There have been investigations into developing standard protocols for connecting spike-based sensors & computational units but these have so far remained disparate academic efforts. As we try to take SNNs & neuromorphics into the commercial arena, having an easily usable & agreed standard will make it easier for groups to explore and inter-operate with each other. A first step would be to achieve a consensus about the requirements & forming a working group with the necessary technical expertise to drive the process.
If it could allow simple plug-and-play between devices that would be helpful. AER could be the underlying information & it may be that a small FPGA- or ASIC-based box may be part of the implementation.
Initial discussions suggest at least 2 requirements and which therefore might require variants or other intelligent handling:
1) A raw, unsynchronous transport of spikes without timestamps and with the absolute minimum overhead.
2) A safer and more conservative transport with reference headers, timestamps and potentially payloads and error-correction.
(1) would be useful for very local connections between high-bandwidth sensors and computing resources e.g. within a robot or between local compute. (2) would be better for more distant connections which are less hard-wired and more likely to be subject to corruption, delays and other uncertainties.
Please put all your thoughts, worries, ideas and everything else on this page where we can collect the discussions:
https://docs.google.com/document/d/1LqjiCzxUczIKLnNxSvZqpIQ-vBTizIllzJyNJoiYoEw/edit?usp=sharing
Transport-Independent Protocols for Universal AER Communications, S. Arik et al., pp. 675–684, 2015. DOI: 10.1007/978-3-319-26561-2 79
Go to group wiki Go to wiki users Info
| Day | Time | Location |
|---|---|---|
| Thu, 02.05.2024 | 15:00 - 16:00 | Sala panorama |
| Wed, 08.05.2024 | 15:00 - 16:00 | Sala panorama |