基于稀疏化的快速扩展信息滤波SLAM算法

Fast Extended Information Filter SLAM Algorithm Based on Sparsification

随着SLAM技术的不断发展,计算效率已经成为制约SLAM发展的主要因素.所提出的算法从稀疏化的角度对扩展信息滤波SLAM算法进行改进.根据信息矩阵几乎稀疏的特点,该算法在合理稀疏化信息矩阵的同时利用环闭合检测技术,不仅大大提高了算法的计算效率,而且所得到的估计结果也很精确.通过仿真对信息矩阵稀疏化、算法效率、重定位以及误差和协方差四个关键问题进行了分析.分别就室内具有摄像头的两轮机器人和室外具有激光雷达的四轮机器人的情况进行了实验讨论.仿真与实验结果表明了所提算法的有效性.

With the continuous development of SLAM technology, computational efficiency has become the main obstacle in SLAM development. From the view of sparsification, an improved algorithm based on extended information filter SLAM is introduced. According to the sparsification feature of information matrix, the algorithm not only improves computation efficiency but also maintains accuracy of the estimated result through using proper sparsification of information matrix as well as loop closure detection. Four key problems of information matrix sparsification, i. e. , algorithm efficiency, relocalization, error and covariance are analyzed with simulation. Two cases are discussed through experiments, i.e. indoor two-wheel robot with camera and outdoor four-wheel robot with laser scanner, and the results of simulation and experiments verify the validity of the proposed algorithm.