基于多尺度网格细胞模型的无人机类脑矢量导航方法

Brain-inspired vector navigation method based on model of multi-scale grid cells

动物具有优秀的矢量导航能力,为发展未知环境下无人机新型自主智能面向目标导航方法提供了较好的生物模型。首先阐述了动物大脑矢量导航机理,然后重点研究了多尺度网格细胞模块编码与解码机制,提出一种大尺度空间下基于多尺度网格细胞模型的无人机类脑矢量导航方法。通过在多个吸引子神经网络中引入指数递增型速度增益因子,提出了多尺度网格细胞路径积分方法(编码);基于余数系统框架提出了大尺度空间网格细胞矢量导航解算方法(解码)。仿真实验结果表明:所提方法能够适应在最大范围38 km的大尺度空间中对速度、方位信息进行准确多尺度路径积分(编码),并准确解算当前距目标矢量(解码);验证了所提方法能够进行大尺度空间矢量导航,对进一步发展大尺度空间下无人机三维类脑矢量导航方法具有较好的参考意义。

Animals have excellent ability to perform vector navigation,which provides a useful biological model for developing a very promising paradigm of autonomous intelligent goal-directed navigation for UAV working in unknown environment.Firstly,the mechanisms of animal vector navigation are discussed.Then the mechanisms of encoding and decoding of multi-scale grid cells are studied,and a brain-inspired large-scale vector navigation method for UAV based on model of multi-scale grid cells is proposed and assessed.The models of path integration of multi-scale grid cells are designed by introducing exponentially increasing velocity gain factors into several continuous attractor neural networks.A method for solving the vector navigation of large-scale grid cells is proposed by employing residue number system.The simulation results show that the multi-scale path integration of velocity and azimuth can be realized accurately in a large-scale space with a maximum range of 38 km,and the vector between the target and the current positions can be calculated accurately in the proposed method,which suggest that multi-scale grid cells can be used for large-scale vector navigation and provide a meaningful reference for further developing brain-inspired large-scale 3D vector navigation method of UAV.