Visión artificial y Filtro de Kalman aplicados al control de un péndulo invertido

Machine vision systems comprise a technology whose development has grown exponentially in recent years. These technologies not only make an essential tool for certain applications, but also enable the substitution of multiple mono-variable sensors (used in process automation) for a single multi-variable sensor adquisition system. In this paper the actual implementation of an inverted pendulum is presented. The control loop is based on artificial vision for the determination of position and the application of a Kalman Filter for the estimation of complete system state. This document describes the model of the system, its stochastic characterization, implementation of Kalman Filter and vision system considerations aimed at improving its robustness. Simulations and tests are presented concluding with the comparative analysis of image-based measurement vs encoders based measurement.