Objectives To explore the effect of robot-assisted gait training（ RAGT） on the gait ability in patients with hemiplegia after stroke. Methods From January 2015 to October 2016， a total of 60 patients with hemiplegia were randomly divided into the RAGT group（ n=30） and control group（ n=30）. Both groups received conventional rehabilitation， while the RAGT group was given robot-assisted gait training and control group was given manually assisted gait training 30 minutes a day， 5 days a week for 8 weeks. L-FORCE/HIP/ Flex（ FHF）， 6-minutes walking distance（ 6MWD）， maximum walking speed（ MWS） and Fugl-Meyer Assessment of legs（ FMA-LE） were evaluated before and after training. Results There was no significant difference in general information and indexes between the two groups before training（ P ＞ 0.05）. The scores improved in both groups after training（ P＜0.05）. The FHF， 6MWD， MWS， and FMA-LE of RAGT group was（ 51.53±22.24）Nm， （80.50±15.62）m，（ 45.54±29.49）s and（ 21.81±7.77）， respectively， which were significantly better than the control group［ （43.65±24.94）Nm，（ 66.44±19.81）m，（ 58.89±21.76）s and（ 15.13±8.67）， P＜0.05］. Conclusions Robot-assisted gait training can effectively improve the gait ability of stroke patients with hemiplegia.