考虑轨迹预测的大曲率道路智能车辆动态避障控制  

Dynamic obstacle avoidance control of intelligent vehicle on large curvature roads considering trajectory prediction

作　　者：史培龙[1] 王彩瑞 马强 刘瑞[1] 赵轩[1] SHI Pei-long;WANG Cai-rui;MA Qiang;LIU Rui;ZHAO Xuan(School of Automobile,Chang’an University,Xi’an 710064,Shaanxi,China;BYD Auto Shanghai,Shanghai 201611,China)

机构地区：[1]长安大学汽车学院,陕西西安710064 [2]上海比亚迪有限公司,上海201611

出　　处：《长安大学学报（自然科学版）》2024年第4期161-174,共14页Journal of Chang’an University(Natural Science Edition)

基　　金：国家自然科学基金项目(52172361);榆林市科技计划项目(CXY-2020-021);中央高校基本科研业务费专项资金项目(300102222201)。

摘　　要：为了提高智能车辆大曲率道路动态避障安全性,提出基于实时轨迹更新长短时记忆(U-LSTM)神经网络的轨迹预测和基于模糊控制重规划逻辑策略的智能车辆动态避障控制方法。通过提取MATLAB工具箱大曲率道路参数,利用当前时域真实轨迹点信息并训练更新迭代的方法,建立U-LSTM神经网络的轨迹预测模型;考虑到大曲率道路场景中智能车辆跟踪、避让与自车周围信息的非线性关系,提出基于模糊控制的重规划避障逻辑策略,运用动态规划和二次规划算法、S-T(累积距离预测时域)图法对局部路径的轨迹跟踪和速度控制进行优化;通过建立跟踪误差模型和速度跟踪模型实现车辆横向和纵向控制,设计用于车辆横向和纵向控制的MPC路径跟踪和速度跟踪控制器,搭建联合仿真模型并验证轨迹预测和控制方法的有效性。研究结果表明:提出的智能车辆动态避障控制方法在大曲率道路上能准确预测车辆的轨迹,U-LSTM神经网络能有效提高预测准确性;重规划避障逻辑策略能够实现动态障碍物有效避让,且满足车辆纵向和横向跟踪控制精度,能够保证车辆的行驶稳定性。In order to improve the dynamic obstacle avoidance safety of intelligent vehicles on high curvature roads,a trajectory prediction method based on real-time trajectory update long short term memory(U-LSTM)neural network and a dynamic obstacle avoidance control method for intelligent vehicles based on fuzzy control replanning logic strategy was proposed.A trajectory prediction model for U-LSTM neural network was established by extracting large curvature road parameters from the MATLAB toolbox,utilizing the current time-domain real trajectory point information,and training an updated iterative method.Considering the nonlinear relationship between intelligent vehicle tracking,avoidance,and surrounding information in high curvature road scenes,a replanning obstacle avoidance logic strategy based on fuzzy control is proposed.Dynamic programming and quadratic programming algorithms,as well as S-T graph method,was used to optimize the trajectory tracking and speed control of local paths.By establishing tracking error models and speed tracking models,vehicle lateral and longitudinal control was achieved.An MPC path tracking and speed tracking controller for vehicle lateral and longitudinal control was designed,and a joint simulation model was constructed to verify the rationality and effectiveness of trajectory prediction and control methods.The results show that the intelligent vehicle dynamic obstacle avoidance control method proposed in the article can accurately predict the vehicle’s trajectory on high curvature roads,and the real-time trajectory U-LSTM neural network can effectively improve prediction accuracy.The replanning obstacle avoidance logic strategy can effectively avoid dynamic obstacles and meet the accuracy of vehicle longitudinal and lateral tracking control,ensuring the driving stability of the vehicle.7tabs,20figs,25refs.

关 键 词：汽车工程 轨迹预测 模型预测控制 跟踪控制 避障轨迹规划 大曲率道路工况 

分 类 号：U461.5[机械工程—车辆工程]