Augmented models for improving vision control of a mobile robot

This paper describes the modelling phases for the design of a path tracking vision controller for a three wheeled mobile robot. It is shown that, by including the dynamic characteristics of vision and encoder sensors and implementing the total system in one multivariable control loop, one can obtain good performance even when using standard low cost equipment and a comparatively low sampling rate. The plant model is a compound of kinematic, dynamic and sensor submodels, all integrated into a discrete state space representation. An intelligent strategy is applied for the vision sensor, including the start up, normal operation, exception handling and shut down phases. Laboratory experiments show the validity of the approach using standard Kalman filter/LQR control design