Robot thinks like a bee, to learn from what it sees
By Ben Coxworth
February 10, 2014
Because of bees' small size, maneuverability and almost machine-like swarm mentality, it shouldn't come as a surprise that scientists are developing tiny flying robots based on the insects. In order to navigate autonomously, however, those robots' artificial bee brains will have to be capable of identifying objects in their environment, and reacting accordingly. Well, thanks to research recently conducted in Berlin, they may soon be able to do so.
The research was conducted by scientists at Freie Universität Berlin, the Bernstein Fokus Neuronal Basis of Learning, and the Bernstein Center Berlin. It incorporated a small camera-equipped robot that looks more like a tank than a bee. The robot was connected to a computer, however, and that computer was running a program that simulated the sensorimotor network of an insect's brain.
The robot was placed in an enclosure, in which both red and blue cylinders were present. When the robot's camera first saw one of the red cylinders, it caused a light to flash, which in turn triggered a virtual "reward sensor nerve cell" in the network. This caused the program to subsequently be drawn to the color red, so when the camera saw it again, the robot would automatically start moving towards it.
The seeing of the blue cylinders, on the other hand, triggered a response in which the robot moved away from the color.
“Much like honeybees learn to associate certain flower colors with tasty nectar, the robot learns to approach certain colored objects and to avoid others," explains Freie Universität Berlin's Prof. Martin Paul Nawrot, who led the research. “Only a single learning trial is needed, similar to experimental observations in honeybees.”
Nawrot and his team are now building upon the artificial sensorimotor network, so that the robot can respond to a wider variety of visual stimuli, with a wider range of behaviors.
Source: Freie Universität Berlin