Cars Could Follow the Flight of the Bumblebee

Bee cognition gives hints of how radios could talk on the fly

2 min read
Cars Could Follow the Flight of the Bumblebee
Photo: Adam Crowley/Corbis

Someday, in the not-too-distant future, the roads may be filled with cars that drive themselves, sharing the space with human-driven cars loaded with collision-avoidance features. Both flavors of cars will happily chatter amongst themselves over a designated portion of the radio spectrum as they seek to make the best and safest driving choices. And the whole system may work thanks to an idea that comes from picking a bumblebee’s brain.

Communication among an ever-shifting collection of moving vehicles—plus roadway infrastructure such as traffic signals—is no easy task, especially given that the bandwidth allocated for the job is relatively small: just 75-megahertz wide, at a frequency around 5.9 GHz. Even with individual channels taking up only 10 MHz (as required by the IEEE 802.11p wireless standard), during rush hour, the airwaves may quickly become as congested as the roads. “You’re going to run out of bandwidth very quickly,” says Alexander Wyglinski, a professor of electrical and computer engineering at Worcester Polytechnic Institute in Worcester, Mass.

Wyglinski’s proposed solution relies on cognitive radios, devices that can scan the radio spectrum and pick the optimum frequency, making independent decisions that nevertheless don’t interfere with nearby broadcasts. The U.S. Federal Communications Commission allows unlicensed use of unused bits of the radio spectrum normally reserved for television broadcasts; Channel 6 might be empty in one city while Channel 11 is free in another. Cognitive radio could rely on whatever spectrum was available locally.

“People usually say ‘cognitive’ must be human cognitive,” Wyglinski says. But he and two colleagues from WPI’s biology and biotechnology department, Robert Gegear and Elizabeth Ryder, think bumblebee cognition might be the better bet. The trio has just received a three-year, $300,000 grant from the U.S. National Science Foundation to pursue their hypothesis.

The idea is to look at the decision-making process that bumblebees go through when foraging for nectar, and use that to build an algorithm that similarly seeks low-noise radio frequencies. Just as the bees look for flowers with the best nectar supply, the radios would look for channels with the highest signal quality. If a bee finds that a flower is running out of food, or is being swarmed by other bees competing for the supply, it can quickly make a decision to switch to another flower of the same color and shape, or pick a different type of flower that may not have as much bee-fueling nectar but is a shorter flight away. All that decision-making involves learning, memory, the ability to weigh costs and benefits, and the capacity to adapt, in a split second, to changes in the environment. Those same skills will be required if a cognitive radio is to find the right channel on the fly, says Wyglinski.

“It really comes down to a collection of learning models,” he says. Computer scientists might eventually work out a scheme on their own. But why not turn to nature, the WPI researchers ask. It has, after all, been optimizing this algorithm over thousands of generations of bumblebee evolution.

Researchers who study networks have looked at communications strategies used by other insects, such as ants and honeybees. But those rely on the transfer of large amounts of information (in ant trail pheromones, for example), whereas as bumblebees share information but can also act independently, Wyglinski says. 

The Conversation (0)

The Future of Deep Learning Is Photonic

Computing with light could slash the energy needs of neural networks

10 min read

This computer rendering depicts the pattern on a photonic chip that the author and his colleagues have devised for performing neural-network calculations using light.

Alexander Sludds

Think of the many tasks to which computers are being applied that in the not-so-distant past required human intuition. Computers routinely identify objects in images, transcribe speech, translate between languages, diagnose medical conditions, play complex games, and drive cars.

The technique that has empowered these stunning developments is called deep learning, a term that refers to mathematical models known as artificial neural networks. Deep learning is a subfield of machine learning, a branch of computer science based on fitting complex models to data.

Keep Reading ↓ Show less