An Adaptive SLAM Algorithm Based on Strong Tracking UKF
R Obot, Wenling Zhang, Mingqing Zhu, Zonghai Chen · Robot · 2010
Abstract: Unscented Kalman filter (UKF) is lack of adaptive on-line adjustment ability that seriously decreases the estima-tion accuracy of system state. To deal with this problem, this paper proposes an improved SLAM (simultaneous localizationand mapping) algorithm that combines the strengths of strong tracking filter (STF) and UKF. Each sampling point of UKFis updated by STF, the effects of noises on system state estimation are suppressed by optimizing filter gains, and the systemstate estimation converges to real values quickly. Performances of several SLAM algorithm in different noisy environmentsare compared by simulation. The experimental results show that this adaptive SLAM algorithm based on STF and UKF is ofbetter adaptability and robustness.Keywords: simultaneous localization and mapping; UKF-SLAM; strong tracking filter; adaptive filter 1 引言（Introduction） 未知环境中，机器人在依靠自身所带的传感器递增地建立环境地图的同时，利用所建立的地图同步刷新自身的位置，即同时定位和地图创建，简称SLAM．SLAM 问题最早由Smith、Self 和Cheese-man 提出 [1] ，他们运用扩展卡尔曼滤波法（EKF），对状态空间中的机器人位姿和地图特征同时进行估计．但是EKF 方法存在计算量过大、精度不高、甚至发散等不足．针对这些不足，有人提出了粒子滤波（PF）