The world’s human population currently stands at around 7.6 billion and is projected to reach 11.2 billion by 2100. We will therefore need a food production and distribution system that can accommodate another 3.6 billion people–ideally while consuming as little additional land and leaving as small an environmental footprint as possible, in order to maintain vital ecosystem services and conserve Earth’s remaining wildlife.
That’s clearly a challenge given that around half of the world’s habitable land is under agriculture of some kind–with a high proportion of this used for livestock farming (Figure A).
In a widely reported recent study, Poore and Nemecek (2018) note that a shift away from meat and dairy consumption would go a long way towards relieving pressure on agricultural land and reducing environmental impact: “Meat, aquaculture, eggs, and dairy use ~83% of the world’s farmland and contribute 56 to 58% of food’s different emissions, despite providing only 37% of our protein and 18% of our calories.”
Moving to a diet that excludes animal products, say the study’s authors, could reclaim 3.1 billion hectares of global farmland (a 76% reduction), while reducing food’s greenhouse gas emissions by 6.6 billion metric tons of CO2eq (a 49% reduction), among other environmental benefits.
Of course, it will take time to effect a major shift in dietary preferences–primarily in developed countries–and global land use patterns, although emerging technologies like lab-grown meat may have an increasingly important role to play here.
On the crops side, big advances in production have been made in recent decades, and modern technology is poised to deliver even more.
Farming output can increase in two basic ways: By increasing the yield per unit area (intensification), or by expanding the area under cultivation (extensification). Increased cereal production has largely been achieved by intensification over the last 50 years (Figure B). Only 16% more land was used for cereals in 2014 than in 1961, for example, while global cereal production increased by 280%. During the same period, the world’s population increased 136%, which means that cereal production per person has increased even as the population has more than doubled.
These increases were largely delivered by the post-WW2 Green Revolution–a suite of technologies and farming practices involving high-yielding crop varieties, agro-chemicals (fertilisers, herbicides, and pesticides), irrigation, and mechanisation. Industrial-scale agriculture, often using genetically modified (GM) crops, has undoubtedly delivered many benefits, but there are costs too. These include high levels of inputs (which can become pollutants if inefficiently applied), the development of resistance to pesticides and herbicides, and the use of large, expensive, and environmentally damaging farm machinery.
These and other issues have sparked interest in sustainable intensification, where the goal is to increase production from existing farmland while minimising environmental damage, thereby maintaining the land’s capacity to continue producing food, and also helping to preserve biodiversity.
What is precision agriculture?
Precision agriculture, also known as “smart farming” or “precision farming,” is a key component of sustainable intensification. This combines remote sensing, IoT devices, robotics, big data analytics, artificial intelligence, and other emerging technologies into an integrated high-resolution crop production system.
SEE: The future of food (ZDNet/TechRepublic special feature)
One of the biggest drawbacks of industrial-scale farming is the use of large, heavy machinery such as tractors, sprayers, and harvesters, which compact the soil and compromise a crop plant’s ability to develop a healthy root system. Soil compaction is an important factor–perhaps the important factor–in the slowing of crop yield increases that has been observed in recent decades–here, for example in the UK (Figure C).
Another drawback of industrial-scale farming is its low resolution. For example, when agrochemicals are applied to fields via sprayers many metres wide, much of it misses the target: Not only is this wasteful, but it can also create environmental pollution, harming beneficial organisms and compromising ecosystem services.
These and other problems can be addressed by replacing big human-operated machines with multiple small autonomous devices, backed up by modern IT infrastructure. The primary goal is to “unlock” flatlining crop yield curves, as shown in the above graph.
Precision farming was the subject of a 2016 report from Goldman Sachs, subtitled Cheating Malthus with Digital Agriculture. According to the investment bank, a technology-driven crop yield improvement of 70% is achievable by 2050 via a combination of precision planting, fertiliser application, irrigation spraying, and autonomous driving applications, with adoption starting in developed markets. This translates (under certain assumptions) to a total addressable market of $240 billion by 2050, with the major components featured in Figure D.
