In this work the problem of controlling a process whose output is irregularly 
measured is addressed. For that purpose a predictor that estimates the outputs 
of the process at regular instants plus a conventional controller that 
calculates the control action from the estimations of the predictor is used 
(technology known as inferential control).
 
The prediction consists of estimating the output variables that want to be 
controlled from the measurements taken with several sensors, using for that a 
mathematical model of the process. The Kalman filter allows to do an optimal 
prediction if the disturbances have a Gaussian distribution of zero mean, but 
with the disadvantage of requiring a high computational cost when using several 
sensors with time-varying delays. In this work an alternative prediction 
strategy of low computational cost is proposed whose design is based on the 
knowledge of measurements availability and delays (of the process, measurement 
system or data transmission line) and the disturbances nature. The proposed 
predictors minimize the prediction error when dealing with random of time-
varying sampling with time-varying delays, disturbances, measurement noise, 
modelling error, delays on the control action and uncertainty on the knowledge 
of measurements instants. The different design strategies that are proposed are 
classified according to the kind of the disturbances information that is known 
and the required computational cost. Designs for monovariable, multivariable, 
linear and nonlinear systems have been proposed. 

In this work it is demonstrated that the inferential control systems that make 
use of the proposed predictors fulfil the separation principle, so the design of 
control systems with irregular measurements is reduced to the problem of 
designing a stable predictor plus the design of a conventional sampling 
controller.