Selecting measurements and controls in LQG problems

Procedures are given for the efficient determination of optimal inputs and outputs of a linear system driven by white noise. Beginning with a large admissible set of inputs and outputs of a linear dynamic system with a quadratic performance metric, an algorithm is given to reduce the number of inputs or outputs to a specified number so as to make the smallest perturbation in the performance metric. Such techniques are useful in selecting and locating sensors and actuators.