Clearly there’s a lot to play for in precision farming–not only in terms of helping to feed the world more sustainably, but also in terms of cold, hard business opportunities. For evidence, look no further than the September 2017 acquisition of machine learning specialist Blue River Technology by agricultural equipment giant John Deere for $305 million.
In the remainder of this article, we’ll examine a couple of UK-based startups in the vanguard of this new agricultural revolution.
Precision agriculture in the UK
“Small Robot Company is inspired by and based on the work of Professor Simon Blackmore at Harper Adams University. He has been working on the idea of small robots rather than big tractors for the last fifteen years or so,” SRC’s co-founder Ben Scott-Robinson told ZDNet.
An entrepreneur with particular expertise in user experience design and AI, Scott-Robinson stressed the emerging status of precision agriculture in the UK: “I heard him [Blackmore] on the radio and was inspired by what he was saying, so I contacted him and asked ‘who are the hot startups that I could join in the small farming robot world?’ He said there weren’t any, but that I should meet this chap called Sam, who’s very interested and brings a different perspective on it.”
“Sam” is SRC’s other co-founder Sam Watson Jones, a farmer and ex-management consultant, who took up the story: “The agriculture business is highly consolidated, both before and after the farm, and there’s really only a handful of companies that supply machinery and equipment into farms. This new technology is highly disruptive to what they’re doing, and potentially cannibalises their business. Also farmers are not really asking for this technology yet–although we’ve spoken to a lot of them over the last 12 months and have 150 on-board. But they understand that their business model is broken in most respects, primarily because their revenues are fixed: On our farm in Shropshire we generate the same revenues we did 25 years ago from our crops, but the costs to achieve those revenues are significantly higher.”
“Farmers don’t have a clear vision of how to get themselves out of this situation,” Watson Jones continued, “which is why, if you follow the farming press, you’ll see so much talk about subsidy, government support, and supermarket bashing. What brought Ben and I together was that we shared a positive vision for technology actually enabling farms to be run in a different way, under a different business model.”
“When we first started looking at this, I got very excited by the technology,” said Scott-Robinson. “But we wanted to know what farmers think about it–because as everybody knows, they don’t adopt new technology. So Sam spent the best part of eight months interviewing farmers–the ‘Aga sessions’ as we call it–to find out what their pain points were, and explore the idea of whether robots would be something they’d be interested in.”
“What emerged is that farmers aren’t scared of technology per se–although some are quite reactionary; what they’re really scared of is being sold something that will break, and then sit in their shed gathering dust. So one of the first things we did was step away from the idea of cool robots and ask what model they’re going to buy into. That’s how we came up with the idea of ‘farming as a service’–a lease service on a per-hectare, per-year basis. First, all the risk is taken off the farmer; and second, they can trial it as they want to–maybe a field, maybe 20 hectares–and see whether it works. Breaking down those barriers to entry and reducing the risk allowed us to have a really open conversation, and they were incredibly positive about it.”
Sam Watson Jones’s farming credentials are another advantage for SRC. “I think this would be much more difficult to launch if it was a bunch of technologists who knew nothing about how farmers work and what they’re worried about. This is an unusual thing–farmers are not used to dealing with startups. But the key point is, we’re not saying this is the end for all uses for the tractor. We’re just asking the question: ‘For the basics of what you’re trying to do–which is to put a seed in the ground, monitor it, and care for it as accurately as you can–is a tractor the best tool to do that job?’ The resounding answer, which all farmers agree with, is that, with the technology available today, it’s not. There are better ways we could do things, and we’ve got an interesting offer around what that better way might look like.”
SRC is developing three robots, named Tom, Dick, and Harry. Tom is a crop and soil monitoring robot, capable of checking each plant individually; Dick is a micro-spraying and non-chemical weeding robot; and Harry does precision drilling and planting. When not collecting data in the field, Tom lives in a charging “kennel” on the farm, where it also uploads its data to WILMA–the company’s Ubuntu-based, AI-driven operating system; Dick and Harry are summoned from SRC as needed.
