To reduce the position error of a parallel robot caused by joint clearances, a method to compensate the joint clearance errors was proposed by optimizing the joint angular displacement parameters of drivers. For the purpose of a planar 3-RRR parallel robot system, an error model of the joint was presented and the change rules of the errors of actual lengths of links and actual joint angular displacements of drivers caused by joint clearances were obtained. According to the inverse kinematics equation, the error analysis model of system was proposed by total derivative theory. The structural errors of system caused by joint clearance errors were compensated by optimizing the joint angular displacements of drivers using Particle Swarm Optimization (PSO) algorithm. By inducing the linear decreasing weight and the compression factor, the standard PSO algorithm was improved, and an unified expression of algorithms was derived. The results of error compensation show that the improved PSO algorithm is effective for improving the convergence performance. The trajectory errors of end-effecter using optimized joint angular displacements are decreased by 99% as compared with that of uncompensated trajectory. Obtained results demonstrate that the proposed method effectively compensates the structural errors caused by joint clearances and guarantees the position accuracy of parallel robot systems.