A Localization and Mapping Algorithm Based on Improved LVI-SAM for Vehicles in Field Environments.

Area of Science:

• Robotics
• Computer Vision
• Autonomous Systems

Background:

• Accurate vehicle localization and mapping are crucial for autonomous driving.
• Complex field environments (e.g., open, bumpy, GNSS-denied) pose significant challenges to existing localization and mapping techniques.
• Current methods struggle with environmental emptiness, terrain variability, and lack of satellite signals.

Purpose of the Study:

• To develop a robust and accurate Lidar, Inertial, and Visual fusion SLAM algorithm for autonomous vehicles operating in challenging field environments.
• To address the limitations of existing methods in GNSS-denied and complex terrains.
• To improve both localization accuracy and real-time mapping capabilities.

Main Methods:

• Proposed an LVI-SAM-based algorithm integrating Lidar, Inertial, and Visual data for SLAM.
• Designed a joint Lidar front end for pose estimation and correction using Super4PCS, ICP, and NDT algorithms.
• Implemented a strategy for balancing localization accuracy and real-time performance through hierarchical pose correction and estimation frequencies.

Main Results:

• The proposed method demonstrated reduced translational error by approximately 4.7% compared to the baseline LVI-SAM.
• Successfully generated real-time 3D point cloud maps of complex environments.
• Achieved high-precision and high-robustness localization and mapping for vehicles in challenging field conditions.

Conclusions:

• The enhanced LVI-SAM algorithm effectively overcomes the limitations of autonomous driving in GNSS-denied and complex field environments.
• The integrated approach provides a reliable solution for real-time, high-accuracy localization and mapping.
• This advancement contributes to the development of more capable and dependable autonomous vehicles for diverse operational settings.