Multiple antenna performance parameters estimation of folded dipole antenna using Adaptive Neuro-Fuzzy Inference System trained with Particle Swarm Optimization

Smarter Antennas for Everyday Wireless Gadgets

From WiFi routers to smart thermostats, many everyday devices rely on tiny metal structures called antennas to send and receive signals. Designing these antennas is usually a slow trial-and-error process that demands heavy computer simulations. This paper shows how a blend of artificial intelligence techniques can learn the behavior of a popular antenna type and predict its performance almost instantly, paving the way for faster and more efficient wireless product design.

Why This Particular Antenna Matters

The study focuses on printed folded dipole antennas, a compact style often used at 2.4 GHz—the same band used by WiFi and Bluetooth. By folding the arms of a classic dipole, engineers can shrink the antenna without sacrificing performance. However, this folding also makes its behavior very sensitive to the material on which it is printed. Small changes in the circuit board’s length, width, or thickness can noticeably shift the antenna’s operating frequency and how well it is matched to the electronics. Because these relationships are highly tangled and nonlinear, engineers usually rely on repeated, time‑consuming electromagnetic simulations to explore design options.

Teaching a Model to Imitate the Simulator

To avoid running thousands of full simulations for every new design, the authors build a "surrogate" model—a fast mathematical stand‑in that mimics the simulator’s output. They start by generating a large dataset of 1,000 different antenna designs, each created by systematically varying the substrate dimensions. For every design, a professional electromagnetic tool calculates two key performance measures: the resonance frequency (the main working channel of the antenna) and the minimum return loss, often written as S11, which indicates how efficiently the antenna draws power from the transmitter instead of reflecting it back.

Blending Fuzzy Rules with Learning and Swarm Search

At the heart of the framework is an Adaptive Neuro‑Fuzzy Inference System (ANFIS), which combines two ideas: fuzzy rules that resemble human “if–then” reasoning, and neural‑network‑style learning that tunes the numbers behind those rules. Rather than using standard gradient‑based training alone—which can get stuck in poor solutions—the authors test four different strategies for adjusting the ANFIS parameters: classic backpropagation, a hybrid backpropagation method, a genetic algorithm inspired by evolution, and Particle Swarm Optimization (PSO), which imitates a swarm of particles searching for the best position. Crucially, they use these optimization methods not to tweak the antenna’s shape, but to improve the internal settings of the ANFIS itself so it can better learn the link between geometry and performance.

What the Model Learns About Performance

The comparison shows that population‑based methods, which explore many candidate solutions in parallel, clearly outperform purely gradient‑based learning. Among them, the PSO‑trained ANFIS delivers the most accurate and stable predictions. For the test designs it has never seen before, the model predicts resonance frequency with an average error of only a few thousandths of a gigahertz and return loss with errors of just a few thousandths of a decibel. Statistical tests confirm that these gains are not just luck: the PSO‑enhanced model systematically tracks the detailed ups and downs of both frequency and S11 across a wide range of board sizes and thicknesses.

What This Means for Future Wireless Devices

In plain terms, the authors have built a highly accurate shortcut for antenna design. Instead of running a heavy simulator for every small change in board dimensions, engineers can use the trained ANFIS‑PSO model to get near‑instant predictions of both operating frequency and matching quality. This can dramatically speed up design exploration, reduce computing costs, and help optimize antennas for compact wireless gadgets in the 2.4 GHz band and beyond. The same strategy—using swarm‑optimized neuro‑fuzzy models as fast surrogates—could be extended to other complex radio components, supporting the rapid development of future WiFi, Bluetooth, IoT, and next‑generation communication systems.

Citation: Haznedar, B., Gençoğlan, D.N. & Haznedar, H. Multiple antenna performance parameters estimation of folded dipole antenna using Adaptive Neuro-Fuzzy Inference System trained with Particle Swarm Optimization. Sci Rep 16, 12214 (2026). https://doi.org/10.1038/s41598-026-41039-8

Keywords: folded dipole antenna, wireless design, neuro-fuzzy modeling, particle swarm optimization, antenna optimization