改进的基于RBF神经网络的网格细胞到位置细胞模型

Improved grid cell to place cell model based on RBF neural network

为解决类脑导航空间表征中位置细胞间距的控制问题,提出一种改进的基于径向基(RBF)神经网络的网格细胞到位置细胞模型。该模型在实现过程中,引入位置细胞间距因子,结合位置细胞放电率阈值,灵活调整网格细胞到位置细胞转换后所生成的相邻位置细胞间距离,最终改变表征空间环境的位置细胞数量及基于位置细胞放电活动的位置估计性能。经仿真验证,本文模型有效,生成的位置细胞具有类似生物位置细胞的放电特性,相邻位置细胞间距离可控,且基于该方式生成的位置细胞进行位置估计时可得到较好的定位结果。

In this paper,an improved grid cell to place cell model based on radial basis(RBF)neural network is proposed in order to solve the control problem of place cell spacing in brain-like navigational spatial representation.In the process of implementation of the model,the place cell spacing factor is introduced and the place cell discharge rate threshold is combined to flexibly adjust the distance between adjacent place cells generated by the conversion of grid cells to place cells,and finally change the number of place cells that represent the space environment and the location estimation performance based on place cell discharge activities.Simulation results show that the model is effective,the generated place cells have similar discharge characteristics to biological place cells,and the distance between adjacent place cells is controllable,and better location results can be obtained when the place cells generated in this way are estimated.