Robustness of the likelihood ratio detector for moderately fluctuating radar targets

The performance of the likelihood ratio detector (LRD) for moderately fluctuating targets (i.e. intermediate between Swerling II and Swerling I models) is evaluated for the case of poor knowledge of the target characteristics. In particular, the influence of incorrect assumptions in the design phase about the signal-to-interference ratio (SIR), the target correlation coefficient, the shape of the target autocorrelation function, and the target mean Doppler frequency is analysed by computing the detectability loss due to mismatching between the design hypothesis and the real target characteristics. It is shown that the LRD is extremely robust to an inaccurate choice of the design SIR. The loss due to correlation mismatching depends not only on the absolute value of the difference between the design value of the correlation coefficient and the real value, but also on the sign of this difference, with larger losses occurring when the design value is larger than the real one. A sensible criterion is given for the choice of the target correlation coefficient to be used in the design process. Also, the performance degradation due to Doppler mismatching can be large and calls therefore for the deployment of a bank of LRDs, each designed for a different value of the target Doppler frequency. The minimum number of LRDs needed to maintain acceptable performance under Doppler mismatching is computed.