Channel state information-aware deep neural networks for scalable approximation of ML detection in massive MIMO

Channel estimation errors significantly degrade signal detection performance in massive multiple-input multiple-output (massive MIMO) systems, limiting the effectiveness of conventional linear, near-optimal, and learning-based detectors under practical channel uncertainty. To address this challenge, this paper proposes a channel state information (CSI)-aware deep neural network-assisted detector (DNN-ML) that explicitly incorporates channel estimation error variance into the detection process. The proposed framework employs CSI feature extraction, dimensionality reduction, error-adaptive feature fusion, and reliability-aware decision refinement to jointly exploit the received signal, estimated CSI, and CSI uncertainty. The detector is evaluated for a 256×256 massive MIMO system employing 256-QAM modulation over Rayleigh fading channels with 10% and 15% CSI error variance. Simulation results demonstrate that the proposed detector consistently outperforms conventional linear detectors (ZFE and MMSE), practical near-ML detection, machine learning approaches (KNN and SVM), deep learning baselines (CNN), and low-complexity massive MIMO detectors including conjugate gradient (CG), approximate message passing (AMP), and orthogonal approximate message passing (OAMP). At a target bit error rate (BER) of \({10}⁻³\) , the proposed detector achieves up to 9.2 dB SNR gain over ZFE under 10% CSI error variance while maintaining robust performance under more severe channel uncertainty. An ablation study further confirms that explicit CSI error conditioning provides measurable performance improvement compared with an identical architecture that does not utilize CSI error variance. In addition, the proposed detector achieves the lowest out-of-band power spectral density (PSD) among all considered methods and replaces the exponential complexity of theoretical maximum likelihood detection with fixed-depth polynomial-time inference. These results demonstrate that the proposed CSI-aware DNN-ML framework provides an effective, scalable, and computationally efficient solution for reliable massive MIMO signal detection under imperfect CSI conditions.