The design of sensing systems for the measurement of multiple physical
quantities related to a dynamical system is considered. A multivariat
e sensor comprises several simple transducers, each measuring a scalar
quantity that comes from the combination of the components of the qua
ntity to be measured. From the collection of measurements of single tr
ansducers at different times, the desired information is extracted by
analogue or digital processing. Besides the choice of technological ch
aracteristics of the transducers to be employed, the designer of multi
variate sensors is usually allowed some freedom in choosing the number
of transducers, their arrangement in the system, and the time schedul
ing of their measurements. These choices are the subject of optimal po
licies in the design phase, whose goal is to maximize some performance
(or minimize some cost) criterion. We survey some of the existing app
roaches to optimal design of multivariate sensors, according to the di
fferent types of systems they are applied to. Two examples of optimal
sensor design are discussed as an illustration of the methods.