In the not too distant future, whenever we see a fly lurking in a room, we may have to wonder if it is really an insect or a robot.
A group of engineers at Cornell University, in New York, has developed programming that mimics the way an insect’s brain works , to make small robots that behave like one.
Using tiny hair-like metal probes embedded in the wings, the robot is able to detect gusts of wind and adjust its flight accordingly , to plan its path and land, for example, on a swaying flower.
These small insect robots can thus be used to explore hostile and complex environments with autonomy, inside and outside our planet.
The first of its generation, called the RoboBee, measures just 3 centimeters and weighs 80 milligrams and is equipped with various vision, optical flow and motion sensors.
But, for the moment, RoboBee remains connected to a power source . Researchers are working, yes, so that soon new sources of energy will turn it into a robot more autonomous and adaptable to complex environments. All of this while maintaining its small size and weight.
The professor of mechanical and aerospace engineering and director of the Laboratory of Intelligent Systems and Controls, Silvia Ferrari, has explained it in a statement: “We are developing sensors and algorithms to allow RoboBee to avoid losing control or, if it crashes, survive and continue flying, “explains Ferrari.
A hover stabilized RoboBee simulation using a biologically inspired neural network, learning to adapt to manufacturing irregularities in real time. Credit: Taylor Clawson.
Thus, not even a gust of wind would disorient RoboBee. “We want to develop learning controllers that can be adapted to any situation,” he adds.
Chips inspired by a neural network
The technology that has been able to achieve this is based on so-called neuromorphic chips. Unlike traditional computer chips, which use binary code by combining zeros and ones, neuromorphic chips process pucks of electrical current that ‘fire’ in complex combinations; similar to how neurons do it in the brain.
As if that were not enough, the Ferrari laboratory is currently developing a new class of detection and control algorithms based on events that mimic neural activity, which would be implemented in these neuromorphic chips.
Because the chips require much less power than traditional processors, engineers can feed more data into the same payload.
To accelerate the development of event-based algorithms, a PhD student in Ferrari’s lab, Taylor Clawson, developed a virtual simulator that can accurately predict RoboBee’s movements during flights through complex environments. To do this, he used programming inspired by the biology of a neural network, which is capable of learning in real time.
In addition to greater range and endurance, Ferrari adds that its lab plans to help equip RoboBee with new micro-devices such as a camera, expanded antennas for tactile feedback, contact sensors on the robot’s feet and airflow sensors , with the look of small hairs.
References: Cornell University, New York.
