Weeding robot that uses computer vision and artificial intelligence
Promising field: California-based FarmWise has developed a weeding robot that uses computer vision and AI

Their machines look more like miniature flying toys than autonomous farming tools, but researchers at Harvard’s Wyss Institute believe their RoboBees — which can achieve vertical take-off, hover and steer — could eventually perform tasks such as crop pollination and environmental monitoring. 

So far, these RoboBees have yet to operate outside the lab, and commercialisation remains far off. But technological advances combined with labour shortages do mean that more robots are becoming economically viable for use on the farm. 

“The low-hanging fruit has been using them in polytunnels and glasshouses, where they don’t get mud and rain on them,” says Belinda Clarke, director of Agri-TechE, a membership organisation that supports innovation in agricultural technologies. “But, now, they’re moving out into the field.”

Their increasing capabilities are making possible a more tailored, plant-by-plant approach to cultivation that can minimise inputs such as water and agrichemicals. For example, California-based FarmWise has developed a weeding robot that uses computer vision and artificial intelligence to distinguish weeds from crops, cutting labour costs and enabling farmers to use less herbicide.

Until recently, agriculture had been slow to adopt robots. High levels of capital investment and limited seasonal deployment made it hard for the machines to generate value. “It has not been a cracking investment theme until now,” says Adam Anders, managing partner at Anterra Capital, a food and agriculture venture capital firm.

But that is changing, he says. In mature economies, restrictions on migrant labour, ageing agricultural workers, and a lack of enthusiasm for back-breaking work that is dull and dirty mean farm workers cost more and are harder to find. In the US, for example, farm workers’ inflation-adjusted hourly wage rose by 28 per cent between 2000 and 2022, compared with 17 per cent for non-farm workers, according to the US Department of Agriculture.

Because technologies such as computer vision and machine learning enable robots to identify and respond appropriately to objects, some machines can now weed or pick fruit and vegetables with a speed, accuracy and reliability that was once achievable only by humans. 

Moreover, “soft robotics”, using rubber cups or small bean bags, can gently clasp and remove delicate, high-value produce such as peaches and raspberries from plants without damaging them. 

Tjarko Leifer, FarmWise’s chief executive, anticipates a new wave of robot-driven agricultural automation. “Computers can now see, and have the dexterity to replace some of these human jobs,” he says.

There may be sustainability pay-offs, too. Drones and robots in the field can use smart sensors and computer vision to collect and transmit data and images in real time — on everything from local weather and soil conditions to plant growth rates.

FarmWise chief executive Tjarko Leifer
FarmWise boss Tjarko Leifer says robotics can tailor nutrient application ‘on a plant-by-plant basis’

Applying machine learning to such large volumes of data can also generate new insights into how and where to cultivate crops. And, by tracking how plants in different parts of a field respond to inputs such as water and chemicals, robots can help farmers minimise use of these resources.

This takes precision farming to another level, says Anders: “Instead of blanket applying fertiliser or pesticide, you can apply less and be more precise.” 

Another benefit of the next generation of robots is that they are smaller than traditional machinery, such as combine harvesters and tractors. “We want these little beasts to be light and agile,” explains Clarke. “Something run by a rechargeable battery could be powered by solar, which opens up a new opportunity to use robots in a sustainable way.”

Being lighter means that the machines compact the soil less when moving over it. Compaction makes it harder for plants to access nutrients and for soil to hold water. “You want the soil to act like a sponge so that the water is there when the crop needs it,” says Clarke.

But, while they are becoming more economically viable in mature markets, few farmers in developing countries are yet able to afford robotics. “These things will have to be largely commercialised in developed markets,” says Anders.

Another hurdle to widespread adoption of farm robots is regulation. “If you’re talking about things that operate on a fully autonomous basis on a farm or in the sky, out of eyesight, there’s still uncertainties,” notes Anders. “In the US, a drone can’t operate out of eyesight, which more or less defeats the purpose in terms of getting to scale.” 

However, he thinks this is likely to change, as governments seek to increase food security and the sustainability of farming, and also as autonomous machines become more intelligent, and people adjust to their presence.

Leifer believes that the success of robotics in advancing precision farming will accelerate its adoption. “We can apply the right nutrients and the right crop protection chemistry to every plant in a field and tailor that on a plant-by-plant basis,” he says. “And through that, we benefit the farmer, the consumer and the environment.”

Copyright The Financial Times Limited 2024. All rights reserved.
Reuse this content (opens in new window) CommentsJump to comments section

Follow the topics in this article