Vision-based Autonomous Crop Row Navigation for Wheeled Mobile Robots using Super-twisting Sliding Mode Control

Abstract

This work presents a new robust image-based visual servoing (rIBVS) approach for wheeled mobile robots (WMRs) endowed with a single monocular camera to carry out autonomous navigation in row crop fields. Then, we design a robust vision-based controller by using the super-twisting algorithm (STA) approach to stabilize the robot motion in the presence of model inaccuracies caused by imperfect camera calibration, and trajectory perturbations due to different plant distributions and high robot driving velocities. The rIBVS approach switches between column and row visual primitives extracted from the images, allowing WMRs to execute the navigation task in two phases: the crop row reaching and the crop row following. To illustrate the effectiveness and feasibility of the proposed control methodology, 3D computer simulations are executed in the ROS-Gazebo simulator using a differential-drive mobile robot (DDMR) navigating autonomously in an ad-hoc developed row crop agricultural environment.