Architecture of predictive control for a Stewart platform manipulator

This paper presents the elaboration of a comprehensive Matlab-Simulink^™ virtual environment of a Stewart platform. This simulation tool includes the kinematic and dynamic modeling of the platform, along with the control loops of the actuators and coupled to a trajectory generation module, it also offers complete visualization features through a dedicated graphic user interface. This simulator was modularly designed following open architecture principles, with plug and play modules allowing a global kinematics, dynamics and control of the machine integral analysis. The first part of the paper focuses on the features of this simulator, by describing the different available models and tools. Based on this computer-aided control system design environment, then the second part develops a comparison between two different control strategies, on the one hand, the classical PID joint control structure and on the other hand, the Generalized Predictive Control (GPC) strategy. Finally, the virtual environment provides time domain results and performance comparison, in addition the spatial tracking error and the disturbance reaction are analyzed for both controllers, showing better accuracy with the GPC one.