杏吧原创

YOU鈥橵E probably heard about Beemageddon. Over the past few years, colony collapse disorder 鈥� thought to be brought on by a pernicious combination of overwork, bad weather, pesticides and infestations of parasitic varroa mites 鈥� has been threatening to wipe out honeybees all over the world, and with them many of our food crops. It is a story that has run and run in the press.

As ever, the reality turns out to be more complicated than the headlines. First to fall under scrutiny was the idea that the global food supply will collapse without honeybees as pollinators; , and cereals like rice, wheat and maize are wind-pollinated. Then beekeepers began to weigh in, refuting the claims of global die-offs. That鈥檚 not to say there isn鈥檛 a problem: evidence surfaced this year that .

All this uncertainty, however, has led some to wonder whether we shouldn鈥檛 have a contingency plan. Can we build an army of robots to pollinate our fields? Although the task turns out to be Herculean, it has helped inspire research into how robots can help bees in an entirely different way.

The idea of flying robot pollinators seems timely in an era when are being tested. Airborne robot technology is advancing at an impressive clip. Robots have replicated complex insect flight: the , mimics a dragonfly, albeit one of prehistoric size. But we are starting to get the dimensions down to the right ballpark: the DelFly Micro, created by researchers at Delft University of Technology in the Netherlands, . At the University of Pennsylvania, researchers have used swarms of quadcopters, each small enough to fit into the palm of your hand, to pick up and move heavy objects.

See more of the insect-drone front line in our gallery: 鈥�Flying robots and cyborgs inspired by insects鈥�

So it didn鈥檛 seem too controversial when, earlier this year, Robert Wood of Harvard University and his colleagues Radhika Nagpal and Gu-Yeon Wei revealed that they had been developing a that could be sent out to pollinate crops. Bee researchers, however, are not convinced. Juliet Osborne of the University of Exeter, UK, says she is sceptical. Dave Goulson of the University of Sussex, UK, is less charitable. 鈥淚t鈥檚 an idiotic idea,鈥� he says. 鈥淲hy waste money on robots that I doubt will ever be as efficient, and certainly won鈥檛 be as environmentally sustainable, as real bees?鈥� Not to mention being unable to make honey.

A look at the alternatives, however, reveals the wisdom of having a plan B for a worst-case scenario. In the apple orchards of south-west China, where pesticide use all but wiped out bee populations in the 1980s, labourers must pollinate flowers using a bag of pollen and a paintbrush. It isn鈥檛 working well. A recent study of apple farming in Sichuan province found that human pollinators are nowhere near as efficient as bees, and that employing them costs about 10 times as much as temporarily bringing in a set of hives. As a result, many farmers have given up on apples altogether in favour of self-pollinating crops like plums, loquats and walnuts ().

You might think spraying pollen from crop-dusting planes is a viable alternative. But a comparison of insect pollination with pollen dusting for apples revealed that crop-dusting yielded 70 per cent fewer fruit, and that than the insect-pollinated variety ().

Plan Bee

There鈥檚 no doubt about it. Bees and other insects make the best pollinators, and so far we haven鈥檛 come up with an alternative that is anywhere near as good.

But delivering pollen is a far cry from delivering a pizza. To approximate the fuzzy, stripy masters of the air, Wood and his colleagues need to build machines that can find flowers of the chosen crop, navigate to them with pinpoint precision, hover, grab the pollen, and then locate the same kind of flower elsewhere to drop off the pollen payload 鈥� all while being small enough to fit inside a flower in the first place. If they were looking for an engineering challenge, Wood and his colleagues have definitely found one.

Most of the problems come down to size. Virtually none of the parts that go into your average drone can be used on one the size of a bee, at least not if you expect it to fly; batteries, motors and sensors are either too big or too heavy (or both). 鈥淲e have had to develop every single component from scratch,鈥� says Wood.

Larger flying robots usually contain electromagnetic motors, but the magnetic forces they rely on become too weak at the centimetre scale. So the team created biomimetic 鈥渇light muscles鈥� out of piezoelectric ceramics. These contract when a voltage is sent through them and, as with real muscles, become more powerful relative to weight the smaller they are. As a result, the team was able to build a 鈥渞obobee鈥� with a wingspan of just 3 centimetres and a mass of less than a gram. It can hover and flit about just like a real insect and, as the team proudly reported earlier this year, is capable of 鈥渟ustaining flights for longer than 20 seconds without ever approaching a crash鈥� ().

So far, however, the robobee has to be tethered to an external power supply. That tether also connects it to a network of infrared cameras that plots its position and relays it to a computer, which then sends instructions back up the wire to tell the drone where to fly next. That鈥檚 because the robobee has no spatial awareness.

For that, it will need a brain. Can that be replicated in silicon? The honeybee brain has only a million neurons 鈥� the same number as a cockroach鈥檚 brain 鈥� but is a remarkable feat of engineering by any standards, allowing its owner to perform some pretty outstanding cognitive tricks. A bee can, for example, make a mental map of its surroundings, remembering not only the location of the flowers it has visited but also their quality. When the environment changes, it adapts its behaviour and what it tells its nestmates. 鈥淢ost people think that bees are stupid individually and only clever when they get together, but that鈥檚 not true,鈥� says James Marshall, a computational neuroscientist at the University of Sheffield in the UK. 鈥淭hey鈥檙e not little simple, preprogrammed robots: they act autonomously, and they can learn and adapt their behaviour.鈥�

Brain buzz

Identifying a particular flower and navigating to it involves three parts of the bee鈥檚 brain: the mushroom body, important for learning and memory; the optic tubercle, which deals with visual information; and the antennal lobe, which tells the bee when something smells good. Marshall is leading a team of researchers at Sheffield and the University of Sussex in the Green Brain Project, aiming to simulate these elements of a honeybee鈥檚 brain and use it to steer their own bee robot.

