计算机应用研究2025，Vol.42Issue(6)：1668-1675,8.DOI:10.19734/j.issn.1001-3695.2024.11.0474
基于改进行为克隆算法的机器人运动控制策略
Robot motion control strategy based on improved behavior cloning algorithm
摘要
Abstract
This paper addressed the challenges of complex motion control strategy training,inefficient and imprecise path planning execution in robots performing fine operations such as dual-arm cooperative insertion tasks.It proposed an improved robot behavior cloning algorithm based on multi-scale feature pyramids and attention mechanisms.The algorithm combined re-sidual networks and feature pyramids to design the backbone network,extracting and fusing multi-scale image features,which enhanced the robot's environmental perception and visual feedback capabilities.It introduced an action segmentation module to improve the accuracy and smoothness of control strategies,reducing compound errors in behavior cloning.Additionally,the algorithm trained the control strategy as a conditional variational autoencoder(CVAE)using the attention mechanism to learn the distribution of demonstration data and capture the correlation between image features and actions.This approach improved the generalization ability and adaptability of the strategy in unfamiliar environments.Simulation results show that the proposed algorithm outperforms five baseline models in terms of success rate and trajectory smoothness in two fine operation tasks.These results demonstrate that the algorithm can execute precise robot fine operation tasks through simple training.关键词
机器人精细操作/运动控制策略/行为克隆/动作序列Key words
precision operation of robot/motion control strategy/behavioral cloning/action sequence分类
计算机与自动化引用本文复制引用
黄小霞,阳波,向鑫,陈灵,陈中祥,孙舜尧,肖宏峰..基于改进行为克隆算法的机器人运动控制策略[J].计算机应用研究,2025,42(6):1668-1675,8.基金项目
国家自然科学基金资助项目(12271525) （12271525）
湖南省科技创新计划资助项目(2024JK2022) （2024JK2022）
