LFVB-BioSLAM: A Bionic SLAM System with a Light-Weight LiDAR Front End and a Bio-Inspired Visual Back End

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Vision

Background:

• Simultaneous Localization and Mapping (SLAM) is vital for autonomous robots but often power-intensive.
• Bionic SLAM algorithms inspired by animal navigation are emerging but face accuracy and robustness challenges.
• Existing bionic SLAM, like RatSLAM, struggles in complex environments.

Purpose of the Study:

• To introduce LFVB-BioSLAM, a novel bionic SLAM system with improved practicality.
• To develop a bionic SLAM framework facilitating research advancements.
• To enhance autonomous robot navigation capabilities through bio-inspired design.

Main Methods:

• Implemented a light-weight LiDAR-based front end using range flow for odometry estimation.
• Developed a bio-inspired, vision-based back end utilizing monocular RGB cameras for loop closure and path integration.
• Integrated LiDAR odometry with a biologically-inspired visual processing pipeline.

Main Results:

• LFVB-BioSLAM demonstrated superior accuracy and robustness in real-world experiments.
• The system outperformed RatSLAM (vision-based bionic SLAM) and RF2O (laser-based odometry).
• The proposed framework offers a practical solution for bionic SLAM.

Conclusions:

• LFVB-BioSLAM presents a significant advancement in bionic SLAM technology.
• The hybrid LiDAR-vision approach enhances navigation system performance.
• This work paves the way for more capable and efficient bio-inspired autonomous robots.