基于改进时间差分的视觉/惯性组合导航研究
2023,31(3):267-274
摘要：针对全球导航卫星系统(GNSS，Global Navigation Satellite System)在军事战争、室内和水下等情况下存在因信号缺失导致的全球定位系统（GPS，Global Positioning System）无法使用和惯性导航系统(INS，Inertial Navigation System)状态误差发散过快的问题，提出了一种基于连续帧时间差分视觉辅助导航的方法。为了抑制状态误差的快速发散，提高INS在长航时工作上的性能，分析了机器视觉连续帧间差分法,并对其进行了计算上的改进，设计了一种时间差分视觉/惯性组合系统，并进行了仿真实验和分析。结果与纯INS相比，均方根误差(RMSE，Root Mean Square Error)在北向位置上减少了19.0%，在东向误差上减少了32.1%，表明该方法有效抑制了纯惯性导航速度和位置的误差发散，延长了惯性导航的可用时间。
关键词：惯性导航;机器视觉;零速校准;卡尔曼滤波;组合导航;时间差分
Research on vision/inertial integrated navigation based on improved time difference
Abstract：Aiming at the problems that GPS cannot be used and INS status error diverges too fast due to signal loss in GNSS during military war, indoor and underwater, a visual aid navigation method based on continuous frame time difference was proposed. In order to restrain the rapid divergence of state errors and improve the performance of INS in long endurance work, the continuous frame difference method of machine vision was analyzed and improved in calculation. A time-difference vision/inertia combined system was designed and simulated. Results Compared with pure INS, RMSE reduced by 19.0% in the northbound position and 32.1% in the eastbound error, indicating that this method effectively inhibited the error divergence of pure inertial navigation speed and position, and extended the available time of inertial navigation.
Key words：inertial navigation; machine vision; zero-speed calibration; Kalman filtering; integrated navigation; time difference
收稿日期：2022-11-12
基金项目：山西省青年科技研究基金资助项目(201901D211251)
