Parallel robot is a kind of nonlinear strong coupling system with many branches and joints. It has obvious advantages of high speed, high stiffness and large load. With the number of joints increasing, its control accuracy is generally not high. In order to improve the accuracy of 3-RRRU parallel robot, kinematic modeling and error calibration method were systematically researched. Firstly, the kinematic equation and error model were derived by DH theory and space vector method. On this basis, the error model of the robot was derived and established with the partial differential theory. Secondly, position data were collected by using laser tracker for straight line and curve path. Lastly, genetic algorithm was optimized and used to complete calculation. The experiment result showed that the tracking error was controlled between 0.14mm and 1.34mm based on the linear trajectory calibration, and the maximum error was greatly reduced from 9.36mm to 1.34mm. But this calibration mode was not suitable for curve path compensation. Its maximum error of curve compensation reached 5.08mm. The linear calibration was just suitable for straight path, and its compensation accuracy was also lower than that of curve calibration mode. After compensation, the maximum error of line trajectory and curve trajectory was respectively reduced to 1.18mm and 1.56mm. According to the experimental data, 3-RRRU robot had better accuracy in the central area of workspace. In summary, the proposed method owned high automation and its feasibility of the method was verified by experiments.