查看文章

This paper investigates the achievable ratesof multi-IRS-assisted multi-hop communications by exploringthe impact of 3D IRS orientation, representedby elevation and azimuth angles relative to the basestation (BS). We firstly formulate the problem as the jointoptimization of the deployment location, 3D orientation, phase shift of IRSs, and power allocation of users, with thegoal of maximizing the sum achievable rate. To addressthis problem, our approach involves the developmentof a novel algorithm, named deep deterministic policygradient (DDPG), which leverages deep reinforcementlearning (DRL). This algorithm iteratively interacts withthe environment, employing a trial-and-error process toimprove its performance. The simulation results demonstratea significant performance improvement achievedby optimizing the IRS orientation compared to othercontemporary approaches that do not consider optimizingthe IRS deployment orientation.