非结构化环境下月球车快速即时定位与制图方法

A Fast Lunar Rover Simultaneous Localization and Map-Building Method under Unstructured Environment

激光雷达的实时环境扫描与已构建环境模型之间的匹配是月球车即时定位与制图(SLAM)过程中的关键步骤,其收敛速度和准确性直接决定了SLAM的成败。月球表面是一种典型非结构化环境,其环境场景特征复杂。若通过三维激光匹配方式则传感器数据量大,特征匹配的难度高,实时性差。针对于此,本文提出了一种基于ICP算法的激光点扫描匹配方法-类等高线匹配方法,将二维激光雷达的实时扫描数据与已构建的三维环境高程图相匹配,利用RBPF粒子滤波实现位姿与地图状态的估计,采用自行研制的小型激光雷达硬件平台,较好地实现了月球车原理样机的快速SLAM过程。实验表明,基于该方法的即时定位与制图过程对于月球车位姿估计的鲁棒性强,实时性好。

Alignment or registration of on-line scan sampled by Lidar with the established environment model is a critical step in lunar rover simultaneous localization and mapping（SLAM） .Performance of SLAM is dominated by the efficiency and the accuracy of the laser alignment procedure.The surface of lunar is a typical unstructured environment because of its complex structure.In this kind of environment,3D laser scan is intractable because the registration tends to be difficult and convergent slowly due to the considerable computation burden resulted by dense laser data.In light of this deficiency,an ICP-based efficient laser scan match approach,named‘Contour Like Registration’,is proposed in this paper.In this method,2D laser real-time scan data is matched with the 3D elevation map built so far.By employing Rao-Blackwellized Particle Filter for rover localization and map estimation,fast on-line lunar rover SLAM is obtained by using a compact Lidar.It is evaluated from experimental results that the proposed method yields consistent real-time lunar rover pose estimation.