The DockWeeder robot enables organic dairy farming by controlling grassland weeds (DockWeeder)

• Unité de recherche: OFAG
• Numéro de projet: 10.20_6
• Titre du projet anglais: The DockWeeder robot enables organic dairy farming by controlling grassland weeds (DockWeeder)

Mots-clé (Allemand)

Stumpfblättrige Ampfer, Robotik, biologische Landwirtschaft, Milchwirtschaft, Grünland, Unkraut, Rumex obtusifolius L.

Mots-clé (Anglais)

broad-leaved dock, robotics, organic farming, dairy farming, grassland, weeds, Rumex obtusifolius L.

Mots-clé (Français)

Patience à feuilles obtuses, robotique, agriculture biologique, élevage laitier, prairie, mauvaises herbes, Rumex obtusifolius L.

Description succincte (Anglais)

Broad-leaved dock (Rumex obtusifolius L.) is a common and troublesome weed with a wide geographic distribution. The weed is readily consumed by livestock but its nutritive value is less than that of grass. The high contents of oxalic acid and oxalates can affect animal health if consumed in larger doses. When left uncontrolled, the weed will reach a high density and reduce grass yield by 10 to 40%. In conventional dairy farming, the weed is normally controlled by using herbicides. In organic farming no synthetic pesticides are used and there is a risk that broad-leaved dock will spread. This is also true in ecologically intensive dairy farming, where one of the goals is to maintain multispecies pastures where use of herbicides would affect desirable species such as clovers and vetch. As an illustration, on 17 organic dairy farms surveyed in The Netherlands, 51% of fields were infested at more than 1,000 plants/hectare. Similarly, of 108 organic farmers surveyed in Germany, 85% indicated having problems with broad-leaved dock. Thus, broadleaved dock may turn out to be a serious obstacle to achieve the European goal of increasing the share of organic farming.

Objectifs du projet (Anglais)

The solution proposed here consists of creating a robot that is capable of exploring a pasture by relying on GPS, equipping it with an array of sensors to detect the weed, and also equipping it with a non-chemical method to eliminate detected weeds. In earlier work, we demonstrated with an experimental robot that under certain conditions adequate weed detection and control is possible. Importantly, we found that the weed population remained low for three years after control. This earlier work had three major shortcomings: the mechanical construction of the autonomous platform was insufficiently robust, the weed detection worked only under a limited set of environmental conditions, and the weed control method was prone to mechanical breakdown on stony ground.

In this project, we address all three shortcomings. We use an existing, robust autonomous platform, we advance the state of the art of weed detection by combining two-dimensional (2-D) and three-dimensional (3-D) imaging, and we adopt and optimize an innovative, environment-friendly hot-water treatment to eliminate weeds. Finally, we will store the geolocation of each detected weed so that in subsequent years the robot can move efficiently to those areas where weeds have occurred in the past. In summary, by combining the expertise of the consortium partners, we will be able to build a robot to detect and control broad-leaved dock which has immediate commercial potential.

Mise en oeuvre et application (Anglais)

The work is organized in 7 work packages.

WP1 (Project management; lead: DLO) comprises overall management as well as development of a business plan for commercial exploitation of project results after the project has ended.

In WP2 (User requirements; lead: AAU) we work with potential users to develop boundary conditions for the remainder of the project.

Of the R&D work packages, WP4 (Weed detection; lead: ZHAW) requires the largest effort: robust detection of weeds under widely different conditions still poses a major scientific and engineering challenge, which is why three academic partners will devote significant resources to this WP.

On the other hand, WP3 (Autonomous vehicle; lead: Pilgrim) requires a relatively small effort because we will adapt an existing robot platform and existing software architecture; likewise, optimizing the hot water treatment method and fitting it on the robot in WP5 (Weed control; lead: AG) requires a relatively modest effort.

The different components of the system are brought together in WP6 (System integration; lead: Pilgrim). Finally, in WP7 (Knowledge transfer & extension; lead: Terrena) we disseminate the results of the project.

The work is organized in time as follows. When user requirements have been made clear (month 3), WP3-4-5 proceed to a large extent in parallel. At month 10, work on the autonomous vehicle and on the weed control method has progressed to the point where the two can be integrated. At month 14, the weed detection system is specified and can be integrated on the robot (work on the weed detection algorithms will continue after this point in time). At month 18, the work will focus entirely on integrating all components (navigation, detection, control) in one functioning system.