Micromachining, Additive Manufacturing and Advanced Encapsulation Techniques for Neuromorphic Sensing and Massive High-Performance Computing Ultra-Low Power Edge Systems with CMOS Nanotechnologies
The present Project is enclosed in the research line of Neuromorphic Systems at IMSE-CNM. With over 30 years of activity, over 15 EU projects portfolio and two successful spin-offs, it develops microchips and edge-computing hardware for bio-inspired event-driven vision sensing, computing, and learning systems. The focus is in the development of vision sensing chips exploiting neuromorphic principles on nanometer-scale CMOS technologies, and low-power Edge-oriented computing chips and systems, including the exploitation of emerging nanotechnology memristive synaptic devices. The research line is highly interdisciplinary, covering from vision sensing, nanoscale memristor based computing and learning hardware, computational neuromorphic algorithms, and applications to high-speed and low-power environments.
On the other hand, IMSE-CNM is setting up a new cleanroom facility for advanced encapsulation, micro-printing, and additive manufacturing. The present proposal is intended to combine the neuromorphic microchip design experience with the new cleanroom facilities to enhance and exploit new technological and research capabilites at IMSE-CNM, applied to artificial intelligence, sensing, high-efficient edge computing and massive data processing of neuromorphic systems. The specific objectives include:
- Training on 3D additive manufacturing techniques for neuromorphic vision sensors: microlense design and printing on top of available/new neuromorphic vision sensors, adding new imprinted light/infrerred/microwave sensors or learning-enabling devices connected to an underneath CMOS chip.
- Training on CMOS neuromorphic nanoscale circuit design technologies, ranging from low-power pico-to-nano ampere analog computing circuits, to fast but power-efficient digital communications.
- Training on nanotechnology synaptic computing devices, such as HfOx memristors, perovskite memristors, nanopore liquid-ionic-based memristors, for exploitation with combined CMOS integrated circuits for low-power learning neuromorphic computing systems.
- Training on neuromorphic computational architectures for massive data processing, to devise hardware-friendly neuromorphic energy-efficient computing systems. This training would be purely software-oriented, acquiring knowledge on available AI computational tools.
- Training on vision sensor chip design for neuromorphic-based vision. These vision sensors are known as "Dynamic Vision Sensors (DVS)". The PI is one of the world-wide pioneers in these sensors, is co-founder of Prophesee, the company commercializing the highest resolution DVS camera in the market, and has 5 patents licensed/co-owned with Prophesee.
