Quantum-Inspired Antenna Subset Selection for IRS-Assisted Wireless Systems

Abstract

Intelligent reflecting surface (IRS) assisted antenna selection systems improve the performance of the wireless communication systems with lower cost and power consumption. We focus on the challenging combinatorial problem of joint optimization of antenna subset at the transmitter and IRS reflection coefficients. Existing approaches, such as manifold optimization based subset selection (MOBSS), rely on exhaustive search over all antenna subsets, resulting in complexity that increases exponentially as the number of radio frequency chains increases. To address this challenge, we propose a quantum-inspired antenna subset selection framework that reformulates the problem as a quadratic unconstrained binary optimization (QUBO) problem. This enables the use of efficient combinatorial solvers, such as simulated annealing and quantum inspired simulated bifurcation, to obtain near-optimal antenna subsets. The reformulated problem is equivalent to the original problem while significantly reducing computational complexity. For a given subset, the IRS reflection coefficients are optimized using a manifold optimization approach. Numerical results demonstrate that the proposed method achieves performance comparable to MOBSS, while offering substantial gains in scalability, making it suitable for large-scale IRS-assisted systems.