计算复杂度降低的基于CDKF的SLAM算法

摘要

In order to reduce the computational complexity of the CDKF SLAM algorithm for large-scale environment, this paper proposed an improved CDKF SLAM algorithm which was presented in the context of the linear-regression Kalman filter. Based on the properties of SLAM, it improved the sampling strategy by reconstructing the estimated state and its covariance during prediction and measurement update. The complexity was thus reduced to O (n2). Simulation experiments in different scale environments and experiments of the car park database proves that the proposed algorithm keep the same accuracy as the general CDKF SLAM, while its running time became much shorter. This makes it more suitable for applications in large-scale environment.%为了降低移动机器人基于中心差分卡尔曼滤波(CDKF)的同时定位与地图构建(SLAM)算法的计算复杂度,使其适于较大规模环境中的应用,提出了一种改进的CDKF SLAM算法.该算法以CDKF的线性回归卡尔曼滤波(LRKF)形式为基础,利用SLAM自身特点,重构其预测和观测更新过程中的状态变量及相应的方差矩阵,改进CDKF的采样方法,从而将CDKF SLAM算法的计算复杂度降为O(n2).不同规模环境中的仿真实验及停车场数据集的实验验证了在不改变CDKF SLAM算法估计准确度的条件下,本文算法的运行时间明显缩短,更适于大规模环境中的应用.