Rehabilitation robots have played an increasing role in improving the quality of life of patients. Combined with the mechanism of pelvic motion, we propose a cable-driven parallel rehabilitation robots and a cable force optimization scheme to help patients perform lower limb rehabilitation training. We introduce an alternative objective function for the case of mutation of the optimization result. Based on analyzing the movement intention of the patient, we implement different assistive functions according to the different movement intentions, such as balance assistance and movement assistance. Our solution eliminates the need for a treadmill and allows patients to train on multiple trajectories over a larger area. We designed a numerical experiment to validate our optimization model, and the results show that our method can well assist patients to perform rehabilitation training.