ROBOTS make unlikely green warriors, but they could soon be doing their bit for the environment. Trials of a Danish robot that maps the position of weeds growing among crops suggest that herbicide use could be slashed by 70 per cent if farmers used it to adopt more selective spraying techniques.
The robot drives across fields scanning the ground for any weeds and noting their positions. A later version will be able to kill the weeds too by applying a few drops of herbicide, says developer Svend Christensen from the Danish Institute of Agricultural Sciences in Tjele. But the longer-term goal is to avoid herbicides altogether by having the robot pluck the weeds out of the ground rather than poisoning them.
In tests on fields of sugar beet, Christensen and his colleagues at Aalborg University have found that selectively spraying weeds identified by robots reduces overall herbicide use by 70 per cent. Herbicide is so cheap that even reducing its cost by that amount won鈥檛 make much of an impact on farmers, but then that鈥檚 not Christensen鈥檚 goal. 鈥淭he most important benefit is to the environment,鈥� he says.
Advertisement
Weedkilling robots have already been put to work in the US, says Christensen, but they can鈥檛 be used for agricultural purposes because they don鈥檛 distinguish between plant species and tend to treat anything green as a weed. Instead, they are used to clear unwanted foliage from railways and airport runways.
The Danish weedkilling robot 鈥� a four-wheeled, battery-powered cart with high ground clearance 鈥� works by scanning the ground with a camera and recognising the shape of particular plants. It does this by harnessing software techniques from face-recognition research. 鈥淚nstead of faces, we describe the shape of weeds in terms of 15 different parameters, such as the size and symmetry of the leaves,鈥� says Christensen. The robot then uses differential GPS to log the positions of the weeds.
The research is far from complete, however, because there are 40 more indigenous broadleaved weed species in Denmark to be modelled before the system becomes practical. This isn鈥檛 a problem, says Christensen, it鈥檚 just a matter of investing the time. The bigger challenge will be to model grassy weeds, which are a more common problem for cereal crops around the world.