News

Autonomous over the field

Robotics in agriculture

Author: Georg Supper

In agricultural sciences and among agricultural machinery manufacturers, work is being carried out on the automation and digitization of agriculture based on Industry 4.0. Today’s automation solutions for field work can be divided into two development areas.
The first would be the integration of more and more automation through the integration of sensors and computer technology into agricultural machinery. With the help of the appropriate software, modern agricultural machines are able to perform field work in a highly automated manner. (Schwich et al., 2019)
Another approach is the development of much smaller machine concepts, the so-called field robots. These robots have the potential to replace especially monotonous and repetitive, as well as physically demanding tasks. Several research papers are therefore concerned with the development of robots for a variety of agricultural work processes. (Bechar and Vigneault, 2016).

Selected field robots

Dino – Najo Technologies

The company Najo-Technologies was founded in 2012 and has three different robots in its portfolio: Oz, Ted (vineyard) and Dino (vegetables).

The Dino field robot is the company’s largest robot. It was developed for mechanical control of weeds. Its tools can handle multiple crop rows. It is controlled in the field by RTK-GPS and a vision system detects the plant rows. According to the manufacturer, the robot can work for 8h in good conditions.
Oz is a robot for weed control with a width of 40cm, a height of 60cm and a length of 100cm (including 130cm tool carrier). It weeds one row at a time and can cover about 48 rows of 100m in about four hours. (Naio Technology 2020)

Robotti – AgroIntelli

The Robotti field robot is designed for agricultural use in the field. Unlike other robotic solutions, this robot is designed as a tool carrier. It has a three-point linkage, a power take-off shaft, and hydraulic connections. This means that existing attachments up to a working width of 3m can be used. It is driven by a diesel engine. An RTK GPS system is used for navigation in the field. (AgroIntelli 2020).

Thorvald 2 – Saga Robotics

Thorvald 2 is a module-based robot that makes it possible to create very different robot configurations with the same basic modules. The main task ranges from phenotyping in wheat, UV treatment in glasshouse or in foil tunnel depending on the configuration. With this system, it is possible to respond to different applications and agricultural environments by customizing the composition of the robot. The robot is equipped with electric motors and powered by batteries. It is automated using a series of sensors such as GPS and LiDAR (Light Detection and Ranging) sensors and cameras that vary depending on the work environment (indoors vs. outdoors). (Grimstad and From, 2017)

A large number of field robots are described in the literature and some are already available on the market. The goal of using robots in agriculture should not only be the automation of processes and the saving of the driver. Robotics makes it possible to break new ground in the cultivation of agricultural land through the equipment and smaller machine concepts. This development of new cultivation scenarios and driving strategies with the help of robots, as well as their evaluation, is part of the current research.

Robot localization

Mobile robots interact with their environment, which they perceive with sensors. Robot programming consequently requires the processing of sensor data. The structure and functionality of sensors have a decisive influence on the design of programs.

Figure 1: Selected sensors: f.l.t.r.: GPS receiver, step encoder/inertial measurement unit (IMU), 3D laser scanner, RGB/stereovision camera
A very important piece of information in robotics is the exact position and orientation of a robot. This can be recorded with a wide variety of sensors. A distinction can be made between absolute and relative positioning.

An absolute position is achieved in the field with the help of GPS systems. By using correction services (Real Time Kinematics – RTK) accuracies of 1-2 cm can be achieved. Likewise, 3-D information from laser scanners and stereovision cameras can be used for absolute positioning. Using this data, a map of the environment, and an appropriate algorithm, such as Adaptive Monte Carlo Localization (AMCL), a position estimate can be calculated. This methodology is primarily used for absolute indoor localization.

Supportive or relative positioning is achieved with sensors, such as step encoders, position sensors, and accelerometers. These sensors provide good short-term accuracy, are inexpensive, and allow very high sampling rates. With the information obtained, such as a robot’s travel speed or distance travelled, conclusions about the robot’s position can be drawn using mathematical modelling (e.g., Extended Kalman Filter) of the robot kinematics. Since the basic idea of a relative position measurement is to integrate motion information over time, this leads to an inevitable accumulation of errors. In particular, orientation errors lead to large lateral position errors that increase proportionally with the distance travelled by the robot. Despite these limitations, this is an important component of a robot navigation system.

Camera systems or laser scanners are used to navigate in rows of plants (e.g., in mechanical weed control) or to detect obstacles. With the help of these systems, the environment of a robot can be recorded. An evaluation of these data requires a correspondingly intelligent software, in which this image and 3D information is processed by the computer system of a robot. (Hertzberg et al., 2012)

Machine safety

The safety of people, animals, and objects, but also of the machine itself, is a central requirement in the automation of agriculture. While with industrial robots the work area and thus accessibility can be delimited, this is often not possible in the agricultural sector. This requires a higher level of safety due to a possible more direct physical interaction or a changed working environment. To avoid collisions, the workspace of a mobile robot must be monitored using cameras, laser scanners or other sensors, which places high demands on the robustness of the systems used, especially in changing environments. If there is a risk of collisions, the necessary protective measures must be initiated by the control technology and/or switched to a safe state.

Figure 2: Excerpt of relevant standards for human-robot collaboration (source: TÜV Austria)
As shown in Figure 2, standards for the safety of robots are subdivided. A-standards are basic safety standards which highlight general aspects that apply equally to all machines and systems. B-standards, in turn, set more specific requirements for individual product groups, and C-standards finally provide information on individual machine types per se. In the event of overlaps between different standards, the most detailed standard applies. Along with the increasing complexity of the tasks of robot systems, the development of new safety standards is also being driven forward. (Jacobs, 2013)

List of references:

AgroIntelli (2020). Robotti. Abgerufen am 21. December 2020, von www.agrointelli.com
Bechar, A. & Vigneault, C. 2016. Agricultural robots for field operations: Concepts and components. Biosystems Engineering, 149, 94-111.
Grimstad, L. & From, P. 2017. The Thorvald II Agricultural Robotic System. 2218-6581, 6.
Hertzberg, J., Lingemann, K., & Nüchter, A. (2012). Mobile Roboter: Eine Einführung aus Sicht der Informatik. Springer-Verlag.
Jacobs, T. 2013: Validierung der funktionalen Sicherheit bei der mobilen Manipulation mit Servicerobotern – Anwenderleitfaden. Frauenhofer- Institut für Produktionstechnik und Automatisierung IPA.
Naio Technology (2020). Dino – Vegetable robot. Abgerufen am 21. December 2020, von /www.naio-technologies.com
SPARC. 2015. Robotics 2020 Multi-Annual Roadmap—For Robotics in Europe [Online]. http://sparc-robotics.eu/wp-content/uploads/2014/05/H2020-Robotics-Multi-Annual-Roadmap-ICT-2016.pdf.
SCHWICH, S., STASEWITSCH, I., FRICKE, M. & SCHATTENBERG, J. 2019. Übersicht zur Feld-Robotik in der Landtechnik. Jahrbuch Agrartechnik 2018, Band