Robot-assisted rehabilitation facilitates high-intensity training of the impaired upper limb in neurological rehabilitation. It has been clinically observed that persons with Multiple Sclerosis (MS) have difficulties in sustaining the training intensity during a session due to the development of activity-related muscle fatigue. An experimental observational pilot study was conducted to examine whether or not the muscle fatigue develops in MS patients during one session of robot-assisted training within a virtual learning environment. Six MS patients with upper limb impairment (motricity index ranging from 50 to 91/100) and six healthy persons completed five training bouts of three minutes each performing lifting tasks, while EMG signals of anterior deltoid and lower trapezius muscles were measured and their subjective perceptions on muscle fatigue were registered. Decreased performance and higher subjective fatigue perception were present in the MS group. Increased mean EMG amplitudes and subjective perception levels on muscle fatigue were observed in both groups. Muscle fatigue development during 15′ training has been demonstrated in the arm of MS patients, which influences the sustainability of training intensity in MS patients. To optimize the training performance, adaptivity based on the detection of MS patient's muscle fatigue could be provided by means of training program adjustment.