摘要
针对拖挂式移动机器人的反向泊车运动控制问题,提出了一种基于策略梯度及强化学习的拖挂式移动机器人控制方法。首先,在Gym软件中搭建了具有单节拖车的拖挂式移动机器人的运动学仿真模型,并设计了稳定的反向泊车运动控制律。其次,构建了基于Tensorflow框架的神经网络模型,设计了相应的损失函数,并利用策略梯度算法更新神经网络的参数,以训练机器人的反向泊车运动。仿真实验结果表明,经过训练的拖挂式移动机器人能够有效地学习反向泊车运动控制策略,并稳定地实现反向泊车运动。不同参数下的实验结果验证了基于策略梯度算法的强化学习模型的有效性。
A control method of tractor-trailer mobile robot(TTMR)is proposed based on policy gradient and reinforcement learning approach for backward parking motion.Firstly,kinematic model of TTMR with one trailer is constructed by Gym software.A stable reverse motion control law is designed.Secondly,a neural network model is constructed based on the Tensorflow framework,and the corresponding loss function is designed.The policy gradient algorithm is used to update the parameters of the neural network and train the robot's reverse parking motion.Simulation experimental results show that the trained TTMR can effectively learn the backward motion control strategy and do stably.The experimental results under different parameters validate the effectiveness of the reinforcement learning model based on the policy gradient algorithm designed in this paper.
作者
林俊文
程金
季金胜
Lin Junwen;Cheng Jin;Ji Jinsheng(School of Electrical Engineering,University of Jinan,Jinan 250024,China;School of Electrical and Electronic Engineering,Nanyang Technological University,Singapore 639798)
出处
《市政技术》
2023年第10期101-105,共5页
Journal of Municipal Technology
基金
国家自然科学基金(61203335)。
关键词
拖挂式移动机器人
强化学习
人工智能
策略梯度算法
反向泊车
tractor-trailer mobile robot(TTMR)
reinforcement learning
artificial intelligence
policy gradient algorithm
backward parking