Key to the whole process is high-resolution digital soil and crop data (via Tom), plus the ability to take action on that data at a similar resolution (via Dick and Harry). Many farms already have soil maps (created by “a bloke with a bucket” as Watson Jones put it), typically with around 200 samples for a 10-hectare field. “But it’s fake digital really,” said Watson Jones, “because there’s only a rough indication where those samples have come from, and the area around those soil samples is assumed to be the same–there’s still a lot of averaging going on. He added: “I haven’t met a single farmer who, hand on heart, would trust those maps as the cornerstone of their farm management decision making.”
“Where we are in terms of our development is taking significant steps to get that base-level data–much more accurate than anything that exists in farming at the moment–and then, when we develop Dick and Harry, farmers are going to be able to take action much more accurately.” Tom (the monitoring robot) can typically cover a 20-hectare field in a day, or a whole farm in 1-2 weeks. “Plants don’t move that fast,” Watson Jones noted, “so in terms of being able to keep on top of changes in soil conditions and the growth of weeds, that will give us as close to a real-time view of the crop as we can hope for.”
The data is a mixture of photographic, near-infrared, and hyperspectral imaging. “Photography determines whether a plant is a crop or a weed, near-IR lets you see the health of the cells in the plant, and hyperspectral is looking at the chemical content of the soil,” said Watson Jones. “We’re collecting a gigabyte of data every ten minutes or so, and one of the reasons Tom is not terribly fast is that a lot of data is collected, and the speed of internet connections on farms means we can’t just shove it up into the cloud. That’s why Tom has to go back to the ‘kennel,’ as we call it, and physically upload data to a local server.” After an initial run through the AI software on the farm’s server, the core data is then uploaded to the cloud (aka “SRC central”).
SRC’s project around data collection, training the AI and running a distributed operating system across the robots, the “kennel” and the cloud received three years of funding from Innovate UK in 2018. “The goal is to gather the data and understand what it means, going from imagery to commands to the other robots, so that they know where to go, giving us a complete end-to-end automated solution,” Watson Jones explained.
SRC’s AI application is developed by a company called COSMONiO, who specialise in deep learning systems that work with small amounts of training data–hundreds or “low thousands” of images, rather than hundreds of thousands. The AI runs on Ubuntu: “It’s a fantastic platform with a lot of expertise and advice out there. It makes absolute sense for us to run our robotic algorithms on Ubuntu,” Watson Jones said.
When ZDNet spoke to SRC in the spring of 2018, the company’s prototype device, called Rachael, was testing the systems for controlling a robot in a rugged outdoor environment, and also collecting data to train the AI. “Over the course of a growing season, we have to understand what wheat plants look like when they’re coming out of the ground, all the way up to when they’re a full plant, so we can determine the difference between wheat and weeds,” said Watson Jones. “That’s Rachael’s core focus–as a testbed for the robot platform and also to gather the training data.” The prototype robot currently uses Raspberry Pi hardware, although SRC is also considering a custom kit going forward.
As SRC worked through the designs of Dick and Harry, it became clear that they weren’t different robots, but the same basic platform with different attachments. On Harry, the planting robot, Watson Jones said: “We won a competition with the Institution of Engineering and Technology at the end of 2017 to develop a way of planting which doesn’t need the plough. One of the reasons tractors are so heavy is they need to drag a plough through the soil, and as our key ethos is to have light vehicles that won’t compress the soil, we need to have a different type of planting. So rather than ploughing the soil and then dropping seeds in the ground using a drill, we look to precision-place the seeds in the ground using a technique that doesn’t require draught force–the force of pulling something through the ground.”
The precision-planting mechanism, which is being co-developed with the Manufacturing Technology Centre in Coventry, takes inspiration from a hand-pushed punch-planting unit called a haraka, which was developed for use in sub-saharan African countries and in India, said Watson Jones, who described it as “a very clever and very simple mechanism.” The big technical challenge facing SRC, he explained, is to get “the same number of seeds in the ground as a tractor going at full lick with a drill behind it, planting at the normal rate.”
