摘要
针对无人驾驶车辆采用纯跟踪算法对不同曲率路径跟踪时,出现道路适应能力弱和跟踪精度差的问题,提出一种基于代价的滚动预瞄模型(rolling preview model,RPM),以提高纯跟踪算法跟踪精度与鲁棒性。首先,根据车辆运动学与阿克曼转向几何,提出预瞄轨迹的确定方法以及预瞄轨迹与待跟踪路径间的几何约束;其次,设计道路弯曲度加权项并构建目标函数对预瞄轨迹进行优化,以获得预瞄距离的最优值;最后,在ROS/Gazebo仿真环境下设置不同初始状态与不同曲率的工况进行对比仿真实验,并在空旷环境中对8字形路径进行实车跟踪实验。实验结果表明,所提出的滚动预瞄模型能够根据预瞄轨迹与待跟踪路径的几何关系有效调节预瞄距离,相较于麻省理工(Massachusetts Institute of Technology,MIT)算法和Stanley算法,滚动预瞄模型在特殊初始状态、大曲率道路下有利于跟踪精度的提高。
A cost-based rolling preview model(RPM)was proposed to improve the tracking accuracy and robustness of the pure pursuit technique for autonomous vehicle with poor road adaptability and tracking accuracy when tracking different curvature paths using pure pursuit algorithms for autonomous vehicles.Firstly,the determination method of the preview trajectory and the geometric constraints between the preview trajectory and the reference path were proposed based on the vehicle kinematics and Ackermann steering geometry.Secondly,the weighted term of road curvature was designed and the objective function was constructed to optimize the preview trajectory to obtain the optimal preview distance.Finally,the comparison simulation experiment were conducted in the ROS/Gazebo simulation environment for different initial states and different curvature working conditions,and the real vehicle tracking experiment was carried out on the 8-shaped path in the open area.The experimental results show that the proposed rolling preview model can effectively adjust the preview distance according to the relationship between the preview trajectory and the path to be tracked.Compared with the Massachusetts Institute of Technology(MIT)algorithm and Stanley algorithm,the rolling preview model proposed has good precision under the special initial state and large curvature path.
作者
胡丹丹
尹鹏飞
牛国臣
赵金聚
HU Dan-dan;YIN Peng-fei;NIU Guo-chen;ZHAO Jin-ju(Robotices Institute,Civil Aviation University of China,Tianjin 300300,China)
出处
《科学技术与工程》
北大核心
2023年第9期3787-3793,共7页
Science Technology and Engineering
基金
天津市科技计划(17ZXHLGX00120)
中央高校基本科研业务费专项资金(3122019043)
天津市研究生科研创新项目(2021YJSO2S30)
