Performance Analysis and Optimization With Deep Learning Assessment of Multi-IRS-Aided IoV Network

Intelligent reflecting surfaces (IRSs) possess the capability to enrich connectivity within dynamic sixth-generation (6G) vehicular communication networks by redirecting signals in desired directions. This study explores a vehicle-to-vehicle (V2V) communication setup bolstered by multiple IRSs, evaluating its efficacy and pinpointing the optimal IRS considering end-to-end channel conditions tailored for applications within the Internet of Vehicles (IoV). The investigation extends to crafting an optimization framework aimed at maximizing data rates while minimizing transmit power. We derive approximate closed-form equations for key performance metrics, such as outage probability (OP), average symbol error rate (ASER), and ergodic capacity (EC), over an independent and nonidentically distributed (i.n.i.d.) double generalized Gamma (dGG) fading channel. To corroborate our theoretical findings, Monte-Carlo simulations are conducted. Moreover, we formulate closed-form expressions for optimization quandaries leveraging the Karush-Kuhn–Tucker (KKT) conditions. Additionally, we introduce a deep neural network (DNN) framework to extract various performance metrics based on Monte-Carlo simulations and to predict optimizations in real-time scenarios. Our findings underscore that the integration of multiple IRSs, along with augmenting the number of elements within each IRS, significantly amplifies system performance in contrast to existing V2V systems detailed in the literature. Furthermore, the DNN framework mitigates computational complexity and streamlines execution times compared to conventional simulation methods.