A variance-reduced simulation for benchmarking CFAR detectors in non-homogeneous environments

Abstract

This paper presents an accurate and stable simulation framework for evaluating the detection performance of constant false alarm rate (CFAR) detectors in the presence of clutter non-homogeneity and interference. The proposed framework incorporates theoretically consistent threshold calibration for CFAR detectors and introduces a variance-reduced estimator for false-alarm probability based on conditional expectation. This approach enables reliable estimation of extremely small false-alarm probabilities without excessive computational cost. Clutter edge scenarios are modeled using an edge-aware formulation that ensures physical consistency between the reference window and the cell under test. Extensive numerical results validate the proposed framework under homogeneous clutter and demonstrate its effectiveness in analyzing detector robustness in the presence of multiple interfering targets and clutter edges. The results provide clear insight into the relative strengths and limitations of different CFAR detectors and highlight the importance of accurate simulation techniques for meaningful performance comparison in realistic radar environments.