Abstract: Aside from the daunting physics based challenges needed to maintain Moore's Law, it's becoming increasingly difficult to justify the huge foundry investments required. This probable constraint to future electronics provides sensor developers with a unique opportunity to create more "elegant" sensing design options. Instead of continually increasing the amount of data collected in an open-loop manner, we should consider providing "intelligent" feedback to and within sensors in a more "closed-loop" manner. In other words, allow the sensor to dynamically learn how best to allocate its sampling, processing, and communication resources based on both external objectives and internal conditions. Imbedding deep learning machines into sensors designs could help mitigate Moore's Law's possible demise.
|