Channel estimation and passive beamforming optimization in intelligent reflecting surface IRS aided mmWave 5G beyond communication systems

Abstract

The emergence of fifth-generation (5G) and beyond mmWavecommunication systems has brought significant improvements in data rates, connectivity, and spectrum efficiency. However, these systems face challenges which includes high signal path loss, and frequent signal blockages. This degrades the overall system performance. Intelligent Reflecting Surfaces (IRS) have recently gained attention as a promising solution to overcome these issues. IRS reconfigures the wireless propagation environment in a cost and energyefficient manner.In highly dynamic environments such as vehicular communication scenarios, the performance of IRS based systems deteriorates heavily. This thesis focuses on two critical aspects of IRS-assisted millimeter-wave (mmWave) vehicular communication systems: channel estimation and passive beamforming. First, we propose a two-stage cascaded channel estimation technique designed for time-varying channels which involves fast-moving users and roadside IRS unit. The first stage utilizes the Compressive Sampling Matching Pursuit (CoSaMP) algorithm to exploit channel sparsity and the second stage applies the Extended Kalman Filter (EKF) for real-time tracking and prediction of channel parameters such as Angle of Arrival (AoA) and Angle of Departure (AoD). This framework ensures highly accurate estimation and significantly reduces the pilot overhead and computational complexity and in the second part, a novel algorithm with Successive Convex Approximation-based Interior Point Method (SCA-IPM) is developed for optimizing IRS reflection coefficients. The method addresses the non-convexity of the passive beamforming problem by iteratively linearizing the objective function. It also efficiently solves the resulting convex subproblems with Interior Point method. Simulation results show that the proposed SCA-IPM based method achieves superior achievable rate and Signal-to-Noise Ratio (SNR) performance compared to benchmark techniques like Alternating Optimization (AO), Semi-Definite Relaxation (SDR),