A mobile machine is developing feelings for blackberry bushes in a West Virginia University greenhouse. It is wingless. It’s not buzzing. It gently opens the flowers by rolling slowly between rows of plants and extending a small arm with a polyurethane brush on its tip. It’s called BrambleBee, and it may be the most subtly unsettling technological advancement in agriculture at the moment—not because it’s scary, but just because it exists.
The fact that engineers were driven to construct it reveals something crucial about our current state of affairs. It’s difficult to look at the numbers without feeling depressed because bee populations have been struggling for years. In a single recent year, the United States lost about 44% of its managed honeybee colonies. There are seven Hawaiian yellow-faced bee species that are in danger of going extinct. Once common in North American fields, the rusty patch bumble bee has all but disappeared. Disease, pesticide exposure, and habitat destruction have come together to create something that feels more like a gradual erasure than a crisis. Farmers are observing. Roboticists are, too.
BrambleBee functions similarly to a cautious autonomous vehicle. It uses lidar to scan its surroundings, creates a three-dimensional map of the greenhouse, and then methodically plans its path to get to as many flowers as it can. Before the brush comes into contact with each blossom, its arm camera takes a high-resolution picture of it. It keeps track of the flowers it has already visited. One day, the information it gathers—following a blackberry from bud to fruit—may enable farmers to forecast yields with a level of accuracy never possible with bees. This aspect of the project quietly excites Yu Gu, the roboticist at West Virginia University who created the system. The pollination itself seems almost incidental.
Japanese researcher Eijiro Miyako discovered a completely different strategy halfway around the world, and the word “stumbled” is appropriate. He had been using a specific ionic liquid gel for electrochemical applications years prior. He put the bottles in a drawer and ignored them after the gel didn’t perform as expected. Something clicked when he discovered them once more while tidying his laboratory. The gel was unique in that it was sticky, durable, and able to repeatedly pick up and release material. He applied the gel and horse hair to small four-propeller drones, launched them into a flower field, and observed the transfer of pollen. The commercially available drones were inexpensive, costing about $100 apiece. Though rough, the idea was genuine.
Whether any of these scales is still unknown. And every robotic pollinator conversation revolves around that unsettling reality. There are one million acres of almond trees in California alone, according to Marla Spivak, an entomologist at the University of Minnesota who has dedicated her professional life to studying bees. To set the nut, each flower must be pollinated. Currently, two million bee colonies with tens of thousands of foragers are needed for that. It is currently unthinkable how many robots would be required to match that. Almonds are only one type of crop.
In addition, the issue of biodiversity is often overlooked in news reports about swarms of robots. At most, a third of crop pollination worldwide is handled by honeybees. More than 20,000 other species of bees make up the remainder, along with butterflies, beetles, moths, wasps, and flies. Each of these insects has a unique shape, is appropriate for a particular flower, and works for free. Even a small portion of that system would require robotic replacements, which would strain credibility in terms of manufacturing, carbon emissions, and resources. According to Dave Goulson of the University of Sussex, replacing the estimated 3.2 trillion honeybees alone would cost £32 billion annually, every year, with the constant replacement of machines lost due to weather and malfunction, even if you could build each robotic bee for one penny.
The claim that robots and bees might coexist rather than compete is more difficult to reject. For example, BrambleBee appears to do better in controlled greenhouse settings than in open fields. It is able to function all night long. The weather doesn’t confuse it. Theoretically, it could be a robotic pollination service that is sent out for the season before moving on, much like rented honeybee hives do today. James Crall, a Harvard biologist who monitors bee colonies by attaching tiny QR codes to individual insects and using cameras to follow them, is not convinced that robotic pollinators will ever significantly offset the decline of bees. However, he also does not discount the research. Even though the solution seems more like a thought experiment than a plan at this point, the science is truly fascinating.
As all of this takes place, it’s difficult to ignore the peculiar situation we’ve put ourselves in: creating complex mechanical solutions for a problem that could be more directly addressed by improved agricultural policy. The bees are still there. In California’s Capay Valley, the flowers continue to bloom, and Brian Paddock still relies on rented hives to set his almond crop each season. One cautious brushstroke at a time, the robots are gathering data and moving pollen as they slowly roll between rows in controlled greenhouses. What we choose to do with the remaining bees will determine whether they are ever required on a large scale, rather than engineering.
