Knowledge Distillation Based Deep Learning Model for User Equipment Positioning in Massive MIMO Systems Using Flying Reconfigurable Intelligent Surfaces

This paper introduces an innovative deep learning-based approach for User Equipment positioning in massive MIMO systems, enhanced by Flying Reconfigurable Intelligent Surfaces. The study emphasizes the transformative impact of Massive MIMO technology, which substantially increases the capacity and efficiency of wireless networks through extensive antenna arrays. We present a knowledge distillation model where a comprehensive teacher model is trained on massive MIMO data to predict User Equipment positions, which in turn guides a more compact student model. This results in a system that not only retains high accuracy with a lower mean absolute error of 3.67m but also benefits from reduced complexity and resource requirements. Our findings reveal that the distilled student model achieves a significant reduction in computational load while maintaining precise positioning capabilities, showcasing the potential for practical deployment in future wireless communication systems.