So what’s next on the SRC roadmap? “We’re looking to trial the Harry mechanism, which will hopefully happen with the planting of winter wheat in September–or spring wheat next year  as a fallback,” said Watson Jones. “Our focus is to prove the point with the punch-planting, and then go into various trials with commercial partners to test the benefits of that.” (The latest Harry prototype was unveiled on 7 November at the Agri-Tech East REAP conference at Cambridge.)
And in the longer term? “We’re aiming for commercial launch in 2021 and are recruiting a group of 30 farmers who are essentially our first customers,” said Scott-Robinson. “They each pay us a pre-sale deposit of £5,000, and we work with them to develop all elements of our service. They become our Farmer Advisory Group, and we’ll be constantly on farms, testing and getting feedback. They will tell us the point at which it’s ready for us to charge for the service.”
The spectre of Brexit
Taking a wider view of agriculture in the UK, the spectre of Brexit looms increasingly large, with key issues being what will replace the annual £3 billion in EU farming subsidies, and the likelihood of an exodus of vital migrant labour as free movement ends.
Given scare stories of vegetables rotting in the fields for want of workers to pick them, for example, are harvesting robots on SRC’s roadmap?
“They are on our roadmap, but not on our initial three-year plan,” said Watson Jones. “Broad-acre arable farms growing things that go through a combine–wheat, barley, and oats, which is where we’re squarely aimed–already use a pretty small amount of labour. Initially, those guys will have a role monitoring robots as they go around the fields, but ultimately our message is to try and encourage farmers to diversify–either adding value to what they’re producing, selling things under their own brand, or expanding into other industries. We can help to increase the economic output of a farm.”
The underlying problem, says Watson Jones, is that UK farmers are essentially producing commodity crops in one of the most expensive countries in the world to produce commodities, which aren’t highly valued by the market. “We think big chunks of that can be automated and done better by robots. The labour that’s made available will have to get up to speed with new technology, but can also be used on projects that add value to the overall farm business.”
As far as vegetables rotting in the fields are concerned, Watson Jones noted that there are a smaller number of farms in this sector that currently require a lot of manual labour to get the harvest in. “I do think that robotics will play a part in things like lettuce and strawberry harvesting, but it will be a longer, slower road,” he said. “One of the reasons we didn’t go down that road is, there are fifteen lettuce farmers in the UK for example–they’re all big players, but the market is pretty tiny. They also do things in different ways, so you’ll be forced into designing and developing expensive bespoke robots to do something very specific–and really, you can’t operate that using a farming-as-a-service model.”
Adoption of vegetable-harvesting robots will be slower, Watson Jones added, requiring more advanced robotic technology than is currently being implemented. “The model that we’re suggesting will be earlier, and easier to adopt,” he predicted.
On the subject of EU subsidies, the likelihood is that “a big chunk” of the annual £3 billion is going to disappear post-Brexit, said Watson Jones (in April 2018). “At the moment, Andersons–a farming consultancy–suggests that 85% of UK arable farms are not profitable without subsidy. So if, as some have suggested, a half or two-thirds of that subsidy disappears, a lot of farms will need a radical change in their business model in order to stay afloat. We think we can answer that problem quite effectively.”
Another project to emerge from Harper Adams University, in collaboration with precision farming services specialist Precision Decisions, is HandsFree Hectare, which kicked off in November 2016 with a simple goal: “To grow and harvest a hectare of cereal crops; all without stepping a foot into the field.”
The idea was to leverage open-source technology and already-available small-scale machinery, adapting these components in the university’s labs for autonomous operation in the field. A crop of spring barley would be sown in March 2017, husbanded via remote agronomy and autonomous input application between April and July 2017, and harvested in August/September 2017.
By February 2017 the three-strong HFHa engineering team had tested an automation system on an all-terrain electric vehicle in the field, and were ready to fit it to the (relatively small, light) tractor selected for drilling and spraying the crop, along with a laser scanner-based safety system.
