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,...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。展开更多
In the process of field operation management,determining how to accurately realize crop row identification and path tracking control is an essential part of tractor automatic navigation.According to the linear operati...In the process of field operation management,determining how to accurately realize crop row identification and path tracking control is an essential part of tractor automatic navigation.According to the linear operation in the process of cotton field management,the tractor path tracking control system was designed based on binocular vision and the pure pursuit model.A new crop row detection method based on the Census transform and the PID control algorithm with dead zone were used.First,the upper computer software was developed by C++with the functions of parameter setting and image acquisition and processing.Second,an automatic steering controller was developed based on microprocessor MC9S12XS128 of Freescale.The control program was developed based on modular design using CodeWarrior during development of the PID-based automatic steering control strategy.Finally,a field experiment platform of tractor path tracking control was built,and field experiments under the actual cotton were conducted.The optimal visibility distance was determined by several previous experiments.When the tractor tracks the path with the optimal visibility distance in the growth environment of actual cotton crops,the mean absolute deviation of course angle was 0.95°,and the standard deviation was 1.26°;the mean absolute deviation of lateral position was 4.00 cm,and the standard deviation was 4.97 cm;the mean absolute deviation of front wheel angle was 2.99°,and the standard deviation was 3.67°.The experimental results show that(1)the crop row detection method based on Census transform can identify the crop line and plan the navigation path well,and(2)the tractor path tracking control system based on binocular vision has good stability and high control precision;thus,the control systemcan realize the automatic path tracking control of cotton line operation and meets the agricultural requirements of cotton field operation management.展开更多
文摘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。
基金supported by the National Key Research and Development Program(No.2017YFD0700400-2017YFD0700403).
文摘In the process of field operation management,determining how to accurately realize crop row identification and path tracking control is an essential part of tractor automatic navigation.According to the linear operation in the process of cotton field management,the tractor path tracking control system was designed based on binocular vision and the pure pursuit model.A new crop row detection method based on the Census transform and the PID control algorithm with dead zone were used.First,the upper computer software was developed by C++with the functions of parameter setting and image acquisition and processing.Second,an automatic steering controller was developed based on microprocessor MC9S12XS128 of Freescale.The control program was developed based on modular design using CodeWarrior during development of the PID-based automatic steering control strategy.Finally,a field experiment platform of tractor path tracking control was built,and field experiments under the actual cotton were conducted.The optimal visibility distance was determined by several previous experiments.When the tractor tracks the path with the optimal visibility distance in the growth environment of actual cotton crops,the mean absolute deviation of course angle was 0.95°,and the standard deviation was 1.26°;the mean absolute deviation of lateral position was 4.00 cm,and the standard deviation was 4.97 cm;the mean absolute deviation of front wheel angle was 2.99°,and the standard deviation was 3.67°.The experimental results show that(1)the crop row detection method based on Census transform can identify the crop line and plan the navigation path well,and(2)the tractor path tracking control system based on binocular vision has good stability and high control precision;thus,the control systemcan realize the automatic path tracking control of cotton line operation and meets the agricultural requirements of cotton field operation management.
