A pixelated circularly polarized dielectric resonator antenna
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-03
- Publication Date
- 2026-08-11
AI Technical Summary
其中，切角和开槽方法结构相对简单，但其扰动形式较为固定，主要依赖少数几何参数对谐振模式进行调节，设计自由度有限；加载扰动结构虽然能够改善模式耦合和相位关系，但往往会增加结构复杂度，并可能引入额外损耗或加工误差；双馈电网络能够较为直接地实现正交模式激励和 90° 相位差控制，但会增加馈电系统复杂度，不利于天线的小型化和集成化设计
本发明提供了一种像素化圆极化介质谐振器天线，通过将介质谐振器划分为多个可独立调控的像素单元，并利用介质与空气的不同分布构建复杂拓扑结构，突破了传统圆极化介质谐振器天线依赖切角、开槽或加载单一扰动结构的设计限制，显著提高了天线结构设计自由度。与传统参数化圆极化介质谐振器天线相比，本发明能够充分利用介质体内部空间的电磁调控能力，通过像素化介质分布改变天线内部电场分布和谐振模式耦合关系，使多个相邻谐振模式在目标频段内有效融合，从而改善天线的阻抗带宽和轴比带宽。在设计过程中引入 CNN-ResNet 代理模型和 NSGA-II 多目标优化算法，能够快速建立像素化拓扑结构与天线电磁响应之间的非线性映射关系，显著减少全波电磁仿真次数，提高优化效率。该方法能够同时考虑反射系数和轴比两个关键指标，避免传统设计中阻抗带宽与轴比带宽难以同时优化的问题。为宽带圆极化介质谐振器天线提供了一种新的设计思路和实现方式，具有较高的工程应用价值和推广潜力。
Smart Images
Figure CN122552802A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a pixelated circularly polarized dielectric resonator antenna, belonging to the field of antenna technology. Background Technology
[0002] Antennas are key components in wireless communication systems, and their radiation performance, polarization characteristics, and impedance matching directly affect communication quality and system stability. With the rapid development of wireless communication technology, communication systems place higher demands on antenna miniaturization, broadband capabilities, and polarization stability. Circularly polarized antennas can effectively reduce signal fading caused by attitude changes or polarization mismatch between transmitting and receiving antennas and have strong resistance to multipath interference. Therefore, they have significant application value in satellite communication, navigation and positioning, radar detection, and mobile communication. A dielectric resonator antenna is a type of antenna that radiates by utilizing the resonant modes within a dielectric block. Compared to traditional microstrip antennas, dielectric resonator antennas offer advantages such as low conductor loss, high radiation efficiency, wide operating bandwidth, and flexible feeding methods, making them particularly suitable for higher-frequency wireless communication systems. Furthermore, the resonant modes of a dielectric resonator antenna can be tuned by altering the dielectric size, shape, dielectric constant, and feeding structure, thereby achieving various radiation characteristics such as broadband, circular polarization, and multi-frequency operation. In the design of circularly polarized dielectric resonator antennas, it is typically necessary to excite two orthogonal resonant modes with similar amplitudes and a phase difference of approximately 90° to enable the antenna to achieve stable circularly polarized radiation within the target frequency band. Traditional methods mainly achieve circular polarization through chamfering, slotting, adding perturbation structures, or using dual-feed networks. Among these, chamfering and slotting methods have relatively simple structures, but their perturbation forms are relatively fixed, mainly relying on a few geometric parameters to adjust the resonant modes, resulting in limited design freedom. While adding perturbation structures can improve mode coupling and phase relationships, it often increases structural complexity and may introduce additional losses or manufacturing errors. Dual-feed networks can directly achieve orthogonal mode excitation and 90° phase difference control, but they increase the complexity of the feeding system, which is not conducive to the miniaturization and integration of the antenna design. Traditional circularly polarized dielectric resonator antennas typically only adjust the external geometry or feed structure, making it difficult to fully utilize the electromagnetic control capabilities of the dielectric's internal space. Since the internal electric field distribution of a dielectric resonator is closely related to its resonant modes, relying solely on local structural perturbations such as chamfering or slotting often fails to simultaneously achieve impedance matching and axial ratio performance over a wide frequency band. Particularly in broadband circularly polarized designs, the antenna needs to maintain amplitude balance and a stable phase difference between two orthogonal modes over a broad frequency range. This necessitates repeated adjustments to multiple structural parameters using traditional methods, resulting in a complex optimization process and a tendency for impedance bandwidth and axial ratio bandwidth to misalign. Therefore, traditional design methods still have limitations in terms of design flexibility, internal field control capabilities, and broadband circularly polarized performance optimization. Summary of the Invention
