Deep Learning-Enabled Angle Estimation in Bistatic ISAC Systems

Integrated sensing and communication (ISAC) is becoming a vital technology for future wireless systems and is particularly relevant for many applications requiring both high-performance sensing and wireless communications. Our work presents a novel algorithm based on deep learning for estimating angles of arrival and angles of departure in bistatic ISAC systems, exploiting orthogonal frequency division multiplexing signals initially designed for communication purposes. Our proposed method incorporates a complex-valued neural network that takes advantage of the estimated channel matrix, along with a preprocessing step for coarse timing estimation. Comparative analysis with an adapted version of the multiple signal classification method demonstrates the effectiveness of our approach, showcasing remarkable performance in terms of mean squared error while requiring lower computational complexity.