As part of the Dutch National Fieldlab for Precision Agriculture (NPPL) project, the Ekobot WEAI is being tested in onion cultivation, including on the farm of the Rodenburg brothers in the Flevopolder. This area is located in the middle of the country. Initial experiences are positive, although some manual labour will always be needed.
Brothers Jaco and Arie Rodenburg are looking for ways to reduce the amount of manual labour required for weed control. “We spend most of our time weeding the onions,” Jaco explains. That is why they are interested in the performance of the Ekobot WEAI, which currently is only able to weed onions. In principle, the Ekobot can also work in other crops, but specific algorithms are still under development.
The Ekobot operates autonomously across the field using GPS guidance. Each crop row is monitored by a camera that detects the plants and determines their precise location. The software analyses the images and determines whether a plant is an onion or something else. If it identifies a plant as a weed, a robotic arm is deployed to move a blade sideways through the soil to dislodge the weed from the crop row. The current version of the Ekobot operates at 0.15 metres per second — equivalent to 540 metres per hour. With a working width of 1.50 metres, this results in a capacity of 810 square metres per hour, or nearly 2 hectares per 24 hours.
Because the robot requires little time to turn at the headland, it managed to cover about 2 hectares per day on the more than 400 metre-long field at Rodenburg’s farm, even working through the night. A 2.25 metre version is already available, and supplier Homburg Holland plans to introduce 3.00 and 3.20 metre models next year.
Due to an early spring, the onions were sown early and weeds emerged early as well. In hindsight, the Ekobot should have been deployed sooner for optimal performance. By the time it arrived, some weeds had already grown too large. The Rodenburg brothers believe the tool is highly capable of removing small weeds. In theory, the Ekobot could keep a field largely weed-free. In practice, however, it is likely to be a supplement to existing weed control methods like flaming, hoeing, and harrowing. These conventional methods remove weeds at high capacity, while the Ekobot can deal with any remaining ones.
Some manual weeding will always be necessary. As the crop grows, there will come a point where the onions become too large for the Ekobot to manoeuvre effectively, even though weeds are still present. Perennial weeds and weeds growing too close to the onions will also pose a challenge. Part of the solution lies in the calibration and software configuration.
Besides saving labour, the Rodenburgs see additional advantages in the Ekobot. Unlike a conventional hoe or harrow, it minimally disturbs the soil, reducing the chance of creating a seedbed for new weeds. With the Ekobot available as a follow-up tool, hoes can be set less aggressively. Working farther from the onions reduces the risk of damaging onion roots, which helps prevent fusarium infections. Completely abandoning hoeing is not the goal, but if the soil surface is compacted, hoeing will still be necessary to loosen the soil.
The Ekobot’s weeding blades may lack the force to penetrate hard soil effectively. Several machine settings are adjustable: the depth of the blade can be configured digitally, as can the degree of crop tolerance and the sideward stroke of the blade. How far must a weed be from the crop (in this case, onions) before the robot targets it? If that margin is wide, no onions will be harmed. If it is narrow, some onions will inevitably be lost. The user needs to find the right balance.
The Ekobot is relatively lightweight but robustly built, notes Rodenburg, which inspires confidence. It is equipped with two 100Ah batteries, which provide enough energy for 8 to 9 hours of continuous operation. The software knows the length of the field and calculates before each pass whether the batteries hold enough charge to complete the full row — preventing the robot from stopping halfway due to an empty battery. The batteries are easy to handle manually and can be swapped quickly.
“I can monitor the machine via my phone, including the battery charge level,” Jaco explains. This allows him to check when it is time to go to the field. The robot is equipped with a safety bumper on the front and sides. If it touches an obstacle, the robot stops and a message is sent to the phone. Moving the Ekobot to another field is easy: it has a three-point linkage that allows it to be lifted and transported by a tractor.
The main conclusion so far is that the Ekobot concept works. There is still room for improvement. One key point is the ability to weed multiple crops — which is only a matter of time. The software counts both crop and weed plants. The weed density per square metre could eventually be used to adjust the driving speed. At present, the speed is fixed, but it could be slowed down in heavily infested patches and sped up in cleaner areas. At the moment, users cannot adjust all settings themselves, such as the blade depth. Eventually, users should be able to change these settings via a user-friendly interface — and that is on the way. Trials are also being carried out with different blade designs depending on soil type or weed species. As experience with the Ekobot grows, so too will user requirements, but the foundation is solid.
Subscribe to our newsletter to stay updated about all the need-to-know content in the agricultural sector, two times a week.