Engineers in Europe have built the world’s smallest autonomous rotor drone equipped with a neural network running on just 1/100th of a watt.
An article in Fast Company says the project is a critical step in the push to develop fleets of self-navigating nano-drones – the term means drones of four inches or less across.
Researchers from Italy and Switzerland equipped a Crazyflie 2.0 drone – a model popular with hackers – with a camera, processor and customized neural network that consumes less than one per cent of the 27 gram vehicle’s power supply. The circuit board for the electronics weighs in at 5 grams, adding almost nothing to the drone’s weight and minimizing the drain on the drone’s battery.
The drone’s “brain” uses a new mobile processor called GAP8, which contains eight processing cores optimized for AI applications such as image recognition and analysis. GAP8 is based on the RISC-V variant of the RISC microchip architecture, which saves power by running a lot of simple operations rather than a smaller number of more complex ones. A couple of months ago the inventors of RISC received the Turing Award, sometimes referred to as the Nobel Prize of computing.
Read more details about the project here.
