Easy pickings? How robot farm hands could revolutionise agriculture

Sergio Martinez has a track record of creating robots to perform unglamorous tasks. “Twenty years ago in Madrid I had an idea to send robots into sewage tunnels to look for fatbergs,” he says. “I built a few prototypes that I was able to drive, with an Xbox-type controller and a headset so I could see what they were seeing. It was pretty cool. But the technology was not quite there yet, so sometimes the connection broke down.”

Having abandoned Spanish fatbergs and moved to the UK, Martinez is now head of engineering at Earth Rover, an ambitious tech company dedicated to robotising agriculture. Over the last year, Martinez and his colleagues have been programming a modified version of the European Space Agency’s ExoMars Rover to tackle their first task: the automation of broccoli farming. The aim is that a scaled-up model will one day be able to zoom autonomously up and down fields of broccoli day and night, monitoring the crop, tending it, protecting it from weeds and then, when the time is right, harvesting it, while the farmer reviews the whole process on their smartphone with their feet up.

The first step will be to gain an unprecedented level of information about a farmer’s fields as each individual plant is assessed by the robot’s sensors. “It will monitor the size, shape and location of your crops,” says Martinez’s colleague Paul Harter, Earth Rover’s chief technology officer. “It will look at them in 3D and get information about their structure. It could use hyper-spectral imagery – which captures pictures at very specific wavelengths of light – to find out things like the nitrogen content of the leaves, which tells you about the plant’s growth stage and health. Then it could do chemical testing on the soil around the plant and even search for micro-organisms. So if you were worried about certain kinds of mildew or certain diseases, you could spot those earlier.”

There are times when Earth Rover’s sophisticated research veers delightfully into Robot Wars territory. “There are a series of options in the frame for weeding attachments,” says Harter. “There are lasers, flames, mechanical cutters and things like mini hoes which would be dragged through the soil… High-powered lasers look promising but there are quite a lot of technical and regulatory things standing in the way of using them. I think realistically a workable version is quite a long way off.”

A field filled with laser-toting robots doing 24/7 battle with the enemies of broccoli may sound like a particularly dystopian episode of The Archers – but its practical impact would be immense. At the moment, farmers spray their entire fields with herbicides to get rid of the weeds which could threaten their crop. If you could take the unwanted leafy intruders out one by one – whether by laser, flame or even individual ‘micro-sprayed’ drops of herbicide – you could prevent a huge amount of harmful materials from going into the soil and ultimately into the water system. Equally, when it came to pests, understanding patterns of infestation on a plant-by-plant basis would enable tailored, targeted responses instead of all-out chemical attacks.

There are 65,000 seasonal migrant workers picking crops in the UK and farmers are looking at robots to help fill the gaps”

But bespoke weeding and bug-zapping are not the only ways robots can increase agricultural efficiency. At the moment, crops are planted in rows separated by access corridors of soil: cut the humans from the equation and your computerised farm hands could fit in far more plants. Harvesting of vegetables, meanwhile, currently takes place all in one go – or at most in a few passes – and much of the produce has to be thrown away because it’s over- or under-ripened. Robots, in contrast, could follow individualised plant-growth maps to selectively harvest one vegetable at a time.

“You’d have an ID for every broccoli plant in the field, and know exactly where it is to within a couple of centimetres,” says Harter. “You’d be able to pull up the growth chart for any particular plant out of the millions you’ve grown. And in theory a robot harvester would be able to go through and pick only the ones that are ready to harvest. You’d be farming on a huge scale but with the same care and attention as if you were working on an allotment.”

 

Transfarmers vs Robo crop

The Earth Rover team is not alone in sensing the opportunities of automated agriculture. Across the world, programmers are teaming up with farmers to figure out how to replace manual labour with robotic serfs.

Abundant Robotics of California is using self-driving vehicles equipped with elaborate vacuum devices to identify ripe apples and suck them out of the tree. A consortium of groups from the Netherlands, Belgium, Sweden and Israel have created Sweeper, a sweet-pepper-harvesting robot which scoots along scanning fruit for ripeness before snipping those that are ready from the stalk. A series of US and Australian companies are working on automated drones that specialise in everything from crop-spraying to tracking rabbit damage. And the University of Plymouth has developed the GummiArm, a machine based on a human arm which can assess and harvest cauliflowers.

At the University of Lincoln, meanwhile, they’ve made an automated strawberry picker as part of their RASBerry (Robotics and Automation Systems for berry production) programme in conjunction with a Norwegian robotics team. Soft fruits are notoriously tricky to harvest by robot as you need to apply exactly the right pressure to pull them free without squishing them into jam – an ability we human fruit-pluckers happily take for granted.

Robots also get easily befuddled when picking fruit. “You can get some serious problems of visual occlusions,” says Professor Simon Pearson, the director of the University of Lincoln’s Institute for Agri-Food Technology (LIAT). “The fruit might be hidden behind a leaf, and if the robot can’t see the fruit, it can’t pick it. That sort of natural variability is probably the biggest challenge [in agricultural robotics].”

When robotics fails, however, selective breeding could step in. “You can use biology to deal with the occlusion problems,” says Pearson. “Conventional crop breeding will be a key enabler: with strawberries you can breed varieties where the fruit is not buried within the leaves or clumped together but is easily identifiable. I’ve no doubt that in the future we’re going to be breeding plants for traits which make them ready for robots.”

