Identification, Control and Prediction Techniques applied to Differential Wheeled Robots
Systems identification, mobile robot, control, linear model, Smith Predictor.
This work presents the identification, control, filtering and prediction of a two-wheeled differentially-driven mobile robot. For this, the variable substitution methodology will be used in the mathematical model of the robot, with its position variables, represented in the Cartesian coordinate system, replaced by its linear displacement. As the model becomes essentially linear, without the use of linearization techniques, classic control approaches, such as PID, can be used, as well as the use of predictors and filters. For the identification of the model, real experiments were performed for data collection and the least squares method was used for different structures considering the order and delay time. Using Akaike information criterion and real robot reaction experiments, it was possible to choose a model that best represents it. Decoupling the system from the robot is proposed to reduce the unwanted cross relationship that exists between inputs and outputs. It is also proposed to use the Smith Predictor and Kalman Filter to minimize the delay time and estimate the filtered values of linear displacement and orientation, respectively. Finally, some identified models and the execution schedule for the completion of this work will be demonstrated.