INFLUENCE OF THE SENSORED MAGNITUDE IN THE PERFORMANCE OF OBSERVERS BASED ON MULTIBODY MODELS AND THE EXTENDED KALMAN FILTER

This work is part of a project aimed to develop real-time observers based on detailed multibody models and the extended Kalman filter (EKF). Detailed models can provide more and more accurate information than the simple models traditionally used to build observers, thus enabling the implementation of more sophisticated control strategies. In a previous work, the authors studied different multibody dynamics formulations, and concluded that the state-space reduction method known as matrix-R was the most suitable for this application. This second work is devoted to study the influence of the sensored magnitude in the performance of the observers. Although the final objective is to address complex industrial systems, preliminary studies are being carried out on a four-bar mechanism with a spring-damper element. The behavior of the matrix-R method has been analyzed for different sensored magnitudes -position, velocity, acceleration-, and for the cases when the sensored magnitude is or is not a generalized coordinate of the problem. The conclusion is that good results can be obtained in all the cases, provided the covariance of the measurement noise is adjusted to a suitable value.