This paper deals with the friction compensation in a two degree of freedom (DOF) planar robot manipulator. The friction model is based on the newest and most realistic model which is described by (Generalized Maxwell Slip) GMS friction model. An online least square algorithm was used in order to estimate the friction force. To do this, two approximations were introduced to linearize the friction model over its parameters and to write it in a formulation convenient for the estimation algorithm. Moreover, since the friction force is unknown, a filtering approach was introduced to build the friction force and its derivative using known signals. Simulation results showed that using very simple proportional derivative controller (PD) combined with a friction compensation, the robot manipulator can track a desired angle position of its two joints with a high degree of accuracy.