INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING INFORMATION AND CONTROL, cilt.8, sa.12, ss.8133-8145, 2012 (SCI-Expanded)
In this study, vision based stabilization control of a real time cart-pole inverted pendulum system was implemented. Inverted pendulum system is one of the classical problems and most widely used experimental setups in control theory. The pendulum angle and cart position were measured using a camera instead of encoders. A graphical user interface (GUI) was developed in Visual Basic for image processing. Visual markers were not used to aid the algorithm in detecting the pendulum angle and the cart position from the camera images. The image processing algorithm was robustly designed against visual disturbances which enter the field of view of the camera. The time delay in measurements that arises in the image processing part of the system was reduced with an effective prediction algorithm. Full state feedback method was used to stabilize the pendulum system. The control algorithms were implemented in Simulink (R) environment and embedded in Dspace DS1103 real time controller. The system was proven to be robust to visual and physical (external force) disturbances. The image processing algorithm was able to process 50 frames per second (fps) with an average of 35ms total time delay and the cart position and the pendulum angle were obtained with 0.86mm and 0.1 degrees resolution, respectively.