SPATIO-TEMPORAL PREDICTION FOR ADAPTIVE OPTICS WAVEFRONT RECONSTRUCTORS

By taking advantage of the spatial and temporal correlation of the phase of the atmospherically-aberrated optical wavefront, we show in extensive computer simulations that the effect of the time delay in the servo loop of an adaptive optics system can be greatly reduced. Further work based on open-loop Shack-Hartmann sensor data from a 1.6-m telescope confirms the results of the simulations. Most of the work so far has explored linear algorithms which predict the output from the wavefront sensor based on immediate past history, although an investigation recently begun into the use of artificial neural networks holds promise for greater robustness in the low signal-to-noise regime, and offers the possibility of continuous on-line training, which can keep the network up to date on changing atmospheric statistics. In addition, in the linear case, we have computed predictors which attempt to track the changing phase during an individual wavefront sensor integration. Substantial improvement in the residual phase error caused by temporal decorrelation is obtained, particularly as the wavefront sensor integration time approaches the atmospheric decorrelation time τ0, which is encouraging for adaptive optics systems pushing towards shorter wavelengths.