For patients with hemiplegia or stroke and other limb locomotion dysfunction due to problems such as neurological, an important therapy is Rehabilitation training. Repetitively task-directed exercise can increase the motion performance of patients with nerve problems. In this paper, an adaptive controller for the lower limb rehabilitation robot (LERR) is proposed to take control of the robot's gait by using the auxiliary training force to track the patient's unknown joint trajectory in order to take into account the patient's demand. The regulator contains a set of stable controllers that are parameterized by Youla-Kucera and an adaptive algorithm has been developed for the best parameters in the online search controller to produce the required auxiliary training force to track the patient's unknown joints gait. The designed controller is validated in Adams-Matlab co-simulation environment. The results show that the adaptive regulator can provide effective training force for the unknown joint trajectory of the patient.