Power-Efficient Symbol-Level Hybrid Precoding: An Adaptive RF Chain Selection Framework

We propose an energy-efficient symbol-level hybrid precoding (SLHP) framework for massive multiple-input multiple-output (MIMO) based on partially connected architectures. A dynamic radio frequency (RF) chain selection mechanism with adaptive connection network (ACN) enhances mapping gains by flexibly activating RF chains. The SLHP optimization is formulated under symbol error probability (SEP) constraints to minimize system power while ensuring robustness against noise. The optimal fully digital precoder is first obtained and then mapped to the hybrid architecture, reformulated as nonlinear least squares (NLS), and solved via Ward clustering-based algorithm. Simulations demonstrate 37.4% power savings compared with benchmarks, confirming scalability and energy efficiency.