Imagine if your smartwatch or home sensor could think like a human brain without draining the battery in an hour. This work creates a new way to build computer chips that process AI data right where it's collected, rather than sending it to a distant cloud. It's like moving the brain's processing power directly into the fingertips of a device to make it instant and energy-efficient.
Current edge AI processors are too power-hungry and take too long to design, preventing the rapid deployment of smart, secure, and battery-efficient AI devices.
An Ultra-Low-Power (ULP) library of hardware building blocks and a compositional SoC generation tool flow to automate chip design.
If you are a wearable device manufacturer dealing with short battery life in AI features — this project developed ultra-low-power building blocks that make AI processing 100x more energy-efficient.
If you are a smart sensor producer dealing with slow time-to-market for new AI chips — this project developed a compositional design tool flow that reduces design time by 10x.
If you are an edge-AI camera developer dealing with data privacy risks — this project developed secure-by-design processors that protect data and privacy directly on the device.
It reduces the design time by 10x through a compositional architecture design-space exploration and SoC generation tool flow, which lowers engineering overhead.
Based on available project data, the project has demonstrated functional operation at the simulation level and is currently working toward a hardware prototype.
The project data does not specify licensing terms, but it involves a consortium of 20 partners including 7 industry players and 3 SMEs.
The project aims for a very short time to market by reducing the chip design time by 10x.
Yes, it includes transparent compilers that support automated code optimizations and domain-specific languages to bridge the gap between software and hardware.
The project is backed by a diverse group of 20 partners across 9 countries. With a 35% industry ratio (7 companies, including 3 SMEs), the consortium balances academic research from 10 universities with commercial application, ensuring the technology is aligned with the $70 billion edge processor market.
Contact the Technical University of Eindhoven (TU/e) regarding the CONVOLVE project.
Contact SciTransfer to connect with the CONVOLVE consortium for early access to ULP library prototypes.
