Abstract

Minimum variance control (residual phase variance) of AO systems under actuator saturation is shown to be the combination of minimum variance prediction and constrained projection onto the DM space. Degradation in performance resulting from saturation is investigated for adaptive optics LQG controllers, which naturally combine (in a saturation-free problem) minimum variance prediction and optimal projection onto the DM space. While optimal prediction can be maintained at negligible cost, computing the optimal projection would require the solution in real time of a constrained optimization problem. Alternatively, simulations conducted on a model of an 8 m. telescope suggest that a linear regularized projection adapted to the saturation threshold can yield a significant increase in performance.

© 2007 Optical Society of America