Mobility-Aware Adaptive Precoding via DNN-Enhanced Channel Estimation for Cell-Free Massive MIMO Systems

Cell-free massive multiple input multiple output (MIMO) systems represent a transformative architecture for next-generation wireless networks, which offers uniform service quality through distributed access points (APs). However, practical deployment faces critical challenges: traditional channel estimation methods suffer from channel aging under user mobility, while fixed precoding schemes cannot adapt to time-varying interference conditions. This paper proposes a unified framework that integrates deep neural network (DNN)-enhanced channel estimation with mobility-aware adaptive precoding selection for cell-free massive MIMO systems. The proposed DNN architecture employs a 52-dimensional feature vector capturing spatial, temporal, and channel characteristics to predict a CSI reliability metric that guides dynamic precoding selection among maximum ratio transmission (MRT), zero-forcing (ZF), regularized zero-forcing (RZF), and conjugate beamforming (CB) schemes. The adaptive mechanism selects interference-suppressing precoders (ZF/RZF) when channel confidence is high, while falling back to robust MRT under high-mobility conditions. Comprehensive simulations with Rician fading channels under urban mobility (5–25 m/s) demonstrate that the proposed framework achieves an average signal-to-interference-plus-noise ratio (SINR) gain of 12.43 dB over conventional MRT-based approaches in the tested setup. In single-user scenarios, DNN enhanced estimation provides consistent gains of 6–8 dB. In multi-user interference-limited scenarios, the adaptive precoding mechanism achieves gains of up to 24.87 dB by effectively suppressing inter-user interference when CSI is reliable. The system maintains robust performance across all tested mobility conditions, validating the effectiveness of joint DNN-based estimation and adaptive precoding for simulated cell-free deployments.