Aiming at the errors of compliant parallel robot, a study on the motion planning and trajectory tracking of robot system was presented to improve the overall performance. According to the performance requirements, the structural parameters of compliant joint were designed. The characteristics of compliant joint were analyzed. The analysis model and the kinematic equation were obtained. Aiming at the length errors of links made by axis drift, a numerial method for determining the actual length of drive and driven links was proposed. Based on the radial basis function (RBF) neural network, a model approximation control method was designed to compensate the vibration and the parameter perturbation errors of parallel robot system for tracking the desired trajectory. The virtual simulation model of compliant joint parallel robot system was estalished by SolidWorks, ANSYS, ADAMS and Matlab/Simulink. The simulation results show that the errors of end-effector trajectory without compensation of compliant joint errors can be reduced by 84% based on the proposed motion planning and control method, which is effective to improve the operating accuracy of compliant joint parallel robot system.