Robust bearing estimation in the presence of direction-dependent modelling errors: identifiability and treatment

This paper presents identifiability and treatment relative to bearing estimation in presence of modelling errors. It introduces the general case of direction-dependent modelling errors. The classical direction-independent case is only a particular case which takes into account a prior knowledge. This general case introduces the important issue of simultaneous sources and perturbation identifiability which is analysed. A new self-calibration technique based on MUSIC algorithm and which is able to treat direction-dependent errors is also proposed. For the purpose of regularization, a “cost term” is introduced in this algorithm. It confers good robustness to the algorithm which usually fails in the presence of great gap between the model and reality. After reduction of the new multidimensional “increased function”, values of azimuths are easily obtained on a monodimensional spectrum. Some simulations support results and verify the improvements expected from the theory