基于牛顿-拉夫逊算法求解的MPC避障路径规划  

MPC obstacle avoidance path planning based on newton-raphson algorithm solving

作　　者：王耀 李占龙[1] WANG Yao;LI Zhanlong(College of Electronic Information Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,Shanxi,China)

机构地区：[1]太原科技大学电子信息工程工程学院,山西太原030024

出　　处：《农业装备与车辆工程》2025年第1期103-109,共7页Agricultural Equipment & Vehicle Engineering

基　　金：山西省重点研发计划“智能网联重卡编队车路协同关键技术研究与示范”(202102070301019);山西省基础研究计划“多尺度空谱三维特征约束的弱成像交通标识感知新方法”(202103021223464)。

摘　　要：在动态环境中,模型预测控制结合人工势场法对于多约束的非线性方程求解能力并不高,缺乏实时性,无法及时调整路径来规避移动障碍物,并且规划的路径并不平滑,从而增加碰撞风险。为了解决这一问题,针对智能车辆提出了一种新方法,在预测控制的目标函数与约束条件下,引入修正的势场函数,并与牛顿-拉夫逊优化方法(NR-MPC)相结合,实现最优轨迹的搜索。仿真结果表明,在超车过程中,算法NR-MPC的横摆速度比MPC减小80%,比AFP-MPC减小37%,前轮转角比MPC小1.63 rad,比AFP-MPC小0.21 rad,并且NR-MPC算法的转角和航向角变化在6 s后很快变得平缓,10 s左右变得稳定,明显优于MPC和AFP-MPC。该轨迹规划器比APF-MPC控制器展现出更理想的稳定性,牛顿-拉夫逊优化算法求解规划的路径比单一人工势场法更快,有更高的实时性,规划的曲线更加平滑安全,跟踪精度也更加精确。In dynamic environments,model predictive control combined with the artificial potential field method does not have a high ability to solve nonlinear equations with multiple constraints,lacks real-time performance,fails to adjust the path in time to avoid moving obstacles,and the planned paths are not smooth,which increases the risk of collision.In order to solve this problem,a new method was proposed for intelligent vehicles,which introduced a modified potential field function under the objective function and constraints of predictive control and combined it with the Newton-Raphson optimization method to achieve the search for the optimal trajectory.The simulation results showed that during overtaking,the lateral speed of the NR-MPC algorithm was reduced by 80%compared to MPC and 37%compared to AFP-MPC.The front wheel steering angle was 1.63 rad smaller than MPC and 0.21 rad smaller than AFP-MPC.Moreover,the angle and heading changes of the NR-MPC algorithm quickly became smooth after 6 s and stable after about 10 s,which was significantly better than MPC and AFP-MPC.The trajectory planner designed by the method exhibited more desirable stability than the APF-MPC controller,indicated that the Newton-Raphson optimization algorithm solved the planned paths faster than the single artificial potential field method,with higher real-time performance,smoother and safer planned curves,and more accurate tracking accuracy.

关 键 词：牛顿-拉夫逊优化算法 MPC模型预测控制 避障 局部路径规划 

分 类 号：U461.91[机械工程—车辆工程]