An Optimized Wearable Textile Antenna Using Surrogate Ensemble Learning for ISM On-Body Communications

In this article, a novel microstrip antenna for the new industrial, scientific, and medical (ISM) band (24-25 GHz) with a flexible substrate is proposed. This antenna has a small dimension, high gain, and large bandwidth (BW) in the ISM band. Several tests such as bending tests and gas environment tests are performed to show the potential of the antenna in body area networks' (BANs) applications. The design procedure starts with a simple monopole antenna with a simple ground plane. Then, to change the effective permittivity of the substrate, an array of half circles is subtracted from the ground plane. To avoid increasing the dimension of the antenna, a triangular part and a shunt are added to the monopole antenna. By doing this, the resonance frequency of the antenna is changed without changing its overall dimension. The theoretical bases for creating an initial antenna are discussed to achieve the values of the initial antenna. Then, to achieve an optimum antenna, the initial antenna is treated with a machine learning approach with the goal of maximizing antenna gain and BW. The so-called DDEA-SE (data-driven evolutionary algorithm based on selective ensemble) is used as the machine learning approach. The optimum antenna has 1.8 GHz (7.6% IBW) BW which covers the 24-25.8-GHz band. The optimum antenna has a 4.9-dBi maximum realized gain on 25-GHz frequency. The optimization procedure increased the average gain of the antenna near 1 dB and the BW of the antenna near 500 MHz with respect to the initial antenna.