[0003] In view of this, the object of the present invention is to provide a pixelated circularly polarized dielectric resonator antenna. To achieve the above objectives, the technical solution of the present invention is as follows. A pixelated circularly polarized dielectric resonator antenna includes a dielectric resonator, a dielectric substrate, a ground plane, a cross-shaped slot, and a microstrip feed structure. The dielectric resonator is disposed above the dielectric substrate, and the ground plane is disposed between the dielectric substrate and the dielectric resonator. The ground plane is etched with a cross-shaped slot for coupling the feed. The microstrip feed line is printed on the bottom surface of the dielectric substrate. Electromagnetic energy is coupled to the dielectric resonator through the cross-shaped slot, thereby exciting the circularly polarized radiation mode. Furthermore, the dielectric resonator adopts a pixelated structure, dividing the design area of the dielectric resonator into N×N square pixel units. Each pixel unit is represented by a binary variable, where 1 indicates that the dielectric is filled and 0 indicates that the air is filled. By adjusting the spatial distribution of the dielectric pixels and the air pixels, the equivalent dielectric constant distribution, electric field distribution, and mode coupling relationship inside the dielectric resonator are changed, thereby achieving coordinated control of the antenna impedance matching performance and circular polarization performance. Furthermore, using the antenna's reflection coefficient and axial ratio within the target frequency band as optimization objectives, a surrogate model with a CNN-ResNet structure is employed to establish the mapping relationship between the pixelated dielectric resonator topology and its electromagnetic response. First, the binary matrix of the pixelated dielectric resonator is used as input, and local geometric and electromagnetic coupling features are extracted through a convolutional module. Then, a residual structure is introduced to fuse shallow and deep features, enhancing the model's ability to represent complex nonlinear electromagnetic responses. Finally, the electromagnetic response within the target frequency band is output through a fully connected layer. This electromagnetic response includes the reflection coefficient S11 and axial ratio AR, enabling rapid prediction of the broadband electromagnetic performance of the pixelated circularly polarized dielectric resonator antenna through the surrogate model. Furthermore, after the surrogate model is trained, the non-dominated sorting genetic algorithm NSGA-II is used to perform multi-objective optimization of the pixelated dielectric resonator topology. During the optimization process, impedance matching performance and circular polarization performance are used as objective functions to select pixelated dielectric resonator structures that simultaneously have a wide impedance bandwidth and a wide axial ratio bandwidth. Furthermore, the dielectric / air material distribution map of N×N pixel blocks is obtained by the NSGA-II optimization algorithm. The pixel blocks are 3mm in length and width, 5mm in height, have a dielectric constant of 9.8, and N is 13. Beneficial effects: This invention provides a pixelated circularly polarized dielectric resonator antenna. By dividing the dielectric resonator into multiple independently controllable pixel units and utilizing the different distributions of the dielectric and air to construct a complex topology, it overcomes the design limitations of traditional circularly polarized dielectric resonator antennas that rely on chamfering, slotting, or loading a single perturbation structure, significantly improving the design freedom of the antenna structure. Compared with traditional parametric circularly polarized dielectric resonator antennas, this invention can fully utilize the electromagnetic control capability of the internal space of the dielectric. By changing the internal electric field distribution and the coupling relationship of resonant modes through pixelated dielectric distribution, multiple adjacent resonant modes can be effectively fused within the target frequency band, thereby improving the antenna's impedance bandwidth and axial ratio bandwidth. The introduction of a CNN-ResNet surrogate model and the NSGA-II multi-objective optimization algorithm during the design process enables the rapid establishment of a nonlinear mapping relationship between the pixelated topology and the antenna's electromagnetic response, significantly reducing the number of full-wave electromagnetic simulations and improving optimization efficiency. This method can simultaneously consider the two key indicators of reflection coefficient and axial ratio, avoiding the problem of simultaneously