This paper presents the design of a parallel robot that is used to play table tennis. The object of this application is to evaluate the level of integration between high-speed robotic systems and vision systems in unknown and dynamic environments. The parallel robot is based on a DELTA robot. The optimization method uses a global condition index based on the condition number of the Jacobian matrix. Firstly, we present the geometric model and kinematic analysis of the parallel robot. The optimization defines the dimensions of the manipulator. Using ADAMS/spl reg/, we obtain the maximum speed of the end-effector and the maximum torque of the motors for a set of typical trajectories of the ball in table tennis. Finally, we display the prototype of the built manipulator.