基于强化学习的同心管机器人逆运动学研究  

Research on inverse kinematic of concentric tube robot based on reinforcement learning

作　　者：刘扬 冯宇 李永强[1] LIU Yang;FENG Yu;LI Yongqiang(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023)

机构地区：[1]浙江工业大学信息工程学院,杭州310023

出　　处：《高技术通讯》2024年第12期1341-1350,共10页Chinese High Technology Letters

基　　金：国家自然科学基金(61973276)资助项目。

摘　　要：同心管机器人(CTR)是一种连续体机器人,具有体积小、可伸缩、高曲率的特点,适用于微创医疗领域。然而,由于其运动学模型的复杂性,实施有效的逆运动学控制是一项具有挑战性的任务。本文介绍了其正运动学模型,实现了机器人从控制输入到其笛卡尔位置的映射。提出了一种基于强化学习(RL)的逆运动学控制方法,根据实际驱动系统设计离散的强化学习训练动作,并结合实际的驱动器限位,设计考虑运动安全的训练奖励,采用强化学习异步优势演员-评论家学习策略(A3C)计算出更符合实际应用的机器人控制输入,为CTR的逆运动学控制提供了一种新的方法。设计并搭建了CTR控制平台,验证了所提逆运动学控制方法的有效性和正确性。在对轨迹的追踪实验中,机器人能以平均1.462±0.483 mm的追踪误差对轨迹进行追踪。Concentric tube robot(CTR)is a type of continuum robot with the characteristics of small size,scalability,and high curvature,which is suitable for the field of minimally invasive medicine.However,due to the complexity of its kinematic model,implementing effective inverse kinematics control is a challenging task.This paper introduces the forward kinematic model of CTR to realize the mapping from the control input to the Cartesian position.An inverse kinematic control method based on reinforcement learning(RL)is proposed.Discrete reinforcement learning training actions are designed according to the actual drive system,and training rewards are designed in combination with the actual drive limit.Moreover,the asynchronously advantage actor-critic(A3C)learning strategy is used to calculate CTR control input that is more suitable for practical applications.It provides a new method for the inverse kinematics control of CTR.In addition,the CTR control platform is designed and built to verify the effectiveness and correctness of the proposed inverse kinematics control method.In trajectory tracking experiments,the CTR can track the trajectory with an average tracking error of 1.462±0.483 mm.

关 键 词：同心管机器人(CTR) 运动学 强化学习(RL) 

分 类 号：TP2[自动化与计算机技术—检测技术与自动化装置]