optimizing impedance bandwidth and axial ratio bandwidth in traditional designs. This study provides a new design concept and implementation method for broadband circularly polarized dielectric resonator antennas, which has high engineering application value and promotion potential. Attached Figure Description Figure 1 This is a schematic diagram of the overall structure of an embodiment of the present invention. Figure 2 This is a schematic diagram of the pixelated dielectric resonator structure according to an embodiment of the present invention. Figure 3 This is a material distribution map of the optimized pixel block according to an embodiment of the present invention. Figure 4 This is the S11 impedance matching curve of an embodiment of the present invention. Figure 5 This is the 3dB axis ratio curve of an embodiment of the present invention. Figure 6 This is a normalized radiation pattern according to an embodiment of the present invention. Figure 7-10 This is a diagram illustrating the construction and optimization process of the neural network in an embodiment of the present invention. Detailed Implementation The present invention will be further described in detail below with reference to specific embodiments. Example 1 like Figure 1-2As shown, a pixelated circularly polarized dielectric resonator antenna includes a pixelated dielectric resonator, a dielectric substrate, a ground plane, a cross-shaped slot, and a microstrip feed structure. The pixelated dielectric resonator is disposed above the dielectric substrate, and the ground plane is disposed between the dielectric substrate and the pixelated dielectric resonator. Orthogonal cross-shaped slots for coupling the feed are etched on the ground plane. Microstrip feed lines are printed on the bottom surface of the dielectric substrate. Electromagnetic energy is transmitted through the microstrip feed lines and coupled to the pixelated dielectric resonator through the cross-shaped slots, thereby exciting the circularly polarized radiation mode. The pixelated dielectric resonator consists of N×N pixel units, where N is a positive integer. Each pixel unit is represented by a binary variable, where 1 indicates that the pixel unit is filled with dielectric material, and 0 indicates that the pixel unit is an air region. By changing the distribution of dielectric material and air in pixel units at different locations, the equivalent dielectric constant distribution, electric field distribution, and resonant mode coupling relationship inside the dielectric resonator can be controlled, thereby achieving synergistic optimization of antenna impedance matching performance and circular polarization performance. To train the subsequent proxy model, it is necessary to collect curves showing the S11 impedance and axial ratio as a function of frequency under different dielectric material and air distributions of pixel units at different locations. This data will be used to train the neural network and construct a positive prediction model from material distribution to electromagnetic response. Using the S11 impedance bandwidth and axial ratio at the target frequency band as optimization objectives, the constructed neural network model optimizes the material distribution of pixel units at different locations. This alters the internal electric field distribution and resonant mode coupling relationship of the antenna, effectively fusing multiple adjacent resonant modes within the target frequency band, thereby improving the antenna's impedance bandwidth and axial ratio bandwidth. In this embodiment, the design area of the pixelated dielectric resonator is divided into 13×13 square pixel units, which serve as the input to the neural network model. Each pixel unit has a length and width of 3mm and a height of 5mm. The pixel unit represented by 1 uses a dielectric material with a dielectric constant of 9.8, and the dielectric substrate uses a material with a dielectric constant of 2.65. The target frequency range is set to 5 GHz to 8 GHz, and sampling is performed in steps of 0.05 GHz, resulting in a total of 61 discrete frequency points. The S11 impedance matching curve in the target frequency band of the above example was analyzed using the commercial simulation software HFSS_2021. The results are as follows: Figure 4 As shown. The 3dB axial ratio curve of the target frequency band in the above example was analyzed using the commercial simulation software HFSS_2021, and the results are as follows: Figure 5 As shown. The normalized radiation pattern of the target frequency band in the above example was analyzed using the commercial simulation software HFSS_2021. The results are as follows: Figure 6 As shown. Reference Figure 4After optimization of the above scheme, in this embodiment, within the