Aiming at the problem that it is difficult to establish accurate mathematical model for robot systems with joints clearances using traditional dynamic methods, a twolink robot with clearances was taken as an example and the least square method, traditional genetic algorithm and improved genetic algorithm were applied to perform parameter identification. With the improved genetic algorithm improved the traditional genetic algorithm from the coding method, crossover operator, mutation operator, and adaptive improvement of crossover probability and mutation probability, the problem of “premature” phenomena was solved. The results indicate that the improved genetic algorithm has the highest identification accuracy and the fastest convergence speed in the parameter identification of twolink robots with clearance using the three identification methods. Compared with traditional dynamics modeling, the identification method is more convenient and fast, and it provides a reference for the model constructing of a robot system with clearances.