By June, the autonomous tractor–using a modified GPS-based drone autopilot system–had successfully applied pre-seeding herbicide and drilled the field to sow the spring barley seeds. With seeds in the ground, the next challenge was agronomy–using a ground rover to take soil and plant samples, and photos, and a drone to capture multispectral imagery to assess plant growth. Various agrochemicals (fungicides, herbicides, and fertiliser) were then applied based on this data, and the (relatively small, light) combine readied for autonomous operation as harvest time approached.
Harvesting of HFHa’s barley field was completed by September, with a yield of around 4.5 tonnes for the hectare. “This project aimed to prove that there’s no technological reason why a field can’t be farmed without humans working the land directly now and we’ve done that,” said Martin Abell, mechatronics researcher for HFHa’s industry partner Precision Decisions, in a statement. “We set out to identify the opportunities for farming and to prove that it’s possible to autonomously farm the land, and that’s been the great success of the project,” Abell added. “We achieved this on an impressively low budget compared to other projects looking at creating autonomous farming vehicles. The whole project cost less than £200k, funded by Precision Decisions and Innovate UK. We used machinery that was readily available for farmers to buy; open source technology; and an autopilot from a drone for the navigation system.”
The HandsFree Hectare project received plenty of media attention (both industry and mainstream), leading to funding from the Agriculture and Horticulture Development Board (AHDB) for a second crop–this time, of winter wheat. The goal for 2017/18 was to increase the yield via more accurate ground-based machinery and improved remote agronomy; after a weather-affected false start, the winter wheat was successfully drilled and sown in November 2017. Fertiliser, fungicide and herbicide were applied through April/May 2018 and remote sensing carried out for spores of yield-reducing fungal diseases of wheat (specifically, Septoria and yellow rust). The wheat was harvested in early August 2018, with an estimated yield (via drone telemetry) of 6.2-7.8 tonnes per hectare (the final figure turned out to be 6.5t/ha). All, again, without a human setting foot on the field.
HandsFree Hectare has won several awards, including the Future Food category at the BBC Food & Farming Awards in June 2018.
Smart farming and mobile coverage
Agricultural IoT devices need to send and receive data over fast, reliable wireless connections, which means that the availability of mobile broadband in rural areas is a critical factor if projects like Small Robot Company and HandsFree Hectare are to scale up.
According to Ofcom’s Connected Nations Report 2017, 4G services are currently available in 61% of the “outdoor geographic area” of England and 60% of Northern Ireland, with Wales (25%) and Scotland (17%) lagging far behind even this moderate coverage.
To improve mobile coverage in rural areas, Ofcom has announced plans to impose obligations on mobile operators bidding for 700MHz spectrum (part of next-generation 5G mobile services), which will be awarded in the second half of 2019 and released in 2020. Ofcom’s 5G roll-out strategy is outlined in its Enabling 5G in the UK report.
In May 2018, 56 MPs from the All-Party Parliamentary Group (APPG) on Rural Services signed a letter calling on Matt Hancock, (then) secretary of state for Digital, Culture, Media, and Sport at the time, to ensure that 95% of the UK gets mobile coverage from all four operators–Three UK, Vodafone, EE, and O2–by the end of 2022.
The letter argued that market forces are not sufficient to meet the needs of rural areas, and that regulation–legally binding coverage obligation on mobile operators–is required. The APPG also expressed concern that Ofcom’s 700MHz conditions will fall short of the 95% coverage ambition (Figure E). As well as rethinking these conditions, the letter suggested that transparency rules be changed to prevent mobile operators hiding behind “commercial confidentiality” when refusing to divulge their roll-out plans.
It’s still early days for precision agriculture, in the UK and elsewhere. But there are clearly huge opportunities for farming with small, lightweight and autonomous equipment that does less damage to the soil, is kinder to the environment generally, and frees up agricultural workers to contribute to projects that add value to the farm’s output. Projects like Small Robot Company and HandsFree Hectare are at the proof-of-concept stage. The next step is to take them to market and scale them up.
Photo credit for hero image: JIRAROJ PRADITCHAROENKUL/iStock