Home / Blog / Robotics and AI for weeding

Robotics and AI for weeding

This is an autonomous robot that is able to distinguish weeds from crop and remove 100,000 weeds per hour, zapping them with a laser beam. Image credit: CarbonRobotics

Possibly the biggest impact of the Industrial Revolution has been on … agriculture. By using machines for farming it made possible to increase productivity 100 folds (both in terms of yield per acre and much more in terms of human work to till, farm, harvest the field). The employment in farming dropped from 70-80% of the XVIII century to less than 2 % today in industrialised Countries (it remains over 50% in poor Countries. According to the latest figure of the World Bank employment in agriculture worldwide dropped from 44% in 1991 to 27% in 2019).

This automation trend continues with the deployment of autonomous robots in agriculture in all phases, including fruit harvesting. This requires not just autonomous robots, it requires smarter and smarter robots able to distinguish a fruit ready to be harvested from one that needs to stay on the plant for few more days or one that can distinguish a weed from the crop.

Weeding has always been a labour intensive activity, so much labour intensive that the approach to weeds has been to find chemicals that can selectively kill the weeds sparing the crop. This approach, still widely used, has the adverse effect in cost (herbicides add to the cost of the harvest) and in pollution: in theory herbicides should not the harmful to us, since some crop contamination is always possible but in practice I would not suggest to ingest a spoonful of herbicide… In addition, even if herbicides would be safe for us they may not be safe for other species. Bees, as an example, have found to be affected by some herbicide.

This is where CarbonRobotics comes in. They have developed an autonomous robot that is able to distinguish weeds from crop, even at the early stage when seeds are sprouting (watch the clip) and seedlings may look very similar. Using artificial intelligence the robots moves over the field “looking” at the soil and when it detects a weed it kills it by zapping it with a laser beam. In this way it can kiil up to 100,000 weeds per hour. The use of laser provides for a very selective action (manual weeding usually uproot the weed and by doing that it can affect the nearby crop). An added bonus is that the laser kills the weed through thermal exposure and this turns the weed into a fertiliser.

The robot is not designed for small patches of land: it cost a few hundred thousands dollars but for large farms it repays itself in 2 to 3 years, by cutting herbicide and labour cost. A single robot takes care of some 20 acres per day and it makes economic sense for farms with arable land between 200 and tens of thousands of acres.

It moves on the field using GPS (with a precision of 15cm -6 inches-) and LIDAR to avoid obstacles. The image recognition uses 12 high resolution cameras whose video streams is processed by 8x NVIDIA 30-Series GPUs (their processing power is huge and it is amazing to see it applied for weeding!).

About Roberto Saracco

Roberto Saracco fell in love with technology and its implications long time ago. His background is in math and computer science. Until April 2017 he led the EIT Digital Italian Node and then was head of the Industrial Doctoral School of EIT Digital up to September 2018. Previously, up to December 2011 he was the Director of the Telecom Italia Future Centre in Venice, looking at the interplay of technology evolution, economics and society. At the turn of the century he led a World Bank-Infodev project to stimulate entrepreneurship in Latin America. He is a senior member of IEEE where he leads the New Initiative Committee and co-chairs the Digital Reality Initiative. He is a member of the IEEE in 2050 Ad Hoc Committee. He teaches a Master course on Technology Forecasting and Market impact at the University of Trento. He has published over 100 papers in journals and magazines and 14 books.