Moving-Bank Multiple Model Adaptive Estimation and Control Applied to a Large Flexible Space Structure.

摘要

The performance of moving-bank multiple model adaptive estimation (MMAE) and control (MMAC) algorithms for large space structure control is analyzed in this thesis. The performance of a six-state filter model and associated controller are evaluated on the basis of estimation/control performance against a 24-estate truth model. A model developed using finite element analysis is used to approximate a large flexible space structure. The space structure is configured as a two-bay truss which is attached to a large central hub, where the mass of the hub is considered to be much more larger than the mass of the flexible structure. The model is developed in physical coordinates and then transformed into modal coordinates, where the method of singular perturbations is used to obtain a reduced order filter model. The actual positions and velocities of various physical points on the structure are used in the evaluation of the moving-bank algorithm performance. Results of the research indicate that appropriate determination of the filter model noise statistics as well as the LQG controller weighting matrices significantly improve performance of the bank throughout the parameter space. The results indicate that the performance of the moving-bank algorithms is seriously degraded by the inclusion of the filter-computed residual covariance in the conditional probability density function for computation of the hypothesis conditional probabilities within the multiple model algorithms.