LIDAR-based detection of furrows for agricultural robot autonomous navigation

Abstract

Robust and accurate autonomous navigation is a main challenge in agricultural robotics. This paper presents a LIDAR-based processing system for autonomous robot navigation in crops with high vegetation density. The method detects and locates the crop furrows and provides them to the robot control system, which guides the robot such that its caterpillar tracks move along the furrows preventing damages in the crop. The proposed LIDAR-based processing pipeline includes various inconsistencies removal and template matching steps to deal with the high noise level of LIDAR scans. It has been implemented in C++ using ROS Noetic and validated in two different plantations with different crop growth status.