Published in IET Radar, Sonar and Navigation Received on 4th November 2008 Revised on 3rd March 2009 doi: 10.1049/iet-rsn.2008.0178 ISSN 1751-8784 Direction-of-arrival estimation in radar systems: moving window against approximate maximum likelihood estimator M. Greco 1 F. Gini 1 A. Farina 2 L. Timmoneri 2 1 Department of ‘Ingegneria dell’Informazione’, via G. Caruso 16, Pisa 56122, Italy 2 SELEX-SI, via Tiburtina, Roma 00134, Italy E-mail: m.greco@iet.unipi.it Abstract: In this study, we compare two radar target direction-of-arrival (DOA) estimation algorithms: the classical moving window (MW) estimator, implemented in many real radar systems and the approximate maximum likelihood (AML) estimator. The first technique exploits multiple detections in the same time-on-target and the second one exploits the fact that the radar antenna mechanical scanning impresses an amplitude modulation on the signals backscattered by the target. Performances of the two estimators are numerically investigated through Monte Carlo simulation and compared in terms of root-mean-square-error (RMSE), probability of detection and probability of target splitting, the latter being defined as the probability of detecting more than one target when instead only one is present in the cell under test. Numerical results show that the AML estimator generally outperforms the classical MW estimator, also in terms of robustness to target splitting. 1 Introduction The estimation of target direction of arrival (DOA) is not a new problem in the radar literature. It shows up both in tracking and search radar systems. One of the most common techniques is the monopulse technique [1], which in principle can work with just a single pulse and uses two tightly matched receiving channels (per angle): the sum and the difference channels. In this work, we compare two algorithms that process multiple pulses and require only one receiving channel, the moving window (MW) [1], often implemented in commercial radars, and the more modern estimator based on the maximum likelihood (ML) principle [2, 3]. The aim here is to compare advantages and disadvantages of the two techniques. The monopulse technique estimates the azimuth DOA exploiting multiple detections in the same time-on-target (ToT), whereas the second technique exploits knowledge of the antenna main beam pattern and the fact that the antenna mechanical scanning impresses an amplitude modulation on the signals backscattered by the target [2]. The rest of this paper is organised as follows. In Section 2, the data model and the problem statement are introduced and explained. In Sections 3 and 4, the MW and the approximate maximum likelihood (AML) estimators are described, respectively. In Section 5, the numerical results are reported for suitable study cases and some conclusions are drawn. The results show that the AML estimator generally outperforms the classical MW estimator, in terms of RMSE, probability of detection and robustness to target splitting. 2 Data model and problem statement Consider a radar antenna that rotates mechanically with constant angular velocity v R (rad/s) and denote by h(u) the one-way antenna beam pattern, by G 0 the maximum gain and by u B the 23 dB azimuth beam width, that is, the angle such that h 2 ( + u B =2) ¼ G 0 =2. The number N of pulses collected during the ToT by the radar within the 23 dB beam width is given by N ¼ u B =(v R T ), where T ¼ 1/PRF is the radar pulse repetition time (PRT) and PRF is the pulse repetition frequency. The number N of 552 IET Radar Sonar Navig., 2009, Vol. 3, Iss. 5, pp. 552–557 & The Institution of Engineering and Technology 2009 doi: 10.1049/iet-rsn.2008.0178 www.ietdl.org