Global Robustness Characterization of DOA Estimators by Breakdown Detection

We characterize the global robustness of direction of arrival (DOA) estimators against additive outliers in the sensor array data by detection of the estimator's breakdown. The proposed method contaminates a given proportion of the sensor array data and observes the corresponding DOA estimate. An oracle compares the DOA estimation error magnitude with a chosen tolerance. The oracle detects that the DOA estimate has broken down when the magnitude of error exceeds the tolerance. The estimator's robustness is thus globally characterized in a robustness operating characteristic (ROC) curve which shows the true positive rate of detected breakdowns versus the false positive rate. The ROC curve is evaluated for several DOA estimators: conventional, phase-only, and weighted phase-only beamformer, as well as least absolute deviation based DOA estimator.