I²SAC: AI-Enabled Digital-Twin Framework on a mmWave 64-Element Phased Array for Integrated Imaging, Sensing, and Communication

This article presents an AI-enabled unified digital-twin (DT) framework for Integrated Imaging, Sensing, and Communication (I²SAC) on a multi-static aperture comprising of a 64-element dual-polarized transceiver phased array and four open-ended waveguides as RF sensing probes operating in 25–30 GHz. The physical layer is emulated through a computationally efficient, physics-informed dipole-moment propagation model that generates large, customizable synthetic datasets as a function of spatially distributed scattering targets and vector measurements. These synthetic datasets are used to train a deep neural network that learns the inverse channel matrix across space, frequency, and polarization. The DT-predicted RF channel is further refined using sparse hardware-in-the-loop calibration measurements to ensure physical accuracy. Using the refined DT, we experimentally demonstrate three integrated capabilities: (1) Ability to learn and sense the mmWave channel with hardware non-idealities. (2) Diffraction-limited computational imaging with a cross-range resolution of 1.5 cm at a 1-m stand-off. (3) Channel equalization for mmWave link achieving 64-QAM data over 1 GHz bandwidth between the phased-array Tx and an arbitrarily placed external Rx within the learned channel space, with an EVMrms of 3.5%.