LiDAR-Based Online Navigation Algorithm For An Autonomous Agricultural Robot

Abstract

In this paper, we propose a new approach for navigation of autonomous robots between crop rows in agriculture. This method is implemented by projecting 2D Light Detection and Ranging (LiDAR) data onto the robot’s movement direction to perform one-dimensional Density-Based Spatial Clustering of Applications with Noise (DBSCAN). The mapping and localization (MAL) are obtained by combining the location of the virtual landmark from the DBSCAN with robot’s position updated from the Particles filter. The results of this method map and estimate the robot’s location simultaneously in one scan. Each robot’s position depends on the LiDAR data information for the current scan and the previous scan. Data association to build global path is achieved by combining data from a number of consecutive scans and by the Kalman filter. The global trajectory, which is created by combining local positions, allows the robot to navigate autonomously in real time without having to go through the prior phases of collecting all of the data from the crop field. The article also performs FIR filter correction with different parameters to enhance the effectiveness of the proposed method.