双程UFIR滤波算法及其在INS/双目视觉机器人组合导航的应用

Double-Pass UFIR Filtering and Its Application in INS/Binocular Vision Robot Integrated Navigation

为了提高移动机器人的室内定位精度,提出了一种惯性导航系统(Inertial Navigation System,INS)/双目视觉松组合导航模型,在INS与双目视觉同步获取到移动机器人位置信息的基础上,通过滤波算法对INS与双目视觉测量得到的移动机器人位置进行数据融合,充分融合INS和双目视觉各自的优势,最终得到当前时刻的移动机器人的位置估计。针对滤波算法,提出了双程无偏有限脉冲响应(Unbiased Finite Impulse Response,UFIR)滤波算法,该算法首先采用自适应UFIR前向滤波,通过引入预先设置的滤波窗口长度以及马氏距离,完成对当前时刻UFIR滤波算法最优滤波窗口长度的计算。在此基础上,通过后向UFIR平滑以及再次前向UFIR滤波,最终得到当前时刻最优的移动机器人位置估计。实际测试结果表明,所提出的双程UFIR算法能够有效地提高室内移动机器人的定位精度,与传统UFIR滤波算法相比,所提出的位置均方根误差(Root Mean Square Error,RMSE)降低了11.64%。

In order to further improve the indoor localization accuracy of mobile robot,a loosely integrated navigation model of inertial navigation system(INS)and binocular vision is proposed.At the first place,INS and binocular vision are adopted to estimate the position of the target robot in parallel.Then,the double-pass unbiased finite impulse response(UFIR)filtering algorithm is proposed to complete the position data fusion of the mobile robot acquired through INS and binocular vision respectively.In detail,the adaptive UFIR forward filtering is first adopted to compute the length of the optimal filtering window by using the preset filtering window length and the Mahal-anobis distance.After that,the optimal mobile robot position estimation at the current time can be obtained via the backward unbiased finite impulse response(UFIR)smoothing and second forward UFIR filtering,fully integrating the respective advantages of INS and bin-ocular vision.The real test results show that the proposed double-pass UFIR algorithm can effectively improve the localization accuracy of indoor mobile robot.Compared with the traditional UFIR filter,the root mean square error(RMSE)of the position of the mobile robot of the double-pass UFIR algorithm is reduced by 11.64%.