Multipath Branch Model-Aided Differential Evolution Algorithm Based on Regional Error

Abstract

This paper introduces a general approach to multiobjective optimization design. In surrogate model-aided algorithms, the prediction accuracy of the surrogate model plays a key point in determining the optimization results. In order to overcome the defect of a single model in the adaptive antenna optimization problem and improving the efficiency and accuracy of the model, a method called multipath branching model-aided differential evolution algorithm analysis (MMDEA-RE) is proposed. The key innovations include: A novel combination model integrating Gaussian process, multilayer perceptron, and radial basis function. Compared with the traditional single model, the combination model can deal with complex and diverse optimization problems more effectively and has stronger robustness and versatility. A method of regional error analysis near the prediction point is proposed, which can select a suitable model with high prediction accuracy after a few iterations. The effectiveness and superiority of MMDEA-RE are demonstrated by the optimization of two linear array antennas and one 5G antenna. The contribution of this study is to improve the accuracy and generality of surrogate models in antenna design. This proposed method has practical implications and potential applications for improving antenna optimization techniques.