Based on online machine learning, the automatic control system for lower limbs gait rehabilitation is designed. The system structure design, motion signal acquisition and rehabilitation analysis processing technology are introduced. The software architecture of the distributed control system, kinematical modeling of the lower limbs exoskeleton, implementation of fundamental motion control algorithm and the design of gait rehabilitation system are expounded. This paper illustrates the software architecture of distributed control system, and the design of motion sensing and controlling algorithm. The system utilizes polynomial regression and online gradient descent to analyze customers’ lower limbs gait, these results are used by motion control system to generate more accurate motion parameter; and has the function of perceiving, recording, and remotely monitoring customers’ motion. The system has friendly interactive interface for staffs of monitor center or relatives of patients to supervise state of patients’ motion and information of environment in real time. Experimental results prove that the system could reliably help customers’ lower limbs gait rehabilitation, possessing some practical value.