Design and control of parallel kinematics platform for nonprehensile manipulation

This work describes recent results on design and experimental validation of the parallel kinematics robotic platform for nonprehensile manipulation tasks. Such type of systems are widely used in flight simulators, automobile simulators, industrial automation such as fast sorting. We built a 2DOF Stewart-like platform with webcam-based vision system. Software integration of the system was performed in MATLAB/Simulink. The platform was specially built to conduct the following experiments: object stabilization in the specific point on the plate, optimal point-to-point motions and desired trajectory tracking as well as object-to-plate contact model identification. This paper presents control-oriented dynamic model of the system and corresponding control algorithms including extension of the output adaptive controllers suitable for discrete time systems. We also compare the adaptive motion control algorithms with conventional approaches like PID-controller.