一种改进的RatSLAM仿生导航算法  

An Improved Bionic Navigation Algorithm Based on RatSLAM

作　　者：张潇[1] 胡小平[1] 张礼廉[1] 马涛[1] 王玉杰[1] 

机构地区：[1]国防科技大学机电工程与自动化学院

出　　处：《导航与控制》2015年第5期73-79,47,共8页Navigation and Control

摘　　要：RatSLAM是模拟鼠类感知环境机制提出的一种定位与构图的导航算法,算法的提出者利用该方法成功地进行了66km的车载试验。在机器人导航领域,RatSLAM是非常优秀的纯视觉仿生导航算法。然而,该算法完全依赖视觉信息,在复杂的环境中存在可靠性低、导航精度不高的问题。在RatSLAM的基础上引入光学双轴速度传感器和MIMU信息,建立了融合光学双轴速度传感器和MIMU信息的航位推算模型,对RatSLAM仿生导航算法进行了合理改进,搭建了硬件系统并进行了动态车载实验。实测结果表明改进的仿生导航算法环境适应性更强、导航精度更高。RatSLAM is a bio-inspired navigation algorithm for simultaneous localization and mapping,simulating rodents’ mechanism to perceive the environment.It is successfully tested on a vehicle platform which travels 66 km long.In the field of robot navigation,RatSLAM is an extraordinary bionic navigation algorithm using pure vision.However,the algorithm suffers from the problem of low reliability and accuracy in complex environment.In this paper,the 2-axis optical velometer and MIMU are introduced into the RatSLAM framework.The path integration model is built by fusing the velometer and MIMU information.In addition,the RatSLAM framework is adjusted and the hardware system for the dynamic vehicle experiment is.constructed.Experimental results show that the improved algorithm has stronger adaptability,higher navigation accuracy.

关 键 词：RatSLAM 仿生导航算法 数据融合 视觉导航 

分 类 号：V249.3[航空宇航科学与技术—飞行器设计]