基于LSTM和SMC的农用履带机器人轨迹跟踪控制  

Trajectory Control of Crawler Robot Based on LSTM and SMC

作　　者：刘东阳 查文文 陶亮 朱诚[1] 辜丽川[1] 焦俊[1] Liu Dongyang;Zha Wenwen;Tao Liang;Zhu Cheng;Gu Lichuan;Jiao Jun(School of Information and Computer,Anhui Agricultural University,Hefei 230036,China)

机构地区：[1]安徽农业大学信息与计算机学院,安徽合肥230036

出　　处：《系统仿真学报》2023年第4期747-759,共13页Journal of System Simulation

基　　金：国家自然科学基金(31671589);安徽省科技重大专项(201903a06020009,202103b06020013);省教育厅自然基金(KJ2019A0209)。

摘　　要：轨迹跟踪是移动机器人控制技术的重要领域之一,拥有广阔的发展前景,而高度非线性的动态特性是控制器设计的主要障碍。提出了一种基于长短期记忆网络(LSTM)和准滑动模态的滑模控制(SMC)方法。给出履带车的运动学模型和动力学模型,并基于动力学模型建立滑模控制系统。设计基于深度学习方法的LSTM网络来对未知干扰项进行控制补偿,降低外部干扰的影响,通过结合LSTM网络和准滑动模态的优势,减弱震颤现象。并给出控制系统的稳定性分析。在不同轨迹上进行长短期记忆网络MATLAB/Simulink仿真实验,并与现有方法进行了比较,证明了其优越性。Trajectory tracking is an important part of mobile robot control technology and possesses prospect.Highly nonlinear dynamic characteristics are the main obstacles of controller design.A SMC method based on LSTM and quasi-sliding mode is proposed.The kinematics model and dynamics model of the tracked vehicle are given,and the sliding mode control system is established based on the dynamics model.LSTM network based on deep learning method is designed to control and compensate the unknown interference items,reduce the influence of external interference,and reduce the tremor phenomenon by combining the advantages of LSTM network and quasi-sliding mode.The stability analysis of the control system is given.MATLAB/Simulink simulation experiments are carried out on different trajectories.Compared with the existing methods,the superiority is proven.

关 键 词：长短期记忆网络 滑模控制 运动学模型 动力学模型 轨迹跟踪 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]