Nonlinear friction is the main factor affecting the dynamic and static performances of robotic excavator electrohydraulic servo systems. The traditional LuGre friction model is only related to the velocity. However, the friction characteristics of the electrohydraulic servo system are also related to the position and direction. An improved LuGre is proposed to describe the friction characteristics between position, velocity and direction, and the instability problem is also solved by stopping the observation of the internal state z . Then, inertia weight, asynchronous change method and elite mutation operation are added into the basic particle swarm optimization algorithm to identify the friction model parameters accurately. Moreover, a friction compensator (FC) based on the LuGre model is designed. Three controllers are implemented with different trajectory experiments on a 23 ton excavator. Experimental results demonstrated that the FC based on the improved LuGre can improve the comprehensive performance of the electrohydraulic servo system.