The process of empowering muscles in order to make them to a normal and common value is an expensive and prolonged work, in common available methods. There are some commercial exercise machines used for this purpose called rehabilitation systems. However, due to their insufficient motion freedom and prospect of being expensive, these machines have limited usage. Hence, it is clearly necessary that Mechatronic technologies should be used in this area. In this paper, an algorithm and an improved rule are presented for controlling a rehabilitation system of lower limbs which is implemented on a 6-Degree Of Freedom (DOF) Stewart parallel robot. Impedance control and adaptive control are used for this purpose. Estimation and optimization of control parameters will be done by artificial neural networks and genetic algorithms, respectively (intelligent strategy). Safety is guaranteed since some of controller parameters can be adapted under the stability conditions given by using Routh stability theory. Thereafter, the results of simulations are presented by defining a physiotherapy standard mode on a desired trajectory. MATLAB/SIMULINK is used for simulations. Finally, a comparative discussion between this strategy and common methods is devised.