Ucinski D. Measurement Optimization for Parameter Estimation..Distrib. Sys. 1999

Textbook in PDF format

It is well-understood that the choice of experimental conditions for distributed systems has a significant bearing upon the accuracy achievable in parameter-estimation experiments. Since for such systems it is impossible to observe their states over the entire spatial domain, close attention has been paid to the problem of optimally locating discrete sensors to estimate the unknown parameters as accurately as possible. Such an optimal sensor location problem has been widely investigated by many authors since the beginning of the 1970s, but the existing methods are either restricted to one-dimensional spatial domains for which some theoretical results can be obtained for idealized linear models, or onerous, not only discouraging interactive use but also requiring a large investment in software development. The potential systematic approaches could be of significance, eg for environmental monitoring, meteorology, surveillance, hydrology and some industrial experiments, which are typical exemplary areas where we are faced with the sensor location problem, especially owing to serious limitations on the number of costly sensors