CNN-Based Fault Detection of Scan Matching for Accurate SLAM in Dynamic Environments.

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Simultaneous Localization and Mapping (SLAM) is crucial for robot navigation.
• Dynamic environments with moving objects pose significant challenges to traditional scan-matching algorithms.
• LiDAR-based scan matching can fail due to environmental changes caused by dynamic objects.

Purpose of the Study:

• To propose a robust method for detecting faults in scan-matching algorithms used in 2D SLAM.
• To improve the accuracy and reliability of SLAM in dynamic environments.
• To leverage Convolutional Neural Networks (CNNs) for scan-matching fault detection.

Main Methods:

• Raw 2D LiDAR scan data is processed using Iterative Closest Points (ICP) for initial scan matching.
• Matched scan data is converted into image format for input into a CNN model.
• The CNN model is trained to identify and classify scan-matching faults.

Main Results:

• The proposed CNN-based method accurately detects scan-matching faults across various dynamic environments.
• Experimental evaluations confirm the method's effectiveness in real-world scenarios.
• The system demonstrates robustness against environmental changes induced by dynamic objects.

Conclusions:

• CNN-based fault detection offers a reliable solution for enhancing scan-matching robustness in SLAM.
• The developed method significantly improves the performance of 2D SLAM in challenging dynamic environments.
• This approach contributes to more accurate and dependable robot localization and mapping.