摘要
针对精密操控中人-机之间缺乏信息双向交互,以及人的精神状态变化会严重影响肢体操控的精度和安全性等问题,提出了一种引入深度强化学习思想的脑-机协作精密操控方法。首先,结合人在上层规划与机器在精细控制上的各自优势,建立由主动操控和被动调控组成的双环路人-机信息交互机制;其次,引入深度强化学习思想,从蒙特卡罗采样原理出发,以表征精神状态的脑电信号(EEG)作为模型的输入,以机器人速度指令作为模型的输出,推导出脑-机协作方法的数学模型;再次,建立具有3个全连接层的精神状态感知网络,从脑-机接口系统的实时监测计算机内存中提取最后1000 ms的EEG作为输入信号,设计开发脑-机协作精密操控算法。最后,创建轨迹跟踪虚拟环境和任务场景,对脑-机协作精密操控方法进行了实验验证,结果表明:该方法在轨迹跟踪任务的控制精度和完成时间指标上均得到了提高;相较于传统方法,该方法的平均轨迹跟踪精度和完成时间指标分别提高了36.55%和22.81%。
A precise control method for brain-computer cooperation with deep reinforcement learning is proposed to solve the problem that the lack of bidirectional information interaction between human and computer and the change of mental state in precise control seriously affect the precision and safety of limb control.First of all,combining the advantages of human in global planning and machine in fine control,a‘double-loop’information interaction mechanism composed of active control loop and passive control loop is established.Secondly,the idea of deep reinforcement learning is introduced,and a mathematical model of brain-computer cooperation is derived based on the Monte Carlo sampling principle,with the electroencephalogram(EEG)representing mental state feature as the input of the model and the robot speed instruction as the output.Thirdly,a mental state perception network with three fully connected layers is established,and the EEG of the last 1000 ms in the real-time monitoring computer memory of brain-machine interface system is extracted as input signal,then a precise brain-computer cooperation algorithm is designed and developed.Finally,a virtual environment and task scene for trajectory tracking is created,and the precise brain-computer cooperation method is experimentally verified.Results and a comparison with a traditional method show that the proposed method improves both the accuracy and completion time of trajectory tracking control task by 36.55%and 22.81%,respectively.
作者
张腾
张小栋
张英杰
陆竹风
朱文静
蒋永玉
ZHANG Teng;ZHANG Xiaodong;ZHANG Yingjie;LU Zhufeng;ZHU Wenjing;JIANG Yongyu(School of Mechanical Engineering,Xi’an Jiaotong University,Xi’an 710049,China;Shaanxi Key Laboratory of Intelligent Robots,Xi’an Jiaotong University,Xi’an 710049,China)
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2021年第2期1-9,共9页
Journal of Xi'an Jiaotong University
基金
国家重点研发计划资助项目(2017YFB1300303)。
关键词
脑-机协作
深度强化学习
脑-机接口
轨迹跟踪
精密操控
brain-computer cooperation
deep reinforcement learning
brain-machine interface
trajectory tracking
precise control
