The NeuroPsense PoC project had two main goals: (1) construct a demonstrator of a novel mixed-signal neuromorphic processor (developed in the ERC NeuroP project), and (2) analyze the business case for the technology.
In the first phase of the project, we developed and produced the first small-scale production run of the Dynamic Neuromorphic Asynchronous Processor (DYNAP). This is the first generally available demonstrator of ultra-low power neuromorphic processor chips, fabricated using mixed-signal (analog plus digital) technology. The demonstrator consisted of a system comprising four DYNAP chips mounted on a Printed Circuit Board (PCB), a Field Programmable Gate Arry (FPGA) programmable interfacing device, a USB interface and additional electronic components for auxiliary functions. As part of the demonstrator, a software development kit was also developed, accompanied by a well-documented user-manual. The system was demonstrated at five global trade shows in robotics and computer vision during 2017 (Singapore, Vancouver, Turin, Hawaii, Baltimore), as well as a number of smaller regional meetings and workshops. This project raised awareness on the NeuroP project research and promoted the use of DYNAP-like technologies. Early prototypes of the demonstrator were also distributed as research platforms. Customer interest has been high; several units have now been sold and the first customers are now using the processor in research and teaching.
The second phase of the project analyzed the business case for the processor. A comprehensive analysis of the use cases led to a decision to focus on ultra-low-power edge computing applications. This analysis led NeuroP researchers and colleagues to start a spin-off company. As a result of this analysis, the startup company aiCTX was founded (
http://www.ai-ctx.com(öffnet in neuem Fenster)) in early 2017. The strong IP and business opportunity of the case, plus the demonstrated availability of the DYNAP, enabled aiCTX to secure its first investment round during the second half of 2017.
Overall, the project was very successful, having achieved and even exceeded its goals of transferring the results of the ERC project to industry with significant positive momentum. It has high potential for making a significant impact on future computing technologies.
