OTA-Validated Hardware-in-the-Loop Framework for mmWave SDR Transceiver Performance Optimization

Abstract

Millimetre-wave (mmWave) communication systems with rapidly changing channel conditions, hardware inefficiencies, and beam misalignments do not have enough plant stability when using software-defined radio (SDR) transceivers, in which setting of the channels ought to be performed by an innovative approach called static configuration strategies. The article is a report of an over-the-air (OTA)-validated Hardware-in-the-Loop (HIL)-based validation framework that incorporates Deep Reinforcement Learning (DRL) as a real-time adaptive method of transceiver performance optimization. The presented architecture creates a closed feedback architecture where live RF measurements such as Error Vector Magnitude (EVM), bit error rate (BER), Adjacent channel power ratio (ACPR) and received signal strength are constantly fed to a DRL agent to dynamically change transmission parameters such as beam index, modulation order, transmit power, and RF front-end control variables. Deep Q-Network (DQN) is used to solve discrete control problems, and Deep Deterministic Policy Gradient (DDPG) and Proximal Policy Optimization (PPO) are used to usher continuous RF parameter tuning. This scheme is experimentally tested by using real time OTA experiments on a mmWave SDR platform, on a controlled propagation channel. The experimental findings exhibit up to 30 percent decrease in the EVM, 28 percent increase in the BER performance at the midSNRs, 5 to 7 dB increase in the ACPR, and 18 percent increment of the throughput relative to the static and heuristic settings. The suggested framework of HIL-DRL reaches fast convergence within under 150 ms adaptation periods and spectral compliance. This supports the viability of closed-loop RF optimization, which includes intelligence and supports the use of the framework in next-generation adaptive mmWave and beyond-5G wireless systems.