This paper considers performance criteria for the identification of sensor error models and the procedure for their calculation. These criteria are used to investigate the efficiency of the identification problem solution, depending on the initial data, and to carry out a comparative analysis of various suboptimal algorithms. The calculation procedure is based on an algorithm that solves the joint problem of hypothesis recognition and parameter estimation within the Bayesian approach. A performance analysis of the models traditionally used to describe errors of inertial sensors is given to illustrate the application of the procedure for the calculation of performance criteria.
Conditions of the gravimetric marine survey have a crucial influence on the gravity anomaly (GA) estimation. These conditions include sea state, speed and type of surveying vessel, and GA variability in the area under survey. It should be considered carefully to design proper models for vertical accelerations and useful signal used in filtering and smoothing algorithms. When models or their parameters in the filter differ from the actual ones, an additional error occurs in the GA estimate with respect to the optimal filter. In the paper mathematical simulation is used to analyze the dependence of GA estimation error from the survey conditions and thereby the potential accuracy of compensation for vertical accelerations. The additional GA filtering / smoothing error due to the uncertainty of the model parameters is studied as well.
The paper considers a problem of designing error models for navigation sensors the data of which are used in fusion algorithms. Within the Bayesian approach, an algorithm is proposed that allows not only estimation of model parameters, but also identification of its structure. As a result of solving the joint problem of hypotheses recognition and parameter estimation, a shaping filter describing sensor errors is designed. An example is given to illustrate the performance of the proposed algorithm.