Their task is no easier than Wood鈥檚. Marshall points out that most animals with similar skills, such as rats, have brains with a hundred times as many neurons. 鈥淚t鈥檚 a miracle of engineering that we are trying to reverse-engineer, basically.鈥� The project has only just begun, but once the robot is fully kitted out, they hope it will be able to see and navigate like a real bee, and to home in on odours just like a bee finding flowers by their scent.

鈥淭he bee brain is basically a miracle of engineering that we are trying to reverse engineer鈥�

Eventually, even the smallest pollinating robots will need this calibre of bee-like intelligence on board. Marshall says miniaturising their artificial bee brain to actual size 鈥� roughly equivalent to a grass seed 鈥� is likely to be 鈥渄ecades away鈥�.

So we鈥檇 better hope that Beemageddon, should it come to pass, doesn鈥檛 happen soon. Other problems will have to be solved in the meantime. As Osborne points out, robots modelled on honeybees will have a limited pollination repertoire, and cannot take the place of the hundreds of other animal pollinators (see diagram), each specialising in different-shaped flowers and with its own way of getting pollen out of plants. Bumblebees, for example, vigorously shake a flower to dislodge its pollen, an effective trick for fertilising plants such as tomatoes. Honeybees don鈥檛 have that up their sleeve. 鈥淵ou would need very crop-specific robots with different shapes, sizes and flight patterns,鈥� she says.

Honeybee robots won鈥檛 make sense as substitute all-purpose pollinators, says Marla Spivak, an entomologist at the University of Minnesota in St Paul. However, should push come to shove, they might be used to help pollinate monocultures.

Many environments pollinated by honeybees are devoid of flowering plants other than their cash crop 鈥� one example is the 鈥� and are saturated with pesticides. These require the bees to be trucked in from all over, which can further damage their already precarious health. 鈥淚f we want to persist with monocultures,鈥� Spivak says, 鈥渢hen we may be forced to do something as drastic and unsustainable as employing robobees.鈥� At best, this could spare the bees from the health risks of pollinating monocultures. 鈥淏ut a better strategy to do that would be to diversify agricultural systems, plant more flowers and reduce pesticide use.鈥� Goulson echoes the sentiment, arguing that we鈥檇 be much better off helping real bees to get on with the job.

This, strangely enough, is something that an entirely different kind of robot bee might be able to help with, and soon. It will not be a fancy, autonomous flying device, but a sneaky hive-based observer that will monitor the bees and help them stay healthy.

Thomas Schmickl has used the robots, which he calls 鈥渟ocial cyborgs鈥�, to guide bees around a metre-long arena. Schmickl, a bee and robot researcher at the University of Graz in Austria, found that he could do this simply by optimising the robots鈥� temperature. As the bees interacted with each other, the news that being next to the robot was the best place to hang out quickly passed through the bee rumour mill.

This made Schmickl wonder about other ways in which bots might be able to steer bee behaviour, or allow beekeepers to observe them in their hives without disturbing them. This would be a boon to beekeepers. 鈥淓very opening of the hive is a problem,鈥� he says. 鈥�.鈥� A robotic hiveminder could be the beekeeper鈥檚 agent in the hive, reading bee behaviour and preventing problems.

For example, they might act to prevent premature swarming. An overcrowded colony will swarm to find a new hive. The odds of swarming increase as the hive becomes overcrowded and bees start to bump into each other more often, says Schmickl. But beekeepers don鈥檛 want the bees to swarm until they are ready with a new box. So Schmickl鈥檚 team is now designing versions of the hive robots that will be able to sense heat and movement in the colony, and regulate both to influence bees鈥� movements. When the hive robots detect the colonies are getting overcrowded and preparing to swarm, they can vibrate or warm up to calm the bees.

That鈥檚 only the start. Link these bee bots to the internet, and they could surf the web on the bees鈥� behalf, passing on important information such as the weather forecast. Weather is no laughing matter for bees, particularly in spring, when a new brood hatches just as older workers are dying off and the winter stores of honey have been depleted. The weather adds one more layer of danger. Bees cannot sustain flight in winds of about 20 kilometres per hour, so they won鈥檛 forage if there鈥檚 a nasty storm. This combination of factors may explain why colonies are most likely to die in the late winter and early spring.

A weather-forecasting bee bot could short-circuit at least part of the problem. 鈥淚f, for example, you know in three weeks there will be bad weather, you know it鈥檚 a good idea to procure food for the brood now, because it will be difficult to do that in bad weather when there is less pollen around,鈥� Schmickl says. Suitably guided, bees that are having a bad year could survive the pollination season, while their neighbours without robotic minders die off.

This would all happen without ever opening the hive to check on the bees. Schmickl envisions a system that alerts the beekeeper via their smartphone when they need to intervene.

All of which would put bees in the position of being the first non-human animals to use the kinds of technology that only we have so far had access to. The question is, what will the most intelligent creatures on six legs choose to do with all this information? Will they still want to work for us?

This article appeared in print under the headline 鈥淚, bee bot鈥