Novel Laser-Based Obstacle Detection for Autonomous Robots on Unstructured Terrain.

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Autonomous robots require robust obstacle detection for navigation in unstructured environments.
• Existing methods may struggle with the complexity and variability of natural terrain.

Purpose of the Study:

• To propose a novel laser-based obstacle detection algorithm for autonomous robots operating on unstructured terrain.
• To enhance the accuracy and efficiency of obstacle identification and characterization.

Main Methods:

• Utilized 3D laser point clouds processed with VoxelGrid filtering and Gaussian kernel functions for edge feature extraction.
• Employed an optimized Euclidean clustering algorithm (super-voxel) for obstacle point cloud segmentation.
• Applied a Levenberg-Marquardt back-propagation (LM-BP) neural network for obstacle characteristic recognition.

Main Results:

• The proposed post-processing algorithm successfully identified obstacle edge features from reconstructed point clouds.
• Clustering and recognition of obstacle characteristics were demonstrated on both existing datasets and real-world data.
• Experimental results validated the feasibility and performance of the novel approach.

Conclusions:

• The developed laser-based method offers a viable solution for obstacle detection in unstructured terrains for autonomous robots.
• The integration of Sobel operator, super-voxel clustering, and LM-BP neural network provides effective obstacle identification.
• This approach contributes to safer and more reliable autonomous navigation in challenging environments.