Hybrid MIMO detectors with quasi-orthogonal space-time block coding and adaptive complexity management based on deep learning

Formulation of the problem. The cubic computational complexity associated with channel correlation matrix inversion in linear Minimum Mean Square Error (MMSE) equalizers constitutes a critical barrier to the implementation of energy-efficient receivers in Quasi-Orthogonal Space-Time Block Coding MIMO systems. This issue is particularly acute given the stringent hardware constraints inherent in the Internet of Things and 5G/6G networks.

Objective. The synthesis of computationally efficient hybrid detection architectures that integrate iterative numerical methods with proactive control based on Deep Neural Networks, aiming to minimize floating-point operations without degrading noise immunity.

Results. Simulation results demonstrate that the proposed algorithms reduce asymptotic complexity to a near-quadratic level while maintaining Bit Error Rate performance comparable to that of the exact MMSE detector. The scalability of the approach for large-scale antenna arrays is also established.

Practical significance. The proposed hybrid algorithms are tailored for hardware implementation in Massive MIMO systems, specifically for Ultra-Reliable Low-Latency Communication scenarios. By mitigating the computational burden, they enable the deployment of less powerful, energy-efficient processors in IoT terminals and base stations for real-time signal processing.