ABB Analyzes The Habits Of Bees To Develop New Computer Vision Tech

ABB is putting the bees back into its name with a new project that was born from one employee’s passion which led him to explore the intersection of beekeeping and technology.

At first glance, beekeeping and technology couldn’t be further apart. Beekeeping involves creating an environment for a colony of living organisms to live and thrive in, tending to their needs at home while allowing sufficient freedom to roam and explore. Technology is the lifeblood of ABB, and after an employee at the company’s Krakow, Poland offices setup a hobbyist beekeeping installation on the roof, they started thinking about how it might be a fantastic opportunity.

“We always try to have in this big corporate body which is ABB, also this kind of small company soul. One day we thought, why not establish this here on our roof? But not only to have the honey out of this, but to try to utilize this big population for our scientific purposes,” explains Marek Florkowski, Head of ABB Corporate Research Center Kraków.

One team of researchers at the Krakow facility is especially keen to leverage computer algorithms to interpret and make sense of large data sets from diverse arrays of sensors. Typically, that means sensors installed in electric motors in industrial facilities, power electronics for massive commuter trains, large boats, or pantograph chargers for electric buses.

The 50,000 to 60,000 bees that call the apiary home created quite a buzz among employees at ABB’s Corporate Research Center, and not long after, the team started working on a new installation. Sensors, cameras, and computers were added and the data started flowing. “We started with some simple sensors for humidity, temperature, air conditions, CO2, then we also added some cameras –fast cameras, infrared cameras, hyperspectral cameras,” Florkowski said.

I mean, you had me at hyperspectral, but what the heck, let’s go one layer deeper into what might be a very sticky situation. To test complex computational algorithms designed to crunch through massive amounts of data, you need large amounts of data, and the beehive with its thousands of bees made for a great testbed.

“In science very often, the problem is – like in material science – to have good specimen. Also for big data and algorithms. We need to have a set of data which can represent certain process – and here with the bees you have thermal processes, you have ultrafast behavior, you have slow behavior, you have movement recognition, pattern recognition. So the feedback we have from colleagues is that it is an extremely useful experiment,” explains Florkowski.

The tracking exercise may seem like a novelty with no real practical application, but that couldn’t be farther from the truth at ABB. “The hives actually allow us to play with and test all aspects of data analysis – sensors, data acquisition, data analysis and data visualization,” says Michal Orkisz, Senior Principle Scientist at the research center. “Take an application like bee identification, bee detection and bee tracking. That could directly translate into tracking objects on the factory floor or tracking shipping containers.”

Bees flying in and out of the factory could be workers moving in and out of a factory or pallets of goods moving around a production line. Developing strategies and solution to count the bees as they surge in and out of the apiary gives ABB an extremely complex, extremely detailed solution that can then be scaled up to the appropriate size for a specific application.

“I feel proud in many aspects,” says Florkowki. “If you want to be a pioneering company, you have to try something unconventional. This is really the point.” The hive has turned out to be much more than just a side project for evenings and weekends, but has allowed ABB to let its team explore and play with technology in a way that’s both exciting and relevant. What a sweet combination.