深度学习理论下移动机器人全局路径规划方法

Global Path Planning Method for Mobile Robot under Deep Learning Theory

移动机器人在复杂环境运动过程中能安全、无碰撞地绕过所有障碍物是对机器人较高的智能化要求。由于障碍物的位置和几何性质等信息具有随机性,机器人的速度和角速度等控制信息易出现随机偏差。为此,提出基于深度学习的移动机器人全局路径控制方法。建立移动机器人动力学方程,依据移动机器人的线速度与角速度,将移动机器人运动行为分为避障行为、转向目标行为、奔向目标行为。结合深度学习方法中的神经网络方法,构建机器人全局路径规划模型。利用模糊控制算法设计模糊控制器,纠正路径规划偏差,实现机器人的全局路径优化控制。实验结果表明,研究方法控制下移动机器人的路径角度偏差距离小于±5cm,能够在10ms之内完成偏差收敛,且规划的路径更短。

In a complex environment,the movement of mobile robots needs to pass around all obstacles safely without collision.This is also a higher intelligent requirement for robots.Due to the randomness of the information such as position and geometric properties of obstacles,the control information such as speed and angular velocity of robots is prone to random deviation.Therefore,this paper puts forward a method of controlling the global path of mobile robots based on deep learning.Firstly,kinetic equations of mobile robots were constructed.According to the linear velocity and angular velocity of the mobile robot,the motion behavior was divided into the behavior of obstacle avoidance as well as the behavior of turning to target and running to target.Combined with the neural network method in deep learning theory,a global path-planning model of a robot was built.Moreover,a fuzzy control algorithm was adopted to design a fuzzy controller correcting the deviation of path planning.Finally,we realized the global path optimization control.Experimental results show that the angle deviation of the path of the mobile robot controlled by the proposed method is between±5cm,and the deviation convergence can be completed within 10ms.Meanwhile,the path planned by the method is shorter.