The path planning for traditional polishing robot is teach-play-back method. That is, after adjusting offset errors, the polishing robot repeats the same fixed program in a pre-determined path. Obviously, fixed program makes transferring the robot to new operations difficult. In this paper, the path planning schemes based on improved genetic algorithm (IGA) are proposed. The encoding ways, crossover operator and mutation operator of genetic algorithm (GA) are researched. Simulation shows that this method of path planning for polishing robot is effective