Sensor Network Design based on the Observability and Precision degree

Abstract

The Unscented Kalman Filter is a state estimation method used in nonlinear dynamic systems to estimate the mean and covariance of a random variable undergoing a nonlinear transformation, knowing the process model and the measurements. Therefore, an adequate choice of the measured variables improves the performance of the filter technique. In this context, the sensor network design problem allows selecting a set of variables that minimizes the global estimation error when the instrumentation budget is limited. This is solved using a level traversal tree search algorithm, whose computation time is reduced by evaluating the design criteria sequentially. In this work, it is proposed to address the effect of the circumstantial loss of measurements on the system observability and the estimates precision. The success of the sensor network design methodology is demonstrated for the copolymerization process of Methyl Methacrylate and Vinyl Acetate, widely studied in the literature.