The study proposes an RNN–MMSE detector for massive MIMO and compares it with conventional and learning-based detectors under channel-estimation errors, Rayleigh fading and over-the-air testing.
Key findings
- At BER 10⁻³ with 20% channel error, the method required about 12.2 dB, a 7.3-dB gain over ZFE and roughly 3–4 dB over CNN/RNN. It required 9.8 dB at 10% error, 8.5 dB under Rayleigh fading and 9.8 dB OTA, with complexity O(TN²+N³).
Why this matters globally
If performance holds at larger antenna scales and on deployed hardware, the detector could improve 5G/6G robustness to imperfect channel estimates and reduce transmit-power requirements.
Thai researcher contribution
Aziz Nanthaamornphong of Prince of Songkla University contributed to the work, linking Thai expertise to signal-processing research for next-generation wireless networks.
Limitations to consider
The abstract does not fully report system scale, training data, channel diversity or OTA replication. Complexity exceeds simpler detectors, and latency, memory, energy use and large-array hardware integration are not quantified, limiting deployment conclusions.