多尺度网格细胞路径综合方法

为实现无人作战飞机(UCAV,Unmanned Combat Aerial Vehicle)认知导航的空间方位自主推算,提出了一种基于多尺度网格细胞的路径整合方法.该方法模拟背侧内嗅皮层(dMEC,dorsal Medial Entorhinal Cortex)的相同区域网格细胞放电特征相同、不同区域放电特征递增变化的特点,构建尺度递增的仿生多尺度网格图组,在各层中引入突触样式(synaptic pattern)计算各细胞权值,通过细胞的活跃度变化表征各网格层中位置的变化,并在各层分别实现路径整合,进而利用低尺度整合结果调整高尺度整合,提高空间位置的推算精度.实验结果表明,所提方法在一定的速度误差与方向误差范围内能够精确推算方位,具有较高的空间位置推算精度,并且方向误差值随运动方向变化呈锯齿状分布.

Method to realize path integration based on multi-scale grid cells

In order to autonomously reckon location and azimuth for unmanned combat aerial vehicle (UCAV)’s cognitive navigation, a kind of path integration based on multi-scale grid cells was proposed. According to the characteristics of neighboring cells in dorsal medial entorhinal cortex (dMEC) sharing common firing traits, while the firing traits of the grid increases isometrically along the dorsoventral axis, a bionic grid groups with incremental scales was constructed firstly, and then synaptic pattern was introduced in each grid layer to evaluate cell weights, therefore, the location and azimuth in each grid layer were calculated by changeable cell activities, and path integration was ultimately achieved in each grid layer. Space location precision was further improved by using integration results of small scales to adjust the grid layer of bigger scale. Simulation results prove that the method can exactly reckon location and azimuth within certain velocity error and azimuth error. It has a higher space location reckoning by adjusting bigger-scale grid layer. And the azimuth error has an indentation distribution accompanying the moving direction changing.