Block-Map-Based Localization in Large-Scale Environment

TL;DR

This work tackles large-scale, GPS-denied robot localization where map size imposes significant computational burdens. It introduces Block Map (BM) based localization, generating overlapping block maps from offline keyframes, using a multi-resolution BM pyramid with Branch-and-Bound Search (BBS) for coarse global localization, and applying a dynamic sliding-window factor-graph optimization for accurate pose tracking across BM switches. The method demonstrates robust global localization up to scales exceeding $6\ \mathrm{km}$ and achieves improved tracking accuracy and efficiency on public datasets (NCLT and M2DGR) compared with state-of-the-art map-based and LiDAR-IMU fusion approaches. These contributions enable real-time, scalable, and reliable localization in large environments, with potential applications in logistics and service robots that operate across expansive areas.

Abstract

Accurate localization is an essential technology for the flexible navigation of robots in large-scale environments. Both SLAM-based and map-based localization will increase the computing load due to the increase in map size, which will affect downstream tasks such as robot navigation and services. To this end, we propose a localization system based on Block Maps (BMs) to reduce the computational load caused by maintaining large-scale maps. Firstly, we introduce a method for generating block maps and the corresponding switching strategies, ensuring that the robot can estimate the state in large-scale environments by loading local map information. Secondly, global localization according to Branch-and-Bound Search (BBS) in the 3D map is introduced to provide the initial pose. Finally, a graph-based optimization method is adopted with a dynamic sliding window that determines what factors are being marginalized whether a robot is exposed to a BM or switching to another one, which maintains the accuracy and efficiency of pose tracking. Comparison experiments are performed on publicly available large-scale datasets. Results show that the proposed method can track the robot pose even though the map scale reaches more than 6 kilometers, while efficient and accurate localization is still guaranteed on NCLT and M2DGR.

Figures (5)

• Figure 1: Dividing a large-scale map into multiple block maps has the potential to enable robots to utilize limited resources to achieve arbitrary scale navigation and service tasks. Consequently, the utilization of block map-based localization becomes indispensable.
• Figure 2: Overview of our Block-Map-Based Localization which consists of four main modules: a block maps generation module, a pose initialization module, a map retrieval server module and a pose tracking module.
• Figure 3: Block Maps (BMs) Generation Method
• Figure 4: The result comparison of different block maps separation methods. Our method considers the overlap between adjacent block maps, and the robot will not lose prior information when switching maps at the boundary.
• Figure 5: Subfigures (a) to (d) show smooth transition and the localization path passing through 4 consecutive block maps on nclt_2. Our switching strategy is able to maintain spatial continuity with good performance.