基于动态视觉范式与实时反馈校验的脑控机械臂

Brain Controlled Manipulator Based on Dynamic Visual Paradigm and Real-time Feedback

提出了一种应用于脑控机械臂的动态脑控视觉范式以及实时反馈的脑控指令校验方法,并基于该范式及方法实现了一套脑控机械臂系统。相比于当前广泛采用的固定视觉刺激,使用者通过注视不同位置控制机械臂末端运动过程的控制实现方法,提出的视觉范式基于动态刺激标签输出面向控制对象的脑控指令。该指令结合机械臂的自主空间运动,实现了基于脑-机协同下的快速抓取。同时,针对面向运动过程脑机接口中脑特征识别准确率低,系统鲁棒性不足的问题,提出了一种实时反馈校验方法,基于该方法对系统使用者的意图进行校验,可有效提升脑控机械臂对人类意图识别的准确率。实验结果表明,研究提出的动态刺激及实时反馈校验方法,其脑控抓取的平均成功率为85%(基于运动过程的控制方法的成功率为47.9%),单次抓取任务下的平均时间消耗为22.46 s(基于运动过程的控制方法为64.76 s)。在未来,提出的脑控机械臂框架可以以节点模块的形式整合入机器人操作系统,并应用于人-机体系协同下的脑机接口闭环控制。

A dynamic brain control vision paradigm and a real-time feedback brain command verification method applied to the brain control robot arm are proposed,and a set of brain control robotic arm system is realized based on the proposed paradigm and method.Compared with the fixed stimulus position and sequential control paradigm of brain computer interfaces,the paradigm outputs object-oriented brain control commands based on the dynamic stimulus,and fast capture based on brain-machine cooperation is realized.On the other hand,in view of the low accuracy of brain feature recognition and insufficient robustness of the system in the process-oriented brain computer interface,this research proposes a real-time feedback verification method.Based on this method,the raw brain command can be verified via dynamic frequency adjustment,which can effectively improve the accuracy of the brain controlled robot arm in brain command recognition.Result shows the dynamic stimulation and real-time feedback verification method has an average success rate of 85%(47.9%for the control method based on motion process)for brain controlled grasping,and an average time consumption of 22.46 s(64.76 s for the control method based on motion process)for a single grasping task.In the future,the framework of the brain controlled robot arm proposed can be integrated into the robot operating system in the form of node modules,and applied to the closed-loop control of the brain-machine cooperative systems.