Kinematic redundancy may be an alternative for enlarging the workspace and for improving the dynamic performance of parallel manipulators. It can be implemented by the introduction of an extra active joint in an active kinematic chain. Kinematically redundant parallel manipulators requires redundancy resolution schemes since their inverse kinematic problem presents infinite solutions. These schemes can be posed as an optimization problem that can be solved locally or globally. These work compares numerical and experimental results of local and global redundancy resolution schemes applied to a three-level kinematically redundant 3PRRR parallel manipulator. The results demonstrate that for the execution of predefined task, global strategies present improved results.