基于视觉记忆想象模式的仿人机器人意念控制研究

项目名称： 基于视觉记忆想象模式的仿人机器人意念控制研究

项目编号： No.61473207

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 自动化技术、计算机技术

项目作者： 李伟

作者单位： 华东理工大学

项目金额： 82万元

中文摘要： 仿人机器人的意念控制是一个新兴的跨学科研究课题，其成果不仅有望缓解人口老龄化带来的社会压力并促进残疾人康复事业的发展，而且具有重要的军事价值。针对现有脑电信号诱发模式用于控制仿人机器人的局限性，本研究致力于建立一种新的模式，即对预先记忆的机器人动作画面进行想象的视觉记忆想象模式。我们的前期研究表明该模式的信号具有不同于现有模式的特征，在仿人机器人的意念控制中更具优势。 本研究拟围绕特定视觉记忆想象中的脑电信号解码和基于脑电信号的人-机决策融合两个关键问题，开展基于该模式的新型脑-机交互实验范式、多域特征表征、多域特征分类模型和复杂环境下的机器人行为融合四个方面的研究，揭示该模式下的脑电信号特征并实现其对仿人机器人的意念控制。 特色和创新：提出了视觉记忆想象这一全新的脑-机接口模式；采用以通道间相位角为主的多域特征来表征脑电信号的状态；借助人-机决策融合增强这一新模式对仿人机器人的控制能力。

中文关键词： 视觉记忆想象；意念控制；仿人机器人；人-机决策融合

英文摘要： The mind control of humanoid robot is a newly emerging interdisciplinary study. It is not only a helpful way to reduce the pressure caused by the aging population and to develop the rehabilitation service for the disabled, but also of important value in military fields. To overcome the limitations of the existing evoked modes of electroencephalography (EEG) signals used in controlling humanoid robots, this study works at establishing a new EEG evoked mode named visual memory imagery (VMI) mode. The so-called VMI means imagining the images of robot behaviors memorized beforehand. It has been proven by our preliminary study that the EEG features evoked by the VMI mode are different from those evoked by the existing modes and the VMI mode has more advantages in controlling a humanoid robot. This study is intended to be developed around the decoding of the EEG signals under specific VMI states and the brainwave-based human-robot decision fusion. The four aspects of the research content are the new BCI (brain-computer interface) paradigm for the VMI mode, the multi-domain features' representation, the classification model for the multi-domain EEG features, and the brainwave-based fusion of humanoid robot's complex behaviors.The objective of this study is to explore the EEG features and their distribution in the VMI mode, and to realize the mind control of humanoid robots under complex environments based on the VMI mode. Innovations and characteristics: We put forward a new mode for BCI research named visual memory imagery, utilize multi-domain features focusing on the phase angles between different channels to represent the states of EEG signals, and improve the control capability of this new BCI mode on humanoid robots under complex environments using decision fusion between the human and the robot.

英文关键词： visual memory imagery;mind control;humanoid robot;human-robot decision fusion