Optimal measurement scheduling for prediction and estimation

A general theory of optimal measurement scheduling for least-squares estimation is developed. The theory is based on the assumption that the cost of a measurement is inversely proportional to the variance of measurement noise, and that it is possible to distribute the total measurement cost arbitrarily among a set of measurements. The theory leads to a nonquadratic minimization problem. An effective algorithm for solving this problem is developed. The theory is applied to the prediction of a discrete-time integrated Wiener process from noisy past samples.<
>