“Target hoeing” –

Using robots to combat weeds in a targeted manner

Author: Georg Supper

Challenges in field robotics

A large number of agricultural robotic developments, some of which are already available on the market, are already described in scientific literature. In most cases, these robots are adapted to a specific task by their equipment. On the other hand, multipurpose robots are capable of performing a variety of different tasks. These developments range from robot use in harvesting fruits, sowing, to mechanical weed control.
Robotics itself provides the opportunity to support or even replace humans, especially in time-consuming manual tasks such as weed control in vegetable or organic farming. With the help of robots, their equipment with sensors and actuators, and the corresponding artificial intelligence, processes can be rethought, and new innovative workflows can be developed. One approach is targeted mechanical weed control using the hoeing technique (target hoeing). In contrast to broadcast cultivation, “target hoeing” uses movable hoeing coulters to cultivate only the areas with weed plants. Possible energy savings by “target hoeing” and new driving strategies for robotic weed control are investigated within the DiLaAg project 1.

Field robot “Mathilda”

The robot platform “Mathilda” (see Fig. 1) was developed at the Institute of Agricultural Engineering as part of a master’s thesis and is intended to serve as a carrier vehicle for a wide range of applications. Its dimensions and parameters were specifically chosen to allow easy transport between different application areas, indoor and outdoor use, and simple and flexible expansion, both in terms of hardware and software.
It has a width of 850 mm, a length of 1197 mm, and a height of 620 mm at the upper edge of the platform and 1020 mm at the upper edge of the control box. The weight of the platform is 255 kg. It is driven by 2 electric motors, which are used to realize the steering by changing their speed. The front drive wheels have a diameter of 500 mm. The rear wheels are freely rotatable and have a diameter of 270 mm. Sensors for navigation are a LIDAR scanner, DGPS system, Inertial Measurement Unit (IMU) and incremental encoders on the drive wheels.
The robot software is based on the Robot Operating System (ROS) framework. For the operation of the robot a software concept is developed, which is based on the ROS Navigation Stack. This concept is complemented by own developments in the area of robot kinematics and localization, which adapt the software to the specific characteristics of the developed robot platform.

Figure 1: Field robot “Mathilda” exiting the machine test station

Development of a “Target hoe”

The described robot platform “Mathilda” is used as a carrier vehicle for target hoeing. For this purpose, a vertical and horizontal shifting frame is developed and attached to the rear end of the robot (see Fig. 2). This is used to raise and lower the hoe, as well as to horizontally align the hoe with the row of plants. The displacement is performed by linear motors LA36 from the company Linak. Force sensors are mounted on the frame to measure the draft force and lifting force. With this shifting frame, various devices can be attached to the robot and experiments can be performed with them. Furthermore, useful data can be collected and integrated with the attached sensor technology.
For mechanical weed control, single hoeing elements are attached to the shifting frame. Each of these hoeing elements can be moved by 80 mm by means of a pneumatic cylinder and thus be moved into the ground if required.

Figure 2: Illustration of the robot, the shifting frame, and the movable hoeing element

Interdisciplinary objectives and collaboration in the DiLaAg

The objective of this project is to perform a side-specific mechanical weed control in a row crop using the field robot “Mathilda” and the developed hoe. For this, detection and in particular localization of weed plants is a necessary precursor. Here, the weed plants are detected by a stereovision system and recognized by deep learning methods (see DiLaAg project 2). The position of the weed plant is detected by the computer and triggers a movement of the relevant hoeing element and cuts the weed plant by means of goosefoot share in the soil (see Fig. 3). A crop management approach (see DiLaAg Project 3), especially the investigation of innovative technologies to optimize agricultural production systems economically and ecologically, should enable responsible and sustainable agricultural production.

Figure 3: Schematic representation of “target hoeing”.

The existing measurement technology on the robot and the hoe are used on the one hand for process control and on the other hand to record parameters such as energy consumption, draft force, and work quality. These agricultural process parameters are recorded by test runs in the field and are used to analyse side-specific driving strategies as a function of weed density, cultivation speed and energy consumption. With the help of Life Cycle Assessment (LCA) (see DiLaAg project 8) and the data obtained, mechanical-robotic weed control can be examined for its environmental impact. This overall scientific view of the project and the investigation in practical trials can provide important insights for field robotics and significantly influence further development in this field.

Cite this post as:
G. Supper, “Target hoeing” – Using robots to combat weeds in a targeted manner In: DiLaAg Innovationsplattform [Webblog]. Online-Publikation: “” 2020.


Coleman, G. R., Stead, A., Rigter, M. P., Xu, Z., Johnson, D., Brooker, G. M., … & Walsh, M. J. (2019). Using energy requirements to compare the suitability of alternative methods for broadcast and site-specific weed control. Weed Technology, 33(4), 633-650.

Duckett, T., Pearson, S., Blackmore, S., Grieve, B., Chen, W. H., Cielniak, G., … & Yang, G. Z. (2018). Agricultural robotics: the future of robotic agriculture. arXiv preprint arXiv:1806.06762.
LINAK (2021). Linearantrieb LA36.

Supper, G., Aschauer, C., & Barta, N. (2019). Planung und Entwicklung einer mobilen, autonomen Roboterplattform für pflanzenbauliche Anwendungen. 39. GIL-Jahrestagung, Digitalisierung für landwirtschaftliche Betriebe in kleinstrukturierten Regionen-ein Widerspruch in sich?.