frequency range of 5GHz to 8GHz, S11 < -10dB is satisfied in 5.47GHz-5.64GHz and 5.83GHz-6.62GHz, and the impedance bandwidth reaches 19% (5.47GHz-6.62GHz). Reference Figure 5 After optimization of the above scheme, in this embodiment, within the frequency range of 5GHz to 8GHz, the AR < 3dB is achieved in the 5.53GHz-7.17GHz range, and the 3dB axial ratio bandwidth reaches 25.8%. Reference Figure 6 In this embodiment, the radiation patterns are all wide-beam radiation patterns. like Figure 7-10 As shown, the specific process of constructing and optimizing the neural network in this embodiment is as follows: The constructed neural network is a CNN convolutional neural network with residual connections. It takes the material distribution of a 13×13 pixel block as input and outputs the S11 impedance and axial ratio within the frequency band. The network consists of an input layer, multiple convolutional layers, group normalization, dropout layers, and residual connections. The neural network was trained on the preprocessed dataset (1900 groups) as input. The hyperparameters of the neural network were adjusted to achieve optimal training results. The Huber loss function was used to evaluate the model's prediction results. The total loss function consists of a weighted sum of the S11 loss term and the AR loss term, where L_S11 and L_AR represent the Huber loss functions corresponding to the reflection coefficient and axial ratio, respectively, and w_S11 and w_AR represent the weight coefficients of the reflection coefficient and axial ratio, respectively. In this embodiment, w_S11 is set to 1 and w_AR is set to 2 to enhance the model's learning ability for circular polarization performance and improve the prediction accuracy of the axial ratio response. To prevent overfitting of the surrogate model, an early stopping mechanism is introduced during training. Model training is terminated when the validation set loss does not improve for 120 consecutive epochs. Simultaneously, an adaptive learning rate decay strategy is adopted; when the validation set loss does not decrease for 30 consecutive epochs, the learning rate is reduced to 0.5 times its original value. The model triggered an early stopping mechanism on the 340th iteration, with training set loss and validation set loss of 1.69 and 1.47, respectively. Optimization process: Based on a well-trained neural network surrogate model combined with the NSGA-II optimization algorithm, multi-objective optimization of the pixelated dielectric resonator topology is performed. During the optimization process, the trained surrogate model is embedded into the NSGA-II optimization algorithm to quickly evaluate the fitness of each candidate structure. The S11 and AR predicted by the surrogate model are used as the fitness evaluation criteria, with impedance matching performance and circular polarization performance as optimization objectives. Through non-dominated sorting and crowding distance mechanisms, the Pareto optimal solution set is obtained while ensuring population diversity. The optimized candidate structures are then re-imported into full-wave electromagnetic simulation software for verification, completing the antenna design. In summary, the invention includes, but is not limited to, the above embodiments. Any equivalent substitutions or partial improvements made under the spirit and principles of this invention shall be considered to be within the protection scope of this invention.
Claims
1. A pixelated circularly polarized dielectric resonator antenna, characterized in that: This includes pixelated dielectric resonators, dielectric substrates, ground planes, cross-shaped slots, and microstrip feeding structures. The pixelated dielectric resonator is disposed above the dielectric substrate, and the ground plane is disposed between the dielectric substrate and the pixelated dielectric resonator. The ground plane is etched with a cross-shaped slot for coupling the feed. The microstrip feed line is printed on the bottom surface of the dielectric substrate. Electromagnetic energy is transmitted through the microstrip feed line and coupled to the pixelated dielectric resonator through the cross-shaped slot, thereby exciting the circularly polarized radiation mode. The pixelated dielectric resonator consists of N×N pixel units, where N is a positive integer. Each pixel unit is represented by a binary variable, where 1 indicates that the pixel unit is filled with dielectric material and 0 indicates that the pixel unit is an air region. Using the reflection coefficient S11 and axial ratio AR within the target frequency band as optimization targets, a deep learning-based optimization method is employed to determine the spatial distribution of the medium material and air region in N×N pixel units.
