Researchers in Berlin have created a robot, with the brain of an insect, not literally, but a model based on the nervous system of the honeybee.
The robot has been able to learn from external sensory data (from a camera) and passing it through a neural network based on the honeybees nervous system which parts of the environment to steer towards or clear of. Through simple color detection the robot was able to navigate to red objects in its environment and avoid blue ones. This was done by reinforcement of red being positive via a flash of light when the bot was near the red objects, similarly to how a bee would learn that going near the correct flower would provide it with pollen while others would not.
While to some of the more artificial intelligently informed readers may be asking what sets this apart from any other artificial neural network, the abstract claims that “Only a single learning trial is needed, similar to experimental observations in honeybees.” and thus a result is produced within seconds.
The full details can be found in the publication, although it’s behind a paywall (IEEE Xplore). A video demonstrating the process is shown below, and the link at the bottom will take you to a more detailed press release page.
Via: Freie University