The Walkbot group received conventional physical therapy (session I for 40 min/day) companied with Walkbot training (session II for 30 min/day) 5 days a week for 4 weeks, 40 session in all. After 4-week intervention all subjects received conventional physical therapy only, 40 min/day, 5 days/week for 4 weeks.

The control group received conventional functional rehabilitation for 40 min/session, 2 sessions/day, 5 days/week for 4 weeks, 40 sessions in all. After 4-week intervention all subjects received conventional physical therapy only, 40 min/day, 5 days/week for 4 weeks. During the test period, general rehabilitation and drug treatment can be done at the same time.

The physiotherapist typed patient's information then Walkbot would adjust itself to this data and the exoskeleton legs lengthen or shorten to the patient. The computer screen would provide information about speed, duration, cadence and distance. The monitor in front of the patient can offer image and feedback during the therapy process for patients.

The whole Walkbot training process was observed by one physiotherapist and should be always ready to adjust the unloading, duration, speed and guidance force of Walkbot according to patients' different situation such as muscle spasticity, gait quality, labor-consuming and observe the changes of patients' state.

Function ambulation category was used for evaluating gait function. The outcome variable were measured immediately before treatment start (pre-training), 4 weeks after treatment (post-training) and 4 weeks after robot-treatment (follow-up). All the evaluations were done by the same professional therapist who was blind to the two group assignment.

Berg balance scale (BBS) were used for evaluating functional performance level after training. The outcome variables were measured immediately before treatment start (pre-training), 4 weeks after treatment (post-training) and 4 weeks after robot-treatment (follow-up). All the evaluations were done by the same professional therapist who was blind to the two group assignment.

EuroQol-5 dimension (EQ-5D) was used for evaluating the effect of robot training to quality of life. The outcome variable was measured immediately before treatment start (pre-training), 4 weeks after treatment (post-training) and 4 weeks after robot-treatment (follow-up). All the evaluations were done by the same professional therapist who was blind to the two group assignment.