The robustness of virtual-ESPRIT against model errors

Virtual-ESPRIT (VESPA) imposes a slight constraint on the array i.e. only a pair of identical sensors are required. We follow the classical procedure to define the sensitivity of VESPA with respect to each model parameter, and derive closed forms for the sensitivity formulas. We justify our derivations by comparing the theoretical formulas with the results from Monte Carlo simulations. Additionally, we demonstrate, by simulations, that VESPA is more robust to model mismatches than ESPRIT in most situations.