AI-Empowered Multi-UAV and IRS Collaboration for Spectrum and Energy Optimization in B5G Networks

Abstract

Unmanned aerial vehicles (UAVs) and intelligent reflecting surface (IRS) assisted networks enable high-capacity, reliable beyond 5G (B5G) communications in complex urban environments. However, existing systems suffer from inaccurate channel state information (CSI) in non-line-of-sight (NLoS) scenarios, inefficient dynamic spectrum sharing under mobility, and unsustainable energy consumption during UAV operations. Therefore, to address these issues, this work proposes an orthogonal matching Markov chain Monte Carlo pursuit (OMCMCP) dynamic multiagent resource optimization framework (DMROF) along with an iterative heuristic algorithm (IHA) for channel estimation, spectrum allocation, and energy-efficient UAV trajectory planning. We formulate an optimization problem for dynamic spectrum allocation in multi-UAV and IRS-assisted networks to maximize the optimal spectrum allocation, power, and trajectory control values. OMCMCP combines sparse recovery and Bayesian inference to acquire NLoS CSI. At the same time, DMROF uses a multiagent dueling deep Q network (MADDQN) for real-time spectrum allocation along with quality of services demand-based power allocation (QDPA) to distribute power. The IHA optimizes the trajectory of UAVs by iteratively adjusting the path to minimize energy consumption, considering factors, such as flight distance, speed, and the energy required for hovering, maneuvering, and communication. Simulation results show that the proposed approach achieved higher spectral efficiency (EE) (16 b/s/Hz), throughput (16 Mb/s), data rate (35 Mb/s), signal to information ratio (SINR) coverage (95%), and lower energy consumption (500 J) compared to the state-of-the-art. These results validate the capability of the proposed framework to jointly optimize UAV-IRS for spectrum utilization and energy EE in smart city deployments and mission-critical Internet of Things (IoT) applications.