This thesis presents a movement recognition and locomotion control system for a hydraulic lower extremity exoskeleton for stable walking on level and non-level ground. A kinematic based decision tree method is proposed to accurately recognize the movement, which is integrated with the kinematics of the robot and individual sensor signals. The proposed method performs a kinematic analysis for each gait mode estimated by ground reaction force. A decision tree algorithm was constructed to recognize the movement mode using the vertical position of both feet. To verify the proposed model recognition performance, walking experiments were conducted for various terrain, such as level ground, stair ascent and descent, and ramp ascent and descent. A dual mode control method is also proposed for natural and stable walking in various terrain that includes active mode, passive mode, and transition control. Hydraulic actuators generate actuating forces to support the load in active mode whereas the joints move freely with no actuating force in passive mode. A pre-transition algorithm is applied to reduce resistive forces generated by residual pressure during gait transition. Walking experiments were performed while carrying backpack loads on level ground, stairs, and ramp terrain to verify the proposed locomotion control algorithm effectiveness. To evaluate the level of assistance, muscle activation levels were measured by electromyography sensors with and without the exoskeleton. The experimental test verified that the proposed method is capable of stable walking on level and non-level ground while supporting a backpack load.