[Paper Review] A Unified 3D Mapping Framework using a 3D or 2D LiDAR

Weikun Zhen, Sebastian Scherer|arXiv (Cornell University)|Oct 30, 2018

Robotics and Sensor-Based Localization3 references4 citations

TL;DR

This paper proposes a unified 3D mapping framework that supports both 2D and 3D LiDARs—fixed or rotating—by extending Cartographer SLAM with a generalized localization module based on error-state Kalman filtering and scan matching. The framework achieves real-time, high-quality 3D mapping across diverse LiDAR platforms with minimal parameter tuning, reducing drift to 0.003° per meter and enabling consistent loop closure across indoor and outdoor environments.

ABSTRACT

Simultaneous Localization and Mapping (SLAM) has been considered as a solved problem thanks to the progress made in the past few years. However, the great majority of LiDAR-based SLAM algorithms are designed for a specific type of payload and therefore don't generalize across different platforms. In practice, this drawback causes the development, deployment and maintenance of an algorithm difficult. Consequently, our work focuses on improving the compatibility across different sensing payloads. Specifically, we extend the Cartographer SLAM library to handle different types of LiDAR including fixed or rotating, 2D or 3D LiDARs. By replacing the localization module of Cartographer and maintaining the sparse pose graph (SPG), the proposed framework can create high-quality 3D maps in real-time on different sensing payloads. Additionally, it brings the benefit of simplicity with only a few parameters need to be adjusted for each sensor type.

Motivation & Objective

• To address the lack of cross-platform compatibility in LiDAR-based SLAM systems, which require separate algorithms for different sensor types.
• To enable real-time, high-accuracy 3D mapping using both fixed and rotating 2D/3D LiDARs within a single framework.
• To minimize parameter tuning across platforms by maintaining a consistent configuration strategy despite sensor differences.
• To improve robustness in environments with limited vertical field-of-view, such as narrow corridors, by leveraging rotating LiDARs.

Proposed method

• Replacing Cartographer’s original localization module with an error-state Kalman filter (ESKF) that performs scan matching against a distance map to estimate pose.
• Using a sparse pose graph (SPG) refinement component to optimize poses and reduce drift, maintaining compatibility with the original Cartographer backend.
• Processing individual laser scans in real time, matching each to the map to improve accuracy over IMU-only integration.
• Supporting both fixed 3D LiDARs and rotating 2D/3D LiDARs by decoupling the localization method from sensor type.
• Adjusting only a few key parameters—such as distance map resolution, scan accumulation count, and optimization frequency—when switching platforms.
• Employing a distance map representation that enables efficient scan matching and supports variable sensor configurations.

Experimental results

Research questions

• RQ1 Can a single SLAM framework support both fixed and rotating 2D/3D LiDARs with minimal reconfiguration?
• RQ2 How does the proposed localization method compare to IMU-only integration in terms of drift and accuracy for rotating LiDARs?
• RQ3 What parameters significantly affect performance across different LiDAR types and environments?
• RQ4 How does field-of-view limitation impact height estimation and loop closure in narrow corridors?
• RQ5 To what extent can parameter tuning be standardized across diverse robotic platforms?

Key findings

• The framework achieves a rotational drift rate of 0.003° per meter, demonstrating high consistency in long-term mapping.
• In a narrow hallway with a fixed 3D LiDAR, height estimation was unreliable due to limited vertical field-of-view, leading to delayed loop closure.
• Rotating LiDARs significantly improved mapping robustness by increasing field-of-view and enabling better 3D structure capture.
• For rotating 2D LiDARs, low moving speed (≈0.5 m/s) was critical to accumulate sufficient scan points for stable submap updates.
• Only a few parameters—such as distance map resolution, scan accumulation count, and optimization frequency—needed adjustment when switching platforms.
• The method achieved consistent loop closure and low error peaks in roll (0.14°), pitch (0.15°), and yaw (0.24°) across all test environments.