室内环境下立体视觉惯导融合定位  

Fusion location method of stereo vision inertial navigation system in indoor environment

作　　者：敖龙辉 郭杭[1] AO Longhui;GUO Hang(Nanchang University,Nanchang 330031,China)

机构地区：[1]南昌大学

出　　处：《测绘通报》2019年第12期12-17,共6页Bulletin of Surveying and Mapping

基　　金：科技部重点研发项目(2016YFB0502002);国家自然科学基金(41764002;41374039);南昌大学研究生创新专项资金项目(CX2018156)

摘　　要：针对室内服务机器人在居家环境下导航定位问题,本文研究了紧耦合非线性优化的立体视觉惯性融合导航方法。本文采用预积分、边缘化、滑动窗口优化等关键技术,提出了一种稳健的视觉惯性导航系统初始化方法。运用于室内家庭服务机器人中,设计并实现了对应的视觉惯性融合导航系统。在搭建的模拟居家环境下,验证了本文系统的初始化方法能够提供稳健、准确的系统初值;最后通过试验验证了本文定位系统的准确性与稳定性,定位误差可控制在0.1 m以内。Aiming at the navigation and positioning problem of indoor service robots in the home environment,this paper studies the stereointegrated inertial fusion navigation method based on tightly coupled nonlinear optimization.The method adopts key technologies such as preintegration,edge-based and sliding window optimization,and proposes a robust initialization method of visual inertial navigation system.Used in indoor home service robots,the corresponding visual inertial fusion navigation system is designed and implemented.In the simulated home environment,the initialization method of the system can provide robust and accurate initial values.Finally,the accuracy and stability of the positioning system are verified by experiments.The positioning error can be controlled within 0.1 m.

关 键 词：IMU预积分 滑动窗口优化 室内机器人 立体视觉惯导融合 导航定位 

分 类 号：P228[天文地球—大地测量学与测量工程]