Saturation is an undesired event in trajectory tracking control of mechanical systems. When the actuators of a robotic device saturate, the solution of the inverse dynamics problem cannot fully be realized, which results in deviations from the desired trajectory and loss of performance. It is generally hard to consider the limited actuator torques and the corresponding nonlinear effects in the control design. The most common way to handle the problem is recalculating the control forces and trying to adjust the desired trajectory such that saturation will not happen. In contrast we propose a switched control approach, where, upon saturation, different sets of inputs are varied periodically to keep the reference point of the robot on the desired trajectory. For this, the desired motion is formulated by means of servo-constraints, and the periodic switching of these constraints is adjusted according to the variation of a new, manipulability type performance measure. It is demonstrated that the proposed controller can effectively reduce the trajectory following error due to actuator saturation. A typical robotic benchmark example is provided to show the application of the method, and to compare it with other approaches taken from the literature.