A low‐cost and efficient autonomous row‐following robot for food production in polytunnels

Abstract

In this paper, we present an automatic motion planner for agricultural robots that allows us to set up a robot to follow rows without having to explicitly enter waypoints. In most cases, when robots are used to cover large agricultural areas, they will need waypoints as inputs, either as premeasured coordinates in an outdoor environment, or as positions in a map in an indoor environment. This can be a tedious process as several hundreds or even thousands of waypoints will be needed for large farms. In particular, we find that in unstructured environments such as the ones found on farms, the need for waypoints increases. In this paper, we present an approach that enables robots to safely traverse not only between straight rows but also through curved rows without the need for any predetermined waypoints. Most types of infrastructure found in agriculture, such as polytunnels, are built on uneven terrain, thus containing a mix of straight and curved plant rows, for which traditional methods of row following will fail. Different from traditional approaches of row following that only consider straight‐line‐of‐sight rows, our approach identifies the rows on each side with the goal of staying in the middle of the rows, even if the rows are not straight. Waypoints are only needed on the very extreme of the rows, and these will be automatically generated by the system. With our approach, the robot can just be placed in the corner of the field and will then determine the trajectory without further input from the user. We thus obtain an approach that can reduce the installation time from potentially hours to just a matter of minutes. The final autonomous system is low cost and efficient for various tasks that requires moving between plant rows inside a polytunnel. Several experiments were performed and the robot demonstrates 1.4% position drift over 21 m of navigation path.

Recognition of Bloom/Yield in Crop Images Using Deep Learning Models for Smart Agriculture: A Review

“…Obstacle-averting decisions in agricultural terrain are complicated. The fusion of sensors with multiple algorithms and multiple robots detects the defined obstacles, controls and navigates the agricultural environment [44][45][46]. The automation system increases the farm proficiency with the self-guided vehicles and autonomous execution of farm duties like spraying, pruning, mowing, thinning and harvesting.…”

“…Dual arm manipulators in a single harvester with one gripper dedicated to moving aside any obstacles, could successfully pick strawberries [16]. An autonomous robot can navigate in straight or curved rows without having hundreds of programmed waypoints, by utilizing a 2D laser scanner [46]. A 3D simulation with real-time geographic of 22 coordinates and a first order approximation model was designed as a skid-steering autonomous robot [60].…”

La revolución tecnológica y los desafíos para los trabajadores

“…El gremio de los horticultores -que hoy lucha cada vez que cae el precio de sus productos-no podrá dejar de ver en las redes sociales y en YouTube la cada vez mayor variedad de implementos automatizados. En el sector hortícola, que utiliza mucha mano de obra, la incorporación de alta tecnología seguramente está muy próxima, quizá sea muy costosa una cosechadora de tomates (Mastermac, 15 de Mayo de 2021) que puede trabajar centenares de hectáreas de cultivos por día, pero están otras como las máquinas low cost que pueden operar incluso en los invernaderos de polipropileno (Le, Ponnambalam, Gjevestad & From, 2020). En el mercado latinoamericano ya están disponibles cosechadoras de tomates, de yerba mate (Quaizel, 2020), y en prototipo la de cítricos (La Universidad patenta una pequeña cosechadora de cítricos, 18 de junio de 2020) que se suman a los ya muy conocidas de cereales y oleaginosas.…”

Robot-supervised Learning of Crop Row Segmentation

“…Onboard sensors such as scanning lidar or machine vision cameras enable direct sensing of the the crops and structures surrounding the robot. Lidar-based navigation has been shown to work well in structured environments such as strawberry polytunnels [1]. Vision-based crop row following using RGB images is a well-established strategy, typically employing colour (e.g.…”