A Study of Direction of Arrival Methods Based on Antenna Arrays in Presence of Model Errors.

• Main Author: Sjödin, Julia
• Format: Bachelor Thesis
• Language: English
• Published: Uppsala universitet, Signaler och system 2022
• Subjects: Direction of Arrival; DOA; WSF; MUSIC; TLS-ESPRIT; Extended Kalman filter; model error; Signal Processing; Signalbehandling; DML
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-488858

Methods for Direction of Arrival, DOA estimation of multiple objects based on phased arrayantenna technology have many advantages in for example electronic warfare and radarapplications. However, perfect calibration of an antenna array can seldom be achieved. Thepurpose of this report is to study different methods for DOA estimation and how calibration-/modelerrors affect the results. Possible methods for quantifying these kinds of errors using measurement data are suggested. This thesis consists of essentially five parts. The different studies have been carried out using MATLAB simulations as well as theoretical considerations, i.e., calculations. In the first study, examples of the possible performance of four DOA algorithms, MUSIC, TLS-ESPRIT, WSF, and DML are provided. Results are given both with and without applying spatial smoothing. The latter scheme is used for handling correlated, or even coherent, sources. The results show that, for the considered scenarios, MUSIC performs the most consistently well, while the performance of DML is inferior. ESPRIT is well-performing when spatial smoothing is applied and performs the best when the angles of two signals are very close. It has been observed that WSF with weighting matrices for optimal asymptotic performance as well as spatial smoothing applied doesn’t perform well. When applying model errors to the systemin the second study, the corresponding conclusions about the algorithms can be drawn. That separation distance between the angles and that higher SNR results in better estimates are also confirmed. Quantification of certain array errors is also considered using methods inspired by a scheme proposed in the context of nonlinear system identification. The results show that the DOA algorithms are very good at dealing with noise and that the attempted method works well when the model error is like the true signals, but different enough that it is not confused with a problem with more signals. The model error that results in the worst results is when it only ...