Project NeurONN: Energy efficient neuromorphic computing

Nature sets a high bar: Our brain is capable of processing and storing huge amounts of information without using more energy than a 20-watt light bulb. This is a good reason for researchers worldwide to take the human brain as a model for circuits in so-called neuromorphic chips. Within the EU project NeurONN, a Fraunhofer EMFT research team is working with six European partners on extremely energy-efficient elements and architectures for neuromorphic computing. Innovative 2D materials are also being used in the process.

The goal is to develop a neurologically inspired computer architecture. There, information is encoded by coupled oscillating elements that are interconnected to form a neural network. Analogous to the brain, the two key components neuron and synapse replicate the distributed computing and storage units. New elements based on vanadium dioxide, which can be 250 times more efficient than state-of-the-art digital oscillators, serve as neurons. So-called memristors - from memory and resistor, storage and electrical resistance - based on new 2D nanomaterials are used as synapses. The tiny components are said to be up to 330 times more efficient than current technologies in terms of switching speed, lifetime and energy consumption.

The neuromorphic chips will be used wherever energy efficiency and low latency are particularly important, for example because devices are battery-powered or there is no time to send data to the cloud and wait for a response. This includes, for example, the processing of sensor data in autonomous driving, satellite applications, predictive maintenance or condition monitoring in Industry 4.0. A major advantage of neuromorphic hardware is also that information is stored locally and not in the cloud, which improves both the security of the devices and data protection. Last but not least, neuromorphic chips serve as a basis for Egde AI applications.

The project is funded under the EU Horizon 2020 research program under grant number 871501.