Range-Aided Pose-Graph-Based SLAM: Applications of Deployable Ranging Beacons for Unknown Environment Exploration
Range-Aided Pose-Graph-Based SLAM: Applications of Deployable Ranging Beacons for Unknown Environment Exploration
复制标题
DOI：
10.1109/lra.2020.3026659
复制
发表时间：
2021-01
影响因子：
5.2
通讯作者：
Nobuhiro Funabiki;B. Morrell;J. Nash;Ali-akbar Agha-mohammadi
中科院分区：
文献类型：
--
作者：
Nobuhiro Funabiki;B. Morrell;J. Nash;Ali-akbar Agha-mohammadi
同时定位与地图构建（SLAM）是机器人在未知环境中探索的关键部分。大规模SLAM通常存在限制漂移的挑战，并且需要环路闭合来校正累积误差。然而，实时环路闭合检测可能受到高计算成本和对离群值的敏感性的限制，特别是在感知上具有挑战性的、未知的和大规模的环境中。这些挑战的一个解决方案是从机器人部署测距信标，并利用它们进行明确的地方识别。在这封信中，我们开发了一个架构，利用部署的测距信标的姿态图SLAM。首先，我们突出了稀疏部署的测距信标的距离测量的固有模糊性，通过分析在应用传统的方法处理距离测量定位的挑战。为了处理这种模糊性，并解决大规模的，感知上具有挑战性的环境中的环路闭合的挑战，我们提出了范围辅助环路闭合（RA-LC），一种算法，结合了基于信标的位置识别与几何环路闭合。我们在硬件系统中实现了RA-LC，并在现场测试中证明了RA-LC的效率。我们的研究结果表明，RA-LC可以实现几何仅基于激光雷达的环路闭合的等效精度，具有显着降低的计算成本（减少95%）和8倍的离群值。
Simultaneous Localization and Mapping (SLAM) is a critical part of robotic exploration in unknown environments. SLAM over large scales typically presents challenges with limiting drift and requires loop closures to correct accumulated errors. However, real-time loop closure detection can be limited by high computational cost and susceptibility to outliers, especially in perceptually challenging, unknown, and large-scale environments. One solution to these challenges is to deploy ranging beacons from a robot and to utilize them for unambiguous place recognition. In this letter, we develop an architecture to take advantage of deployable ranging beacons in pose-graph SLAM. First, we highlight the inherent ambiguity in range measurements for sparsely deployed ranging beacons by analyzing the challenges in applying conventional methods of processing range measurements for localization. To both handle this ambiguity and address the challenges of loop closure for large-scale, perceptually challenging environments, we propose range-aided loop closure (RA-LC), an algorithm that combines beacon-based place recognition with geometric loop closures. We implement RA-LC in our hardware system and demonstrate the efficiency of RA-LC in field tests. Our results show that RA-LC can achieve equivalent accuracy of geometric-only lidar-based loop closure with significantly lower computational cost (95% reduction) and eight-times fewer outliers.