手臂抓取时空协调神经网络模型构建方法研究

Research on Construction Method of Neural Model with Spatiotemporal Coordination during Arm Grasping

基于VITE(vector integration to endpoint)点对点运动轨迹生成模型,提出一种具有生物学意义的手臂抓取神经网络模型,用以解释延伸和抓取过程中手臂运动、抓取角度以及手掌朝向三个组件之间的时空协调问题。模型主要利用基底神经节丘脑皮层环路门控调节组件通道的全局运动速度,设置可进行空间状态信息交流的耦合神经元动态调节抓握孔径,从而实现手臂延伸与抓取运动的时空协调,为更好地进行3D手势跟踪奠定基础。模型改变了门控信号的输入方式,优化了最大抓取孔径的更新方法,还增设了用于检测组件目标状态扰动情况的监督细胞。相同条件下的仿真结果表明,相较于原孔径组件的二目标值给定法,新模型的运动调节时间缩短了13%,较好地体现了抓取运动的动力学关键特征,增强了运动协调性和抗干扰性。

Based on the VITE(vector integration to endpoint)point-to-point trajectory formation model,we propose a new arm grasping neural network model with biological meaning to explain the spatiotemporal coordination among arm transport component,hand preshape component and palm orientation component during reach-to-grasp tasks.In addition to adjusting the overall speed of movement components,the basal ganglia thalamus cortex loop is exploited and coupled neurons are set up to ensure the exchange of spatial status information with one another.In this way,the spatiotemporal coordination of reach-to-grasp movements is achieved,laying a solid foundation for robust 3-D gesture tracking.Model changes the gating input signal,optimizes the updating method of maximum grip aperture,and also adds monitoring cells to detect the state disturbance of component’s target state.Simulation results under same conditions demonstrate that comparing with the two-value given method of original aperture component,the proposed model robustly achieves spatiotemporal coordination during prehension and effectively reproduces the important kinetic characteristics of arm prehension movement,meanwhile the setting time is reduced by 13 percent.