Upper limb dysfunction is one of the most common sequelae of stroke and robotic therapy is considered one of the promising methods for upper limb rehabilitation. This study aimed to explore the clinical effectiveness of upper limb training using a rehabilitation robotic device (Rebless®) for patients with stroke. In this prospective, unblinded, randomized controlled trial, patients were randomly assigned to receive robotic training (experimental group, n = 15) or conventional therapy (control group, n = 15). Both groups received upper limb training lasting for 30 minutes per session with a total of 10 training sessions within 4 weeks. Motor function, functional evaluation, and spasticity were clinically assessed before and after the training. Cortical activation was measured using functional near-infrared spectroscopy at the 1st and 10th training sessions. The experimental group demonstrated a significant improvement in the Fugl-Meyer assessment-upper extremity score and the modified Ashworth scale grade in elbow flexors. The cortical activity of the unaffected hemisphere significantly decreased after 10 training sessions in the experimental group compared with the control group. The experimental group showed significant improvement in the Fugl-Meyer assessment-upper extremity score and spasticity of elbow flexors and had significantly decreased cortical activity of the unaffected hemisphere. Training with Rebless® may help patients with chronic stroke in restoring upper limb function and recovering the contralateral predominance of activation in motor function.