27 February 2025 in IEEE Antennas and Wireless Propagation Letters
A Reinforcement Learning (RL)-based Image Mapping Topology (IMT) optimization approach is proposed. To enable RL to optimize antenna topology and increase the efficiency of the optimization process. Two crucial components, Stratified Sampling Balance (SSB) and IMT are introduced. SSB is utilized to balance different types of data samples. IMT is utilized to facilitate the interaction between RL and simulation software, it also pretrains the actor network within the RL algorithm. This methodology allows the RL agent to acquire knowledge from the interaction of the antenna topology with the environment. Subsequently, the agent outputs antenna topology with the maximum reward signal. Compared with other machine learning-based optimization methods, the proposed method significantly alleviates the issue of learning inefficiency and reduces the exploration time for feasible solutions. To validate the effectiveness of the optimization method, a microstrip antenna is optimized. The measurement results show that the impedance bandwidth is doubled while maintaining the realized gain with fewer simulation samples and less optimization time.
7 March 2017 in Electronics Letters
The design and results of a D‐band frequency doubler using two parallelly mounted Schottky dual‐diode chips are presented, which is considered as a unit in the doubler design. The doubler features four Schottky anodes closely located at a compact region, being no need for additional power dividing and combining circuit. The purpose is to verify the feasibility of a simple and compact hybrid microwave integrated circuit (HMIC) circuitry for frequency doubling with a higher power‐handling capability in this band, and an effective fast method for designing the matching networks of the doubler. With an input driving power of about 20 dBm at the E‐band and bias voltage of 0 V, it delivers an output power of 6 ± 1 dBm across the band of 140–152 GHz and 2–4.7 dBm over the bands of 152–160 and 135–140 GHz.
9 December 2024 in IEEE Antennas and Wireless Propagation Letters
A Simulation-Driven Surrogate Parallel Improved Adaptive Genetic Algorithm (SDS-IAGA) method is proposed. This method aims to improve the efficiency of topology optimization for the automated design of antenna. The optimization process involves two stages: initialization population screening and algorithm application. In the first stage, a coarse-mesh electromagnetic (EM) simulation model combined with current-driven search is utilized to provide a high-quality initial population. In the second stage, variable-fidelity surrogate and correction technology assist the IAGA in optimizing the antenna topology. During this stage, the IAGA uses new adaptive crossover and mutation operators based on nonlinear improvement to enhance the efficiency in reaching the target solution. To verify the efficacy of the proposed SDS-IAGA, the design task of a planar tri-band antenna with center frequencies at 2.45/3.5/5.8 GHz is completed. The experimental results demonstrate that, compared to AGA and IAGA, the SDS-IAGA enhances the optimization efficiency of antenna topology by 62.97% and 54.22%, respectively. Furthermore, compared to existing optimization methods, SDS-IAGA can complete the target design task with fewer full-wave EM simulations.
14 May 2024 in IEEE Antennas and Wireless Propagation Letters
This letter introduces a wide impedance bandwidth microstrip comb-line array antenna in the millimeter-wave band. To achieve wide impedance bandwidth performance, a seriesfed array (SFA) antenna consisting of three types of radiating structures is designed on a substrate utilizing printed circuit board (PCB) technology. Different from the traditional combline array, parasitic components are introduced in the vicinity of the radiating elements to obtain low sidelobe levels (SLL). The antenna is constructed and measured to confirm its viability. The measured impedance bandwidth (VSWR ≤ 2) is 30.02% (20.1–27.2 GHz) with satisfactory impedance matching. Meanwhile, the SLL of the comb antenna is about -22 dB at 26.6 GHz. Therefore, the proposed antenna boasts advantages including a simple manufacturing process with low fabrication cost, ease of integration, wide impedance bandwidth, and low sidelobes.