Adaptive Kalman Filter-Based SLAM in LiDAR-Degenerated Environments

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• 2D LiDAR is crucial for mobile robot SLAM due to its cost-effectiveness and ease of installation.
• Traditional 2D LiDAR SLAM methods struggle with robustness and accuracy in environments with sensor degradation.
• Enhancing SLAM performance in challenging conditions is vital for reliable mobile robot navigation.

Purpose of the Study:

• To develop an innovative SLAM method that improves robustness and accuracy in LiDAR-degenerated environments.
• To enhance the pose estimation and mapping capabilities of mobile robots.
• To provide a more reliable navigation solution for mobile robots operating in complex settings.

Main Methods:

• A novel SLAM approach combining front-end positioning with back-end optimization.
• Utilizing an adaptive Kalman filter (AKF) for robot pose, sensor bias, and calibration estimation.
• Employing a particle filter (PF) with front-end pose constraints for optimized pose estimation and mapping.

Main Results:

• The AKF dynamically adjusts noise variances based on residuals for improved estimation.
• The PF incorporates front-end pose constraints to prevent mismatches and boost accuracy.
• Experimental results demonstrate significant improvements in positioning precision compared to existing SLAM algorithms.

Conclusions:

• The proposed SLAM method offers superior robustness and accuracy in challenging environments.
• The integration of AKF and PF with pose domain constraints effectively addresses SLAM limitations.
• This advancement contributes to more reliable and precise mobile robot navigation systems.