The use of Pneumatic Muscle Actuator (PMA) in medical robots for rehabilitation has changed due to the requirements for a compliant, light weight and user-friendly robotic system. In this paper, a control system for controlling the bi-articular actuators (PMA) is proposed. Based on the information obtained from the positional input data (hip and knee joint angles), a contraction model is derived using mathematical equations to determine the contraction patterns of antagonistic mono- and bi-articular actuators, and then implemented it into the control system. Anterior and posterior muscle activation levels are introduced into the model to manipulate its magnitude. There are two tests for the control system; first is with antagonistic mono-articular actuators alone; second is along with antagonistic bi-articular actuators. The contraction model control scheme was tested on a healthy subject in a robot assisted walk test, and satisfactory performance was obtained. The result showed that, the cycle time of the gait training system is improved up to 3 seconds gait cycle compared to 5 seconds gait cycle used in previous research. However, a little time shift and inertia occurred when the controller is tested at faster gait cycle time of 2 seconds and 1 second. Thus, the potential field and iterative learning control are suggested to improve the gait cycle of the system.