ROBOT ARM CONTROL USING CORTICAL MULTINEURONAL RECORDING

该提案提出了利用从大脑中同时记录的神经元群中提取的“运动”信息来补救与瘫痪、截肢和其他神经系统疾病相关的运动功能丧失的可能性。这项工作在科学上也很重要，因为它直接解决了大脑中神经群编码的问题，以及通过生物反馈控制这种编码的可能性。我们最近在老鼠和猴子身上证明了在运动皮层同时记录神经元数量来控制机械臂运动的可行性。尤其是老鼠，它们能够利用大脑活动来精确地定位（一维）机械臂在水滴下的位置，然后把水滴带回嘴里。此外，在这种“神经-机器人”模式下的持续训练中，这些动物能够逐渐将这种神经活动与通常与它们相关的明显运动分离开来。该方案有三个具体目标：1 .利用猴子的慢性神经集合记录直接控制机器人手臂的多向运动。主要问题是神经机器人是否能够在不同的负载条件下控制多个方向的运动。2。开发、实施和优化将神经元群体活动转化为实时神经机器人控制信号的新方法。主要问题是简单的线性神经种群编码算法是否可以用来产生最优的神经机器人控制函数，或者是否需要非线性网络。3。探讨感觉运动去神经后神经机器人控制的可行性。主要的问题是瘫痪后神经机器人控制是否可行。这将在可逆的前肢去神经或截肢的大鼠中进行研究。

This proposal addresses the possibility of utilizing "motor" information extracted from simultaneous neuronal population recordings in the brain to remedy the loss of motor function associated with paralysis, limb amputation and other neurological conditions. This effort is also scientifically significant because it directly addresses the problem of neural population coding in the brain, and the possibility of controlling such coding through biofeedback. We have recently demonstrated in rats and monkeys the feasibility of using simultaneous neuronal population recordings in the motor cortex to control movement of a robot arm. The rats, in particular, were able to utilize their brain activity to accurately position (in one dimension) the robot arm under a water dropper, and then carry the water drop back to their mouths. Moreover, over continued training in this "neuro-robotic" mode, these animals were able progressively decorrelate this neural activity from the overt movements with which they were normally associated. This proposal has three specific aims: I. To utilize chronic neural ensemble recordings in monkeys to directly control multi-directional robot arm movement. The main issue is whether neuro-robotic feasibility be demonstrated for control of movements in multiple directions and under varying load conditions. II. To develop, implement and optimize new methods for transforming neuronal population activity into realtime neuro-robotic control signals. The main issue is whether simple linear neural population coding algorithms can be used to produce optimal neuro-robotic control functions, or whether nonlinear networks will be necessary. III. To investigate the feasibility of neuro-robotic control after sensorimotor denervation. The main question is whether neuro- robotic control is feasible after paralysis. This will be investigated here in rats subject to reversible denervation or amputation of the forelimb.