Graph SLAM Built over Point Clouds Matching for Robot Localization in Tunnels

Area of Science:

• Robotics
• Computer Vision
• Geospatial Analysis

Background:

• Simultaneous Localization and Mapping (SLAM) is crucial for autonomous systems.
• Feature-poor environments, such as tunnels, pose significant challenges for traditional SLAM algorithms.
• Accurate environmental reconstruction is vital for applications like autonomous navigation and infrastructure inspection.

Purpose of the Study:

• To present a novel, modular, and sensor-agnostic graph SLAM algorithm tailored for challenging environments.
• To improve the accuracy of environmental reconstruction and robot localization in feature-poor settings.
• To develop a robust system for generating precise 3D models of surroundings.

Main Methods:

• The algorithm employs a graph-based SLAM approach, integrating point cloud data.
• It features three core modules: initial pose estimation, pose refinement using point clouds, and over-constrained graph generation.
• The system utilizes loop closures to enhance trajectory accuracy and environmental consistency.

Main Results:

• The developed graph SLAM algorithm demonstrates effectiveness in environments with limited features.
• It enables precise environmental reconstruction and accurate robot trajectory estimation.
• The modular design allows for expandability to various point cloud-generating sensors.

Conclusions:

• The proposed graph SLAM algorithm offers a robust solution for localization and mapping in difficult environments.
• It significantly improves environmental modeling capabilities, particularly in tunnel applications.
• The algorithm's generic nature and sensor expandability make it a versatile tool for robotic applications.