To improve the response to unexpected thermally induced disturbances,
two new controllers are designed for the Hubble Space Telescope (HST)
using covariance control techniques. The first controller minimizes th
e required control effort subject to inequality constraints on the out
put covariance matrix. The second controller is designed to satisfy bo
th output covariance constraints and controller covariance constraints
. The importance of the controller covariance constraint is to properl
y scale the controller for digital implementation in the control compu
ter using fixed-point arithmetic. We provide a technique to integrate
modeling, control design, and signal processing (in HST's fixed-point
control computer), since these problems are not separable. Covariance
control can easily ac commodate the roundoff error and computational t
ime delay. Compared with the existing HST controller design, the resul
ts show that the required pointing error specifications can be achieve
d with 85% less control effort and that the error due to the finite-wo
rdlength implementation of the controller can easily be included in th
e optimal control design process.