Bees' brains help scientists develop decision-making model for robots

Led by Professor Andrew Barron from Macquarie University in Sydney in collaboration with Dr HaDi MaBouDi, Neville Dearden, and Professor James Marshall from the University of Sheffield have conducted a recent study to enhance our understanding of insect brains.

This has inspired them to develop a new model of decision-making that could help design better robots. 

Cognition power

Decision-making is at the core of cognition, says Barron.

“It’s the result of evaluating possible outcomes, and animal lives are full of decisions. A honey bee has a brain smaller than a sesame seed. And yet she can make decisions faster and more accurately than we can."

Furthermore, Barron highlights that a robot programmed to do a bee’s job would need the backup of a supercomputer. For instance, he refers to the functioning of drones:

“Drones are relatively brainless, they have to be in wireless communication with a data center. This technology path will never allow a drone to truly explore Mars solo – NASA’s amazing rovers on Mars have traveled about 75 kilometers in years of exploration.”

Since bees have to work instantaneously and efficiently when finding nectar before returning to their hives, they face numerous challenges.

Honey bees have to not only avoid predators, but need to quickly determine flowers with the highest nectar. 

When flying, bees face the danger of being aerially attacked, and when they land to feed, spiders and other predators could strike them. Because predators are often camouflaged, it’s harder for bees to dodge them.

Color coding

Dr MaBouDi says that 20 bees were trained to recognize five different colored ‘flower disks’. Blue-colored flowers were the only blooms to contain sugar syrup while green flowers had quinine [tonic water] with a bitter taste for bees where as other colors sometimes had glucose.

Each honey bee was introduced to a ‘garden’ where the ‘flowers’ contained solely distilled water.

The bees were filmed and observed for 40 hours on video. The scientists tracked their path and recorded the amount of time each bee took to decide.

Findings showed that when bees determined whether a flower contained food, they landed on it in an average of 0.6 seconds.

They made a decision just as quickly when they realized that the flower did not have food. 

When bees were unsure about the flowers’ food status, they took more time, approximately 1.4 seconds on average, and the time usually represented the probability of the flower being equipped with food. 

Bee brain

The researchers paved the way for scientists to build a computer model with the ambition to replicate the bees' brains and their decision-making process.

Upon compiling data, the structure of the model looked similar to that of the physical layout of a bee brain. 

Dr Marshall says the study demonstrated complex autonomous decision-making with minimal neural circuitry. The research uncovered the ways in which bees make smart decisions.

“We are studying how they are so fast at gathering and sampling information. We think bees are using their flight movements to enhance their visual system to make them better at detecting the best flowers.”

The scientist say that AI researchers could learn from insects and other simple animals as millions of years of evolution have led to enhancements in their brains that require little power. 

“The future of AI in the industry will be inspired by biology,” says Professor Marshall.

The research was published in the journal eLife on 27 June

Abstract

Honey bee ecology demands they make both rapid and accurate assessments of which flowers are most likely to offer them nectar or pollen. To understand the mechanisms of honey bee decision-making, we examined their speed and accuracy of both flower acceptance and rejection decisions. We used a controlled flight arena that varied both the likelihood of a stimulus offering reward and punishment and the quality of evidence for stimuli. We found that the sophistication of honey bee decision-making rivalled that reported for primates. Their decisions were sensitive to both the quality and reliability of evidence. Acceptance responses had higher accuracy than rejection responses and were more sensitive to changes in available evidence and reward likelihood. Fast acceptances were more likely to be correct than slower acceptances; a phenomenon also seen in primates and indicative that the evidence threshold for a decision changes dynamically with sampling time. To investigate the minimally sufficient circuitry required for these decision-making capacities, we developed a novel model of decision-making. Our model can be mapped to known pathways in the insect brain and is neurobiologically plausible. Our model proposes a system for robust autonomous decision-making with potential application in robotics.