Use of robots for efficient weeding

A project supported by the Federal Office for Agriculture (FOAG) in partnership with the University of Applied Sciences for Agriculture, Forestry and Food (HAFL).

When we defined the planning for the year 2020 with numerous agronomic tests, we expected our AVO robot to encounter certain obstacles but also some great victories. The HAFL project was one of them because in the process of the overall operation of the robot, we were able to validate certain functionalities and obtain an extremely encouraging study result.

Deciphering an important step in the development of our technology

Last spring, we were fortunate enough to conduct a pilot project with our
HAFL partner. In the predefined objectives for this test we find

  • Comparison of AVO technology to conventional weed control
  • Evaluation of the effectiveness of the plant recognition system
  • Examine the accuracy of the spraying device


  • Location: Aarberg (BE)
  • Date: June 2020
  • Variety of beet: Smart Belamia

For these trials, 3 treatment modalities were considered:

  • M1 – Full herbicide treatment (traditional application)
  • M2 – Herbicide band treatment coupled with AVO passages
  • M3 – Herbicide treatment carried out only with AVO

In order to evaluate the effectiveness of the different approaches, surveys of the weedy flora
have been regularly realized by our partner.

Study results

The results presented in the table above demonstrate a good
overall performance in terms of stability and detection rate of the detection device.
spraying. The values obtained allow a validation of the
algorithmics and mechatronics in field conditions!
For all the tests carried out, the modalities corresponding to a use
complementary (M2) or total of our technology (M3) have demonstrated results
comparable to a traditional application (M1). These results put
also evidenced substantial savings in herbicides.


Work continues for our engineers to optimize the robot’s feed speed.
as well as the efficiency of spraying. One of the major challenges is to obtain an
regularity of the robot’s performance in all types of terrain and independently
weather conditions.

To date, algorithmic and mechatronic validations allow us to use
our technology behind a tractor travelling at 8km/h.