无人驾驶方程式赛车轨迹预测跟踪控制

Trajectory Tracking Control of Driverless Formula Racing Car

为提高车辆轨迹跟踪的稳定性与实时性,并针对传统MPC算法轨迹跟踪精度不足的问题,提出了一种基于LQR求解的多帧预瞄的最优控制方法,并将其应用于FSAC赛车上。首先建立赛车的运动学模型,再将非线性系统作线性化处理,获得所需的线性时变误差模型。结合控制序列递推赛车往后一段时域内赛车状态的预测模型,通过离散参考轨迹函数,获得预测时域内的最佳参考点序列,利用赛车实时状态和控制输入序列构造二次型目标函数,迭代求出预测时域内的最优控制序列,输出车轮转角和赛车纵向速度来实现赛车对期望轨迹的跟踪。最后,通过构建仿真环境测试和实车测试验证算法的跟踪精度和稳定性,算法达到预期的效果。

In order to improve the stability and real-time performance of vehicle trajectory tracking, and solve the problem of insufficient trajectory tracking accuracy of traditional MPC algorithm, an optimal control method of multi-frame preview based on LQR solution is proposed and applied to FSAC racing car. Firstly, the kinematic model of the racing car is established, and then the nonlinear system is linearized to obtain the required linear time-varying error model. Combined with control sequence of state prediction model within a time domain recursive racing back racing, by the discrete reference trajectory function, the best reference point is obtained to predict time domain sequence. The quadratic objective function is constructed by using the real-time state of the racing car and the control input sequence, the optimal control sequence in the prediction time domain is obtained by iteration, and the wheel angle and longitudinal velocity of the racing car are output, so that the racing car can track the desired trajectory. Finally, the tracking accuracy and stability of the algorithm are verified by constructing simulation environment test and real car test, and the algorithm achieves the expected effect.