Design and isolation prediction of a quad-port orthogonal MIMO antenna with defective ground structure for sub-6 GHz applications using machine learning
Аннотация
The rapid growth of 5G Sub-6 GHz applications, including vehicular communication, wireless networks, and 5G-enabled smartphones, has intensified the demand for compact, high-isolation MIMO antenna systems. However, achieving improved isolation, minimized Envelope Correlation Coefficient (ECC), and high radiation efficiency within limited space remains a critical challenge. To address these challenges, a machine learning (ML)-based compact quad port orthogonal Multiple Input Multiple Output (MIMO) antenna is proposed. The proposed MIMO antenna enhances isolation by chamfering edges on rectangular radiators and integrating a defected ground structure. This design employs an FR-4 substrate, with compact dimensions of 50.475 × 50.475 × 1.6mm 3 . Performance evaluation of the MIMO antenna demonstrates better isolation of −40.938 dB, radiation efficiency of ≥ 89%, an ECC less than 0.001, and CCL < 0.08 bits/s/Hz. To overcome the computational complexity of traditional isolation prediction, ML techniques are employed to estimate the transmission coefficient, where the Random Forest Regressor (RFR) attains high prediction accuracy with an R 2 of 97.6% and minimal error metrics. In addition, the Specific Absorption Rate (SAR) is evaluated using a multilayer human tissue structure at 4.5 GHz, confirming suitability for wearable applications. The simulated performance closely matched the results obtained from the measurements of the fabricated antenna.