 

Down on the farm

The new race to robotise agriculture is partly being propelled by the escalating sophistication and plummeting cost of technology, including AI. But in the UK, says Pearson, there’s also an extra factor at work. “Brexit and the uncertainty of access to seasonal migrant labourers is a very big driver,” he says. “It’s driving lots of robotic applications for all those crops which need many people to harvest them. There are 65,000 seasonal migrant workers picking crops in the UK and farmers are looking at robots to help fill the gaps.”

There is evidence that the UK’s referendum vote to leave the EU is already having an effect on recruitment: at its meeting on 24th April 2018, the National Farmers’ Union reported a 29 percent shortfall in farm labour in the UK in 2017, with 60 percent of growers reporting shortages. “There are no horticultural businesses without people,” said NFU horticulture and potatoes board chairwoman Ali Capper. “If we can’t pick it, we aren’t going to grow it.” The union has pointed the finger of blame at the government’s refusal to introduce a Seasonal Agricultural Workers (SAWS) visa programme to help make up the shortfall, saying that a lack of workers could drive many farms out of business.

Investors see potential in a world of ripe fruit with only unenthusiastic, out-of-shape urbanites to pick it”

Recruitment has been further hampered by the fall of the pound against the euro since the referendum vote, which has lessened the appeal for eastern European students looking to do a season’s picking and send the money back home. “No one wants to do these jobs,” says Martinez. “The salary is not good enough and Brexit is like the country saying ‘no’ to them, so 
why bother?”

But the problems with agricultural labour go way further than Brexit. “There’s a fast burn, which is Brexit and then a slower burn, which is demographics,” says Pearson. “And of course that’s not limited to the United Kingdom. That’s in Japan and China, in the US and the EU. It’s everywhere: ageing populations, moving from rural areas to urban areas, with fewer and fewer people wanting to work in farming. And of course you’ve also got to be relatively fit to do some of these jobs.”

Money is pouring in to agricultural robotics companies from big landowners and investors who can see potential profits in a world of ripe fruit with only doddering, unenthusiastic, out-of-shape urbanites available to pick it. A report published on the Agricultural Robots Market in January 2018 predicted that the sector would grow from US$6.374 billion in 2017 to $17.614 billion by 2023, an increase of 176 percent.

Farms are big and spread-out and muddy. It’s a hard environment to run technology in”

Agricultural robotics firms in the UK face a problem that farmers’ expectations are out of kilter with robotic reality. “The biggest issue is validating expectations,” says Pearson. “If you’re a farmer and you’ve got fruit and no pickers, you want a solution now, not next week, next month or next year. But realistically it’ll take three to five years before a lot of this technology has an impact on farms.”

One of the biggest challenges for agricultural roboticists trying to get their products into the field is the context in which their creations have to work. “Farms are big and spread-out and muddy,” says Harter. “It’s a hard environment to run technology in.” Many technologists are factoring in smaller, more achievable goals on the pathway to full automation. While perfecting their strawberry-picking robot, for example, the RASberry team has also created a “co-bot” which can ferry boxes of fruit from human pickers to faraway collection points, saving time and labour costs.

 

Fields of dreams

The automation of agriculture is still at a fledgling stage, but its potential 
is extraordinary. “Looking at where things are going, I don’t think we’re going to have so many fields,” says Martinez. “I think we’ll have warehouses, inside the cities.”

He cites a blue-skies project which would see supermarkets instal robotised greenhouses in the space above shoppers’ heads. “If you take all the heat that the building is generating and you direct the CO2 into a fully automated greenhouse that’s inside the 
building, you don’t need a tractor or a lorry to transport the food,” says Martinez. “A customer would just 
push the button to say they want a lettuce and the robot picker would check which one is ready and give it 
to them. And if nobody’s buying broccoli this week, 
the greenhouse just brings the temperature down a bit so the broccoli crop lasts for another week 
before ripening.”

The robotisation process could create a lot of new opportunities for humans. “The problems are all solvable,” says Pearson. “It just takes time and human expertise – we’ll need more and more skilled people to build these robots and then to service them. We need to be recruiting more robot engineers, computer scientists and electronic engineers. People are going to have to retrain. We are going to have a much more technologically advanced industry.” At the moment, says Martinez, there’s a real problem of supply. “The UK doesn’t have enough roboticists,” he says. “Everyone is fighting for the same ones.”

Eventually you could use your mobile phone as the brain of the robot… 
Then you could get your fields to work by themselves”

Martinez believes that the future benefits of 
robotic farming could be felt most keenly in poorer nations. “What happens now is that every few years, westerners go to developing countries and they say, 
‘We are going to help you with your agriculture,’” he says. “Then they bring tractors and other technology and 
they just leave them there. A year later, those things 
start falling apart. It’s impossible to get replacement parts, so people go back to farming in the traditional, manual way. But if you could manage to build an agricultural robot that you could 3D print parts for, then when it broke you could easily fix it. And if electronics keep becoming cheaper and faster, eventually you could use your mobile phone as the brain of the robot, so you could build a basic model for very little money. 
Then you could get your fields to work by themselves.”

On 5th September 2018, the British government announced a two-year SAWS pilot that would grant visas for agricultural work to 2,500 non-EU workers. Spokespeople for the fruit-farming industry responded with frustration, saying that a minimum of 10,000 visas would be needed immediately just to deal with current shortages. As the pool of people prepared to do backbreaking, repetitive seasonal work in remote parts of the countryside shrinks, you should keep your eyes peeled. It may not be long before Earth Rover and its automated brethren – with their assorted battery of snippers, lasers and miniature hoes – clank out to make their debut in a field near you.

 

Posted on Friday, October 12th, 2018 - Categories: #31 - Apr-Jun 2018, Delayed Gratification magazine, From the archive, Long reads.