Agribots and robo-seeding: the future of farming
Image credit: Case IH
Agriculture - humanity’s first industry - has taken centuries to automate, but the pace of change may be about to increase dramatically.
The world needs more food. With the global population expected to reach nine billion by 2050, a 2015 report written for the World Resources Institute argued that agriculture will need to increase production by about 25 per cent to meet demand. You would expect farmers to prosper in this environment. Yet a report on the UK agricultural industry published in mid-2016 by the Department for Environment, Food and Rural Affairs revealed some harsh realities for farmers.
The authors estimated total income from UK farming fell by 29 per cent in real terms, to £3.8bn, between 2014 and 2015, largely as a result of low commodity prices. This kind of mismatch points to one of the major problems facing an industry that needs to find new ways to increase productivity to meet long-term demand.
The UK’s shortage of workers willing to work the fields - and the need therefore to import them from other countries - is mirrored across the developed world. Automation through robotics provides farmers with a way to cope with a labour shortage that, in the case of a hard Brexit, could well arrive quickly.
Robotic seeding and weeding systems help farmers increase yields. Automated combine harvesters have been used in large-scale wheat farms for decades. Aerial drones can help farmers to herd sheep and analyse soil conditions so fertiliser and water can be used precisely. From robot milkers to autonomous vehicles, automation is set to transform global agriculture over the next decade. Best of all, agribots can operate 24/7, 365 days a year and they do not need to be paid, just maintained.
Sudden short-term drops in income such as the one seen between 2014 and 2015 present problems for farmers and the producers of agricultural robots, or agribots. Farmers may not have the resources required to invest in expensive robot technologies.
Even when there is a strong economic case for automation, a six-figure capital outlay is often prohibitive for cash-strapped farmers, says Sara Olson, an expert on agricultural robotics and analyst at Lux Research. There are also risks for early adopters. As Olson put it, mess up a test run at a manufacturing facility and you could ruin a day’s work; mess up with an agribot and you could ruin an entire season’s crop.
“With agricultural robots the outlay is often too much regardless of how much of a case the robot could make for itself. But on top of that most agricultural robots can’t make a good case yet because there isn’t good data,” says Olson, who is upbeat about the future of agricultural robotics, but with the caveat that it will take eight to ten years before we are likely to see widespread adoption.
Growers all over the world are typically conservative and quite sceptical about innovations and disruptive technologies, says Gad Kober, co-founder and vice president of business development at FFRobotics, an Israeli firm that developed the Fresh Fruit Robotic Harvester - a fruit-picking robot that the company plans to make commercially available in 2017.
“The same is true of venture capitalists - they are rather distant from the agro world, though this is slowly changing. It is so much easier to obtain financing for internet-based applications, and there are a lot of fields in which venture capitalists can expect a higher return on investment than agriculture,” says Kober.
Venture capitalists (VCs) often expect to get their money back in five years or less, Olson adds. “The time to legitimate revenue in agricultural robotics can often be significantly longer than 10 years. It ends up being really difficult, especially for corporate VCs, to think of making investments that aren’t going to pay any dividends until after the careers of everyone involved in that decision are done,” she says.
Despite the investment challenges, the agribot has begun to take root. The unmanned aerial vehicle (UAV) and autonomous steering technologies have seen meaningful levels of market penetration in some regions. The market for agribots overall is set to grow from $0.8bn in 2013 to $16.3bn by 2020, according to market analyst Wintergreen Research.
A 2016 report from online VC middleman AgFunder found that investors poured $4.6bn into food and farm technologies in 2015, up from $2.3bn in 2014. Eight per cent of the total went to drones and agribots.
One of the success stories so far for robots in agriculture lies in dairy farming - an industry that has been plagued by wafer-thin margins for years. In use for well over a decade, milking robots are currently used on an estimated 5 per cent of UK farms, according to Liz Snaith of the Royal Association of British Dairy Farmers, but they comprise 30 per cent of all new milking systems in the country.
In a typical system, cows line up for milking up to five or six times per day. The robot scans, maps, and cleans the cows’ underbellies and teats, while a computer tracks each animal’s milking speed - an important piece of data for those running a 24-hour-a-day dairy farming operation.
The nature of the take-up of robots in milking indicates the way in which the business of agriculture may shift in the future. Milking systems are six-figure investments, so they tend to be purchased by large dairy farms and cooperatives.
Some argue that milking robots do not have sufficient autonomy and intelligence to be classified as robots. It is a different matter for crop-harvesting robots such as FFRobotics’ Fresh Fruit Robotic Harvester. The robot platform has 12 autonomous robotic arms: six on each side each able to pick and handle fruit. Sensors supported by dedicated image-processing software map the trees and use an algorithm to identify the location, size and colour of fruit. It is not autonomous enough to roam the fields. When the fruit in one area has been harvested, a human drives the system to its next location where the autonomous harvesting process can begin all over again.
The system’s total picking capacity for apples is 10,000 fruits per hour, says Kober. The robot can work three shifts a day, is 10 times faster than human workers and is capable of picking apples and all kinds of citrus fruit, as well as peaches and pomegranates.
