基于CNN的康复机器人逆运动学分析与轨迹规划
CNN-Based Inverse Kinematics Analysis and Trajectory Planning for Rehabilitation Robotics
徐玉杰 1李宪华 1林凤涛2
作者信息
- 1. 安徽理工大学 机电工程学院,安徽 淮南 232001
- 2. 华东交通大学 载运工具与装备教育部重点实验室,江西 南昌 330013
- 折叠
摘要
为提高脑卒中患者上肢康复训练的运动舒适性,设计一种具有运动依从性的上肢康复机器人.用D-H参数法建立机器人模型,进行正运动学分析,针对传统逆运动学求解方法的高复杂性,搭建一种卷积神经网络进行逆运动学求解,96.24%的预测角度所计算得到的关节位置,与标签位置误差在0～0.025m范围内.采用解析法分析机器人的工作空间,使用七次多项式插值法规划了一组康复训练动作,仿真结果符合患者的康复运动要求,能够提供高质量康复训练.
Abstract
To enhance the comfort of upper limb rehabilitation training for stroke patients,a motion-com-pliant upper limb rehabilitation robot is designed.The robot model was established using the Denavit-Harten-berg(D-H)parameter method,and positive kinematic analysis was performed.Given the high complexity of traditional inverse kinematics solving methods,a convolutional neural network was implemented to solve the in-verse kinematics.The neural network achieved the accuracy of 96.24%in predicting joint angles,with errors in joint positions within the range of 0 to 0.025 meters compared to the labeled positions.Analyzing the robot's workspace using analytical methods,a set of rehabilitation training movements was planned using a seventh-de-gree polynomial interpolation method.Simulation results meet the rehabilitation exercise requirements of pa-tients,enabling the provision of high-quality rehabilitation training.
关键词
卷积神经网络/上肢康复机器人/逆运动学/轨迹规划Key words
convolutional neural network/upper limb rehabilitation robot/inverse kinematics/trajectory planning引用本文复制引用
出版年
2024
NETL
NSTL
万方数据