的抗噪性.路径点筛选器提升运算速度并减少筛选过程中的关键信息损失,速度规划器在安全行驶前提下生成最优速度曲线.路径跟踪控制器考虑跟踪偏差软约束,提升跟踪效果.LMS状态估计器基于在线矫正的动力学模型,对横摆角速度与横向速度在线估计.搭建dSPACE-TX2硬件在环仿真环境,在高速公路工况及双移线工况下对比所提出方案与传统运动跟踪控制.半实物仿真结果表明,所提出的运动控制架构提升了抗噪性能与21％的跟踪精度,且满足50 Hz高频控制的要求.
LTV-MPC-based Real-time and Anti-noise Motion Control for High-speed Vehicle
A two-layer control architecture is proposed for the real-time high-precision motion control of high-speed autonomous vehicles,in which an upper layer consists of a path point filter based on a path point Cost function and a velocity planner based on tire force analysis in both lateral and longitudinal directions,and a lower layer consists of a path tracking controller predicted by a linear time-varying dynamic model and a velocity controller.A least mean square(LMS)adaptive state estimator is introduced to enhance the system's noise immunity.The path point filter improves the operation speed and reduces the loss of crucial information during selection,and the velocity planner generates the optimal speed curve under the premise of safe driving.The path tracking controller considers the tracking deviation soft constraint to improve the tracking effect.The LMS state estimator estimates the lateral velocity and yaw rate online based on an online-corrected dynamic model.A dSPACE-TX2 hardware-in-the-loop simulation environment is constructed,and the proposed path tracking architecture is compared with traditional motion tracking control under high-speed and double lane change scenarios.The hardware-in-the-loop simulation results demonstrate that the proposed motion controlling architecture improves the noise immunity and the tracking accuracy of 21％while meeting the 50 Hz high-frequency control requirement.
linear time-varying modelmodel predictive controlreal-time motion controlpath trackingquadratic programming
万方数据
维普
