Hybrid ANN-GA optimization method for minimizing the coupling in MIMO antennas

In this paper, we propose a novel approach to address the challenge posed by large datasets in the context of antenna optimization through machine learning (ML). The approach involves a modified artificial neural network (ANN) technique. To tackle the issue of training with limited data, we introduce a method that leverages prior knowledge to create initial small datasets. Throughout the training process, these datasets are iteratively enhanced by incorporating new simulation data obtained from optimization results. In the applications presented here, ANN is used to establish a relationship model between antenna geometric variables and transmission co-efficients (S21). Subsequently, a genetic algorithm (GA) is used to optimize the decoupling of the antenna array. By implementing wideband and narrowband optimization of the S21 between two antennas, the accuracy and efficiency of this method are verified. The initial design and optimized antennas were fabricated and measured. In summary, this proposed method not only can effectively mitigate the expenses associated with design but also generates antenna optimization solutions of high quality.