Deep learning-based energy harvesting with intelligent deployment of RIS-assisted UAV-CFmMIMOs.

The ever-evolving internet of things (IoT) has spawned hundreds of wireless sensors that communicate via the internet infrastructure. The lifetime and self-sustainability of these sensors are pivotal factors dictating the performance of respective application infrastructure. Radio frequency energy harvesting (RFEH) technology has exhibited the capability of effectively augmenting the battery lifetime of these sensors. In this work, we introduce a novel framework called CURe, which combines the advantages of cell-free massive multiple-input multiple-output (CFmMIMO) and reconfigurable intelligent surfaces (RISs) to provide uninterrupted energy harvesting for IoT devices through RFEH. CFmMIMO integrates the advantages of distributed systems and massive MIMO, while RIS improves the signal strength of the information transfer and RFEH via its passive reflection capabilities. Moreover, we consider unmanned aerial vehicles (UAVs) equipped with CFmMIMO as mobile access points (APs) to better serve the moving sensory devices. To further enhance RFEH, we propose DeNCE, a channel estimation technique based on deep learning (DL) that eliminates the need for traditional closed-form equation-based channel estimation methods. Through evaluation, we first validate the performance of CURe by comparing it with the modified bisection search for max–min fairness (MBS-MMF) algorithm and later corroborate that DeNCE significantly improves the performances of both models. Finally, to optimize the UAV deployment and ensure continuous RFEH coverage, we propose dARL, a deep reinforcement learning (DRL)-based scheduling framework that enables UAV-CFmMIMO swarms to perform continuous energy harvesting in the coverage area collaboratively.