2. The pixelated circularly polarized dielectric resonator antenna as described in claim 1, characterized in that: By adjusting the spatial distribution of dielectric pixels and air pixels in a pixelated dielectric resonator, the equivalent dielectric constant distribution, electric field distribution, and resonant mode coupling relationship inside the dielectric resonator are changed, enabling multiple adjacent resonant modes to be effectively integrated within the target frequency band, thereby achieving coordinated control of antenna impedance matching performance and circular polarization performance.
3. The pixelated circularly polarized dielectric resonator antenna as described in claim 1, characterized in that: Using the reflection coefficient S11 and axial ratio AR in the target frequency band as optimization objectives, a surrogate model with a CNN-ResNet structure is used to establish the mapping relationship between the pixelated dielectric resonator topology and the electromagnetic response; The pixelated dielectric resonator topology is a binary matrix corresponding to N×N pixel units, and the electromagnetic response includes the reflection coefficient S11 curve and the axial ratio AR curve within the target frequency band.
4. A pixelated circularly polarized dielectric resonator antenna as described in claim 3, characterized in that: The surrogate model of the CNN-ResNet structure takes the binary matrix of the pixelated dielectric resonator as input and extracts local geometric features and electromagnetic coupling features through the convolution module; Subsequently, a residual connection structure is introduced to fuse shallow and deep features, thereby enhancing the model's ability to represent complex nonlinear electromagnetic responses. Finally, the reflection coefficient S11 and axial ratio AR in the target frequency band are output through a fully connected layer.
5. A pixelated circularly polarized dielectric resonator antenna as described in claim 3, characterized in that: The surrogate model uses the Huber loss function to evaluate the model's prediction results. The total loss function is composed of a weighted average of the reflection coefficient S11 loss term and the axial ratio AR loss term, and its expression is as follows: ； Among them, L S11 and L AR Let w represent the Huber loss function corresponding to the reflection coefficient and axial ratio, respectively. S11 and w AR These represent the weighting coefficients corresponding to the reflection coefficient and the axial ratio, respectively.
6. A pixelated circularly polarized dielectric resonator antenna as described in claim 3, characterized in that: After the CNN-ResNet surrogate model is trained, the non-dominated sorting genetic algorithm NSGA-II is used to optimize the pixelated dielectric resonator topology. During the optimization process, the reflection coefficient S11 and axial ratio AR predicted by the surrogate model are used as the fitness evaluation criteria, and the impedance matching performance and circular polarization performance are used as optimization objectives. The Pareto optimal solution set is obtained through non-dominated sorting and crowding distance mechanism, thereby selecting the pixelated dielectric resonator structure with both wide impedance bandwidth and wide axial ratio bandwidth.
7. The pixelated circularly polarized dielectric resonator antenna as described in claim 6, wherein the objective function for multi-objective optimization is defined as: ； ； in, N represents the total number of frequency sampling points, AR(f i ) and S11(f i ) represent the axial ratio and reflection coefficient predicted by the surrogate model at the i-th frequency point, respectively; by minimizing F AR and F S11 This enables the antenna to achieve better circular polarization and impedance matching performance within the target frequency band.
8. A pixelated circularly polarized dielectric resonator antenna as described in claim 1, characterized in that: The pixelated dielectric resonator is designed with a 13×13 square pixel unit, each pixel unit having a length and width of 3 mm and a height of 5 mm; the pixel unit represented by 1 uses a dielectric material with a dielectric constant of 9.8, and the dielectric substrate uses a material with a dielectric constant of 2.65.