This study aimed to identify the effects of rhythmic arm swing during robot-assisted walking training on balance, gait, motor function, and activities of daily living among patients with subacute stroke. Twenty patients with subacute stroke were recruited, and thereafter randomly allocated to either the experimental group that performed the robot-assisted walking training with rhythmic arm swing, or the control group that performed the training in arm fixation. In total, 30 training sessions were carried out. The outcome measures included the 10-m walk test, Berg balance scale, timed up-and-go test, fall index that was measured using the Tetrax system, motor function test of Fugl-Meyer assessment, and modified Barthel index. The patients of both groups showed significant improvement in all parameters after the intervention (P<0.05). The Berg balance scale, Fugl-Meyer assessment, and modified Barthel index scores at post-test appeared to be significantly higher for the experimental group than for the control group (P<0.05). These findings indicate that more favorable effects from robot-assisted walking training in patients with subacute stroke may be obtained by the use of rhythmic arm swing.