Robustness of a moving-bank multiple model adaptive algorithm for control of a flexible space structure

The robustness capabilities of moving-bank multiple model adaptive estimation and control algorithms are investigated for application to a flexible space-structure in which parameter uncertainties dictate a need for online adaptivity. For both estimation and control, the algorithms are sufficiently robust to allow a six-state dynamics design model to yield excellent performance against a 24-state truth model of the actual system. Particularly for control, where nonadaptive controllers were shown to lead readily to unstable closed-loop systems, the moving-bank MMAC (multiple model adaptive control) provided control that was essentially equivalent to that of an artificially informed benchmark controller. It is pointed out that the values of the measurement noise covariance play an important role in the performance potential of the moving-bank algorithms. A range of admissible measurement precisions exists beyond which the effective movement of the bank in parameter space is seriously impaired