Efficient and Reconfigurable Directional Beam Steering in Phased Arrays using AI and Edge Computing

This paper presents an Artificial Intelligence (AI) based method for controlling the radiation pattern of phased arrays. The method employs a feed-forward Artificial Neural Network (ANN) that is trained to steer the beam towards a desired direction by changing the radiation pattern. The ANN acts as a function that takes the pointing direction as input and returns the corresponding phase shift matrix as output for each radiating element. To ensure efficiency in terms of computational complexity and time response, specific layers are extracted from the level curve of the array factor at -3dB before training the neural network by assigning the connection weights. This approach achieves a balanced trade-off between the number of phase-shifting processes and the spatial resolution, which is crucial in contexts such as IoT and 5G. The proposed AI-based method has been successfully tested and verified using an AI-oriented Microcontroller for Edge Computing applications, where a specific neural network is implemented and used to compute phase matrices of a 4x4 phased array for a number of pointing directions obtained through data fusion of IMU data.