【摘要】 在弯道等大曲率场景中，车辆转向系统的迟滞和车辆模型的线性化会导致转向不足和稳态误差，从而影响自主车辆路径跟踪精度和响应速度。为了解决这一问题，提出了一种路径跟踪框架。该框架在弯道等大曲率场景，触发前馈控制控制器，输出理想转角序列，提前使转向机构到达最优转角附近；随后将引入排斥目标函数的非线性模型预测控制器优化求解出的最优控制序列作用于车辆，刷新理想转角序列。搭建自主车辆实验平台，在不同场景下进行仿真验证，结果表明，与忽略滞后的传统模型预测控制相比，前馈非线性模型预测控制器跟踪精度和响应速度方面的性能有所提高。特别是在弯道等大曲率场景中，所提出的框架将横向均方根误差降低了近30%。更多还原

【Abstract】 In high-curvature scenarios like sharp bends, the hysteresis in the vehicle’s steering system and the linearization of the vehicle model may cause insufficient steering and steady-state errors, thereby impacting the precision of autonomous vehicle path tracking and its response speed. To address this issue, we introduce a novel path tracking framework, which triggers a feedforward control controller to generate an ideal steering angle sequence, proactively guiding the steering mechanism to approach the optimal steering angle in advance. Subsequently, a nonlinear model predictive controller incorporating a repelling target function optimally determines the best control sequence, which is then applied to the vehicle to update the ideal steering angle sequence. An autonomous vehicle experimental platform is built, and simulation verification is conducted in various scenarios. Our results indicate that, in comparison to traditional model predictive control methods that disregard hysteresis, the feedforward nonlinear model predictive controller improves the tracking accuracy and response speed. In high-curvature scenarios in particular, our framework reduces the lateral root mean square error by nearly 30%.更多还原