Complex Neural Network based Joint AoA and AoD Estimation for Bistatic ISAC

TL;DR

This paper introduces two innovative methods, including a deep learning approach, for joint AoA and AoD estimation in bistatic ISAC systems, enhancing efficiency and maintaining high accuracy.

Contribution

The paper presents a novel deep learning-based method and a parameterized algorithm for joint AoA and AoD estimation in bistatic ISAC, improving computational efficiency and performance.

Findings

DL approach achieves comparable accuracy to the parameterized method.

DL method significantly reduces computational complexity.

Preprocessing with coarse timing estimation enhances efficiency.

Abstract

Integrated sensing and communication (ISAC) in wireless systems has emerged as a promising paradigm, offering the potential for improved performance, efficient resource utilization, and mutually beneficial interactions between radar sensing and wireless communications, thereby shaping the future of wireless technologies. In this work, we present two novel methods to address the joint angle of arrival and angle of departure estimation problem for bistatic ISAC systems. Our proposed methods consist of a deep learning (DL) solution leveraging complex neural networks, in addition to a parameterized algorithm. By exploiting the estimated channel matrix and incorporating a preprocessing step consisting of a coarse timing estimation, we are able to notably reduce the input size and improve the computational efficiency. In our findings, we emphasize the remarkable potential of our DL-based approach, which demonstrates comparable performance to the parameterized method that explicitly exploits the multiple-input multiple-output (MIMO) model, while exhibiting significantly lower computational complexity.