摘要
为实现人体运动意图预测,该文提出一种基于粒子群优化(particle swarm optimization,PSO)算法优化回声状态网络(echo state network,ESN)的模型预测控制(model predictive control,MPC)方法。首先,利用运动捕捉系统获得人体动作运动学信息,通过OpenSim软件反解算获取动力学信息;其次,以动力学信息为输入,运动学信息为输出,构建人体骨骼肌肉系统的ESN模型,并利用PSO算法优化ESN模型的关键参数;同时,将线性化后的ESN模型作为MPC控制对象,通过运动学信息,反优化MPC目标函数,求解目标结果,完成对人体运动意图的预测;最后,通过比较实验验证了所提方法的有效性。该方法对人体运动意图预测及穿戴式机器人控制算法设计等相关应用研究与教学实践具有实际意义。
Aiming at the prediction of human motion intention,a model predictive control(MPC)method based on particle swarm optimization(PSO)algorithm with optimized echo state network(ESN)is proposed in this paper.Firstly,the human motion kinematics and dynamics information are obtained by the motion capture system and OpenSim software calculated,respectively.Then,the ESN model of human skeletal muscle system is constructed with dynamic information as input and kinematics information as output,and the PSO algorithm is used to optimize the key parameters of ESN model.At the same time,the linearized ESN model is regarded as the control object that used to obtain the object result through using the kinematics information and inverse-optimizing the MPC control function in order to finish the prediction of human motion.Finally,the proposed method is verified by the comparative experiments.This method is of actual significance for application research and practical teaching with relation to human intention prediction and wearable robotics control algorithm design.
作者
罗晶
李思
唐必伟
向馗
庞牧野
LUO Jing;LI Si;TANG Biwei;XIANG Kui;PANG Muye(School of Automation,Wuhan University of Technology,Wuhan 430070,China)
出处
《实验技术与管理》
CAS
北大核心
2023年第4期40-46,共7页
Experimental Technology and Management
基金
国家自然科学基金项目(62203341)
中央高校基本科研业务费专项资金资助(2022IVA044)。
关键词
人体运动意图预测
粒子群算法
回声状态网络
模型预测控制
实验设计
human motion intention prediction
particle swarm optimization
echo state networks
model predictive control
experimental design
