High-Frequency Ground Segmentation for Autonomous Mobile Robots: A RANSAC-Based Approach

In this work, a RANSAC-based algorithm was developed for ground segmentation on point clouds obtained from 3D LIDAR sensors. The algorithm employs both distance and normal angle criteria to construct a robust ground plane model, even in the presence of noise and outliers. In the initial stage, a height filter is applied to analyze only the points associated with the ground. Subsequently, the RANSAC method identifies the plane model with the highest number of inliers, dividing the point cloud into two groups: ground and obstacles. The proposed method demonstrated real-time performance with a 20 Hz LIDAR sensor, delivering higher speed and accuracy compared to alternative approaches. This study provides an effective and reliable solution for ground segmentation in autonomous systems.