Not all orchards are good candidates for fruit-picking robots, however. Those that use ‘fruit walls’ - which were originally developed to make fruit-picking easier for human workers - are ideal, but while the majority of orchards and groves in Europe use this technique, in the US that number is around 20-25 per cent, Kober says.
High-value crops make for better-value agribots. With six-figure price tags, machine harvesting tends to make more economic sense today in the wine industry because of the high value of some of the fruit. Mechanical harvesting in California costs somewhere between $100 and $300 per acre (0.4 hectare) of grapevines compared to a hand-harvesting cost of up to $750 per acre.
The strawberry industry - worth $2.9bn in the US alone in 2014 - is another target for agribot developers. One example is the Agrobot - a $100,000 robot, also with 12 arms - developed by Spanish engineer Juan Bravo.
Equipped with several robotic manipulators, the Agrobot analyses each strawberry individually, by looking at the shape and colour to identify which strawberries are ripe. Having judged the ripeness of the fruit, small metal baskets attached to the robot’s arms collect the strawberries; they are then transported via a conveyor system to a packaging area.
In a similar vein, Abundant Robotics, a spin-out from California-based SRI Ventures, has developed a robot for apple harvesting. The company was launched in August and aims to become an important player in the fruit-picking agribot space. It hopes to have robotic apple pickers working in orchards within two years.
In Japan, Panasonic plans to start selling a tomato-harvesting robot in 2019.
Not needing the delicacy and millimetre precision of dexterous fruit pickers, tractors look to be prime targets for robotic autonomy. Case IH demonstrated a completely autonomous tractor as a concept vehicle in August to gauge farmers’ interest in a vehicle with no seat for a driver that could plough fields unsupervised.
In reality, tractors are already evolving into self-driving vehicles; US farmers have been using GPS assistance since the 1990s. In fact, agricultural equipment manufacturer John Deere estimates that about two-thirds of large farmers in the United States use self-driving technology. This lets the vehicles operate over longer hours, at night and when visibility is low. GPS ensures millimetre-level accuracy, which improves seeding and harvesting operations and reduces crop damage and losses caused by drivers going off track.
John Deere has also developed AutoTrac Vision, which uses cameras to detect growing crops and guide chemical sprayers. The firm’s Machine Sync system enables a combine harvester to control the speed and location of a tractor and grain cart for automatic unloading.
UAVs provide another degree of freedom and have the benefit of being much cheaper than most other agribots. A farmer can buy a drone suitable for mapping for less than £1000. However, many farmers prefer to hire specialist agricultural drone operators that provide mapping and analysis services on a contract basis.
One problem with the current approach to robotics is that many are single-purpose machines which are only used, albeit intensively, for short periods during the entire year. Multi-purpose devices would make it easier for farmers to justify the outlay. Is it happening? Not yet, says Lux Research’s Olson, but it’s the direction agribot developers should be moving in.
Taking the example of Harvest Automation’s small mobile robots, which were originally designed for the single purpose of moving potted plants but have since developed into a multi-purpose robot platform, Olson suggests that the firms which focus on developing versatile robots will have the edge in years to come.
As to the long-term prospects for the agribot sector, FFRobotics’ Kober expects up to five major agribot players to emerge “within five years or so” from the currently diverse crop of start-ups and other entrants.
Olson is more cautious: “Five years might be a little bit optimistic. Somewhere in the seven to ten year range, we’re headed for an agricultural robotics industry that looks a lot like the agricultural machine industry today from an organisational standpoint.”
That works out to be a small number of major players but a proliferation of service suppliers and a thriving used-device industry.
Agricultural workers: picking off the pickers
Across all industries, workers born in the EU but not in the UK account for 5 per cent of the workforce but in agriculture that figure soars to 65 per cent, according to Office for National Statistics. That does not include seasonal agricultural workers.
With UK agriculture relying so heavily on foreign labour, the implications of Brexit could be considerable. Should Brexit result in a rapid fall in the number of available workers because of immigration restrictions, UK farmers could face the prospect of food rotting away instead of being harvested.
That prospect has become reality in parts of the United States, primarily because of a decline in the number of immigrant Mexican farm workers. The Partnership for a New American Economy claims US agricultural labour shortages are reducing fruit and vegetable production by an estimated 9.5 per cent, or $3.1bn a year. This is bad news for farmers, but good news for agribot makers. In the future agribots may not be considered an expensive luxury, but rather a necessity.
“Growers are hesitant to bear the capital cost of a robotic harvester,” says FFRobotics’ co-founder Gad Kober. “But [they are] slowly realising - in view of ever-increasing labour shortages and constantly rising labour costs - that they will have no choice but to use such robotic harvesters. Unless, of course, they elect to close down their business.”
Although robots may be recruited to deal with immediate labour shortages, over the longer term the perception that mechanisation has reduced the need for human labour is likely to affect the further take-up of robots, says Gaetan Severac, co-founder of Naio Technologies, a French agribot company best known for its Oz weeding robot. “The point of view that robots are ‘bad things that take jobs from humans’ is often false and very simplistic but it exists and if we don’t communicate properly about this issue it could even begin to slow down the adoption of robots in general.”