This paper is devoted to the problem in controlling a compliant robotic system by means of actuators with musclelike properties, which underlie prescribed bounds due to the natural muscle behavior. A typical example to demonstrate the effectiveness of developed control schemes is the choice of a (inverted) pendulum with higher degree of freedom. Due to the force restriction of the driving muscle forces, we have to sought (saturated) feedback strategies to control the system behavior (e.g. tracking of paths) which have to be limited a-priori. A suitable control variable can be generated by adaptive controllers, e.g., a PID-λ-stabilization. But, the classical torque control variable has to be converted to the muscle force at the joints, and the joint angle velocity has to be converted to the contraction velocity. The effective force at every joint is the difference of the antagonistic muscles pairs with the muscle characteristic curve of HILL (force-velocity-relation). The aim is now, to hold the control variable inside the area restricted by the muscle pairs. Several simulations show the effectiveness of the designed controllers.