Nonlinear friction is one of the main factors that effect both static and dynamic properties of high-precision mechanical servo systems. Considering the different friction characteristics of high-precision experimental platforms at different positions，a modified friction model based on the LuGre model is presented. By taking velocity and position signals as input variables

both static and dynamic parameters of the model are identified using the genetic algorithm. Based on the modified friction model

the experiments and simulations of the friction phenomenon and the friction feed-forward compensation are performed by a high-precision experimental platform and a corresponding simulation servo platform

respectively. Experiments show that the following error after compensation is 1/3 that of original one and the static error of system decreases from 1.4 μm to 0.4 μm with compensation

which is basically consistent with the simulation results without and with compensations. This modified friction model can describe accurately the friction characteristics of the experimental platform

and can be used in the feed-forward compensation to decrease following errors and improve the positioning accuracy for the system.