野外无人车路径规划与轨迹规划技术

Path Planning and Trajectory Planning Technology of Unmanned Ground Vehicle(UGV)in the Field

常规布局及结构化的城市环境中的无人地面车,其环境感知和路径规划技术的研究相对成熟。然而,无人车在信息未知、环境复杂的野外进行自主规划和障碍规避尚存在困难。针对复杂环境中无人车的路径规划和轨迹规划技术展开研究,首先,考虑环境相对未知、存在凹凸障碍物和起伏地面等因素,利用激光雷达、相机和IMU组成车载多传感器系统获取复杂的野外环境信息并进行数据处理和校正;其次,通过训练径向基神经网络(RBF网络)对野外环境进行建模;然后,基于环境模型引入起点至目标位置的距离、环境高度和梯度以构建约束,通过优化约束函数实现复杂环境中无人车的路径规划;最后,基于Minimum Snap的思想,利用N阶多项式拟合得到的路径,将规划后生成的折线路径优化为最终需要跟踪的轨迹。通过仿真实验验证了方法的有效性:所提出的方法实现了无人地面车在野外环境中自主路径规划和轨迹规划。

For the Unmanned Ground Vehicle(UGV)in conventional layout and structured urban environments,the environment perception and path planning technologies are relatively mature.However,the UGV has difficulties in autonomous planning and obstacle avoidance in the field with less information.This article focuses on the path planning and trajectory planning technology of UGV in the complex field environment.Firstly,considering factors like the presence of convex obstacles,concave obstacles and the undulating ground,the on-board sensors system composed of lidar,camera and IMU are used to obtain complex field environment information and data processing.Then,the field environment can be modeled by training Radial Basis Function neural networks(RBF networks).After that,the constraint is constructed by the obtained environment model and the constraint function is optimized for path planning of UGV in the field environment by introducing the distance between start position and the target position,the environment height and gradient.Finally,the path is fit by the N-order polynomial based on the Minimum Snap so that the obtained path can be optimized to the trajectory that needs to be tracked.The effectiveness of this method is verified by simulation experiments that it can achieve the purpose of autonomous path planning and trajectory planning of the UGV in the complex field environment.