鼠类脑细胞导航机理的移动机器人仿生SLAM综述

摘要

Aiming at the probabilistic algorithms that have shortcomings such as large computation, high complexity, and failure to find the global optimum, a variety of cells, including border cells, view cells, grid cells and speed cells, are applied to simultaneous localization and mapping (SLAM) in order to construct a BVGSP-SLAM model with multi-celled navigation. A loop closure detection algorithm with keyframe matching is added to SLAM to avoid lighting that changes based on the direction and angle of the light. Speed cells and border cells are added to SLAM to avoid the influ-ence of mobile obstructions. A mathematical model of mixed cells that analyzes robustness and real-time performance of the system is proposed. This project will develop an integrated approach for modeling, simulation and experimental veri-fication, which provide an important theoretical reference on SLAM.%针对同步定位与地图构建(SLAM)问题中传统概率算法存在计算量大、复杂度高、易陷于局部最优解等问题,本文提出一种未来深入研究的方法建议,将鼠类脑细胞中边界细胞(border cells)、局部场景细胞(view cells)、网格细胞(grid cells)、速度细胞(speed cells)、位姿细胞(pose cells)等具有定位导航功能的细胞应用于SLAM研究中,构建一种基于多细胞导航机制的BVGSP-SLAM模型.结合具有实时关键帧匹配的闭环检测算法以避免光线变化对SLAM的影响,融入速度细胞和边界细胞以避免移动障碍物对SLAM的影响,利用鼠类混合细胞衍生出的数学模型分析该系统的鲁棒性和实时性.将生物细胞模型引入SLAM,并形成了建模、仿真与实验验证一体化的研究体系,为移动机器人SLAM研究领域多样化提供重要的理论参考.