中文摘要：

同步定位与地图构建（SLAM）是实现机器人自主定位的核心问题之一,Rao-Blackwellised粒子滤波器（RBPF）作为一种SLAM定位的有效方法,被广泛应用在实时定位领域中,但由于其随着粒子数目的增加会频繁重采样从而导致粒子退化问题。为了解决该问题,改善SLAM性能,提出了一种基于改进小生境遗传优化的RBPF SLAM算法INGO-RBPF,采用改进的Rao-Blackwellised粒子滤波器解决SLAM路径估计问题,采用扩展卡尔曼滤波器解决SLAM地图估计问题。最后通过MATLAB仿真表明INGO-RBPF算法具有较高的估计精度和稳定性,抗干扰能力较强,定位较准确,比较适合应用在SLAM实时定位中。

英文摘要：

Simulation localization and mapping （SLAM） is one of the key problems in realizing robot self-navigation. As an effective method for SLAM location, it widely used Rao-Blackwellised particle filter（RBPF） in the field of real time location. However, the RBPF behavior of frequent resampling results in particle impoverishment problem along with particles increased. In order to solve the problem and improve the algorithm performance, this paper proposed a RBPF SLAM algorithm based on improved niched genetic optimization （INGO-RBPF）. The INGO-RBPF algorithm solves the robot path estimation using im- proved Rao-Blackwellised particle filter（ PF）, and solves the map estimation using extended Kalman filter （EKF）. Finally the MATLAB simulations prove that INGO-RBPF performs well on estimated accuracy, stability, disturbance and location accura- cy, and therefore it is suitable to apply in SLAM real-time location.