Selection of a suitable training modality according to the status of upper limb function can maximize the effects of robotic rehabilitation; therefore, it is necessary to identify the optimal training modality.

This study aimed to compare robotic rehabilitation approaches incorporating either resistance training (RET) or active-assisted training (AAT) using the same rehabilitation robot in people with stroke and moderate impairment.

In this randomized controlled trial, we randomly allocated 34 people with stroke who had moderate impairment to either the experimental group (RET, n = 18) or the control group (AAT, n = 16). Both groups performed robot-assisted therapy for 30 min, 5 days per week, for 4 weeks. The same rehabilitation robot provided resistance to the RET group and assistance to the AAT group. Body function and structure, activity, and participation outcomes were evaluated before, during, and after the intervention.

RET led to greater improvements than AAT in terms of smoothness (p = 0.006). The Fugl-Meyer Assessment (FMA)-upper extremity (p < 0.001), FMA-proximal (p < 0.001), Action Research Arm Test-gross movement (p = 0.011), and kinematic variables of joint independence (p = 0.017) and displacement (p = 0.011) also improved at the end of intervention more in the RET group.

Robotic RET was more effective than AAT in improving upper limb function, structure, and activity among participants with stroke who had moderate impairment.