Abstract: To avoid discontinuous parameters and low-speed convergence in kinematic parameter error identification, an error model for 6R serial robots based on a 6-parameter model is established in the link coordinate systems set up with DH (Denavit-Hartenberg) method. A formula for parameter conversion is derived. In order to compare the results produced by the MDH (modified DH) model and the 6-parameter model, a computer simulation experiment in presence of random noise errors in drives and measuring instruments is designed. The average absolute error is decreased by 96.1% and 52.9% respectively by using the 6-parameter model and the MDH model. The results show that the 6-parameter model has completeness and continuity. The range of model parameters can be derived from the manufacturing tolerances. The 6-parameter model is superior to the MDH model in terms of identification speed. By limiting the parameter range, the influence of lack-of-minimality on error identification is eliminated. The method is applied to parameter identification of an SR165 robot, and the average positioning error is reduced from 2.5 mm to 0.35 mm.