Rapidly aging society faces with increases in neurological disorders including stroke. Hemiplegia, which is one of the common sequelae due to stroke, causes difficulties in activities of daily living. As the number of stroke patients grows, demands for gait training increases, where robotic gait training systems are necessary. A robotic gait training system, called 'COWALK-I,' is designed to provide pelvic motion on the transverse plane as well as leg motions in the sagittal plane during gait training sessions. The pelvic motion allows weight-shifting as well as more natural gait patterns during gait training. In this research, effect of the pelvic motion during waking in the COWALK-I system is studied. Interaction force between the healthy subjects and the COWALK-I and electromyography(EMG) sensor data are measured. The average interaction forces did not show significant difference while each subject exhibited diverse patterns. The EMG signals shows that more activation of rectus femoris and less activation of gastrocnemius and gluteus medius.