The project aims to develop and apply new Artificial Intelligence (AI) techniques and tools in the design and operation of Radio Frequency (RF) to Digital interfaces for Software Defined Radio (SDR) transceivers intended for the Artificial Intelligence Internet of Things (AIoT). The specifications of the RF-to-Digital converter, which constitutes the front-end of the system, involve digitizing a wide spectrum of signals with a programmable resolution of 8-12 bits within a programmable bandwidth of 30 kHz to 300 MHz, and a carrier frequency that will vary between 0.4 GHz and 6 GHz. The operation of the system will be managed by an Artificial Neural Network (ANN), which could even be integrated alongside the electronic circuitry of the converter, and which will "learn" to predict the best frequency band to establish communication with a reduced number of interferences.
The device will combine Spectrum Sensing (SS) techniques with the capabilities of ANNs to extract patterns and make short-term predictions (hundreds of ms) about occupancy/interference/noise levels in the available bands, and make autonomous and adaptive decisions about which band to place itself in. In addition, the microelectronic design itself will be
supported by an AI algorithm-based design methodology that we have already begun to outline (https://doi.org/10.1109/TCSII.2023.3323886) and that will be subject to evolution and improvement during the course of this project. The objective is to obtain neural networks capable of, starting from very high-level system specifications, first providing valid designs at block level, which will be cross-verified with behavioural simulations, and even proposing electrical designs at transistor level of the same, which will be validated through electrical simulation. Thanks to the existence of other sources of funding in the group, these developments can be put into practice in a real implementation of the system in a 28nm technology.
The research carried out in this project is aligned with the technological fields of "Artificial Intelligence", "Sensorisation" and "Advanced data analytics/edge computing" of Annex I of this call, and addresses some design challenges towards a more efficient digital transformation, linked to the strategic actions of the National Scientific and Technical Research and Innovation Program (PEICTI 2021-2023), and more directly with the strategic action AE4 "Digital World, Industry, Space and Defense".
