Imagine a computer chip that sees and thinks like a honeybee, using almost no energy. Instead of taking constant photos like a normal camera, it only notices things that move or change, which saves massive amounts of power. This allows tiny drones or glasses to 'see' their surroundings instantly without needing a giant battery or a connection to a cloud server.
Current AI processing in IoT devices is too power-hungry and slow because data must be sent to remote servers. This creates bottlenecks in latency, security, and battery life for small autonomous devices.
A 3D-stacked neuromorphic chip architecture featuring digitally-foveated and light-field Dynamic Vision Sensors (DVS) and Spiking Neural Networks (SNNs).
If you are a micro-drone manufacturer dealing with short battery life and slow reaction times — this project developed neuromorphic vision chips that provide 100x energy-efficiency improvement and 50x latency reduction. This enables the creation of autonomous drones the size of bumblebees.
If you are an AR glass developer dealing with bulky hardware and overheating due to heavy AI processing — this project developed a 3D-stacked sensing-processing solution. It allows lightweight augmented reality wearables to run computer vision tasks within energy envelopes of tens of mW or less.
If you are an IoT sensor provider dealing with high data costs and privacy risks from sending video to the cloud — this project developed endpoint processing that handles AI tasks locally. This reduces the need to travel to remote data centers, improving security and reducing latency to units of ms or less.
Based on available project data, specific pricing or cost-per-unit information is not provided.
The project utilizes 3D-integrated circuit technology and microelectronics to create chips for endpoint devices, though specific manufacturing volume capacities are not mentioned.
Based on available project data, the specific licensing terms for the neuromorphic architecture and DF-DVS technology are not disclosed.
The architecture includes a DVS front-end that creates data structures compatible with mainstream AI models, bridging the gap between neuromorphic hardware and standard AI engines.
The project period runs from 2022-10-01 to 2026-03-31, suggesting the technology will be developed through early 2026.
The project is highly industry-driven with a 38% industry ratio, comprising 21 partners across 9 countries. The balance of 8 industry players (including 6 SMEs), 7 universities, and 6 research centers indicates a strong push to move the technology from lab to market, specifically targeting the microelectronics and sensor supply chain.
Contact IKERLAN S. COOP in Spain for technical specifications
Contact us to identify licensing opportunities for neuromorphic vision hardware.
