Tiny drones dressed in horsehair and coated with a sticky goo have been attempting to pollinate lilies in a Japanese lab. The $100 quadcopters, measuring just 42mm-wide and weighing 14.8g, can be heard buzzing in the background on a short film, before zipping in to make clumsy contact with the target.
“For now, my robot is possibly just a proof-of-concept,” admit Eijiro Miyako, a chemist from the National Institute of Advanced Industrial Science and Technology (AIST) Nanomaterials Research Institute (NMRI). “But it does make sense for the future. We're doing the best; we’re looking forward,” he told WIRED. Read more: Amazing animals: how wildlife is inspiring science and technology
The project came about entirely by chance. A decade ago, Miyako was investigating liquids to be used as electrical conductors. One output, a sticky gel made using an acrylic to polymerise an imidazolium salt, failed to work. It was left in a storage cabinet, in a bottle with no cap, and forgotten. Two years ago, when Miyako had to move out of the lab, he decided to tidy up. “I noticed the gels collected dust, which made me think they might work to attract pollen.” At the time he was “emotionally motivated” by what was happening to honeybee populations. “I thought we urgently need to create something for these problems.
"This project is the result of serendipity. We were surprised that after eight years, the ionic gel didn't degrade and was still so viscous. Conventional gels are mainly made of water and can't be used for a long time, so we decided to use this material for research."
Ants were the first test subjects, with 30 anointed with a gel droplet and left in a jar of tulips. “After three days, we carefully observed their body by electron microscopy to count pollens. We [compared] the hybrid ants and wild ants without the gels - hybrid ants had 1,000 times more pollen than wild ants.” The gel also has a camouflage effect, thanks to the photochromic chemical compounds in it, says Miyako. “I'm sure that camouflaging properties of this gel would help our artificial pollinators to protect against attacks from enemies like clever birds and bigger insects.”
To help the gel adhere to drones, then brush off on flowers, Miyako’s team turned to horsehair. “The diameter of horsehair was ideal for coating the gels and collecting pollen. More importantly, horsehair is a biodegradable, eco-friendly material.” The flying robots could carry about 2mg of pollen per flight, attached to the vertically-aligned horsehairs. But the drones need to be painstakingly manoeuvred to the flowers using a radio-wave controller.
“Indeed, it was very hard work to control the robotic pollinators to precisely hit the target sites,” Miyako explained. “I believe that a form of AI, GPS and high-resolution cameras would be very useful for the development of automatic machines in the future. In particular, AI will help provide the intelligence in a robotic drone. They will [autonomously consider] something such as exploring the shortest path and the highest efficiency for pollination. That must be helpful and useful for farmers, right?”
A 2016 study looking at 18 years of data on bee populations in England found that insecticides were linked to a longterm decline in the pollinators. Another paper out the same year identified 57 pesticides licensed for use in the European Union that are poisonous to honeybees. Meanwhile, an estimated $235 billion to $577 billion worth of global crops depend on pollinators, a report by the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES) revealed in 2016.
Miyako believes the project could help counter this enormous problem, by automating manual crop pollination. “We hope the robotic pollinators might replace bees [to counter] the problem of honeybee decline - but more importantly, we should use both together. They might cause good synergistic effects.”
And his team isn't the only one attempting to solve this issue. University labs in Sheffield, Sussex and Harvard have been building robo-bees of differing kinds for several years now.
In particular, the computer scientists at the universities of Sussex and Sheffield have been working on a bee brain AI, first through the Green Brain Project and now Brains on Board, which counts Nvidia and Movidius as partners. The team has built software made up of thousands of virtual neurons linked together to mirror a real bee’s brain network. In Green Brain, the drones were taught to fly down corridors without bumping into the sides, linked wirelessly to a base workstation. With Brains on Board, the operation will be taken into the real world where the robot will have to autonomously navigate, identifying the horizon and avoiding obstacles. A credit card-size Nvidia CPU accelerator will be embedded onboard.
Aside from the AI onboard, there’s one pretty stark contrast to Miyako’s project.
“Our quadcopter drones are about 40cm by 40cm,” professor of Informatics at the University of Sussex, Thomas Nowotny, told WIRED. “I personally think [Miyako’s project] is a bit ambitious. Small drones have almost no electronics onboard so you can’t put much AI on there. Things are developing rapidly of course - the kind of processing power you have in smartwatches is more than in computers 15 years ago. But still, it’s a bit of an ask in these microdrones.”
Nowotny and his team aren’t looking to build an army of artificial pollinators. They want to build autonomous flying robots that have the same incredible navigational and learning abilities of honeybees - the drones could help in search and rescue operations, or bomb disposal. It’s the bee’s brains they are interested in, not their pollinating skills.
“Bees can do certain things like navigate autonomously in an environment with no GPS or supervision,” Nowotny continued. “They just fly out and see the world and are able to come back autonomously. They can also make quite sophisticated decisions. We want to replicate these abilities for search and rescue. The drone could fly into an unstable building that doesn’t let GPS penetrate, then look for people and fly back without the need for supervision.”
The team is additionally working on the AI bee’s ability to decipher smells as well as sights. “They could find a source of air pollution or detect explosives - bees use scent to find flowers, we were inspired by that and in Green Brain it could distinguish between two odours based on chemical sensors.”
Although pollinator drones could be a reality one day, Nowotny says we are already “right at the edge of what can be done”, and the drones are still not all that robust.
“The problem is this needs a lot of compromises - even though we are using the newest embedded controllers, we expect our models to be a bit slow.”
Read all about Miyako's work in the journal Chem.
This article was originally published by WIRED UK
