Robotic exoskeletons inspired by the animal’s external covering are wearable systems that enhance human power, motor skills, or support the movement. The main difficulty, apart from the mechanical structure design, is the development of an exoskeleton control system, as it should recognize the movement intended by the user and assist in its execution. This work is devoted to the exoskeleton of the upper limbs that supports movement. The method of controlling the exoskeleton by means of electromyograms (EMG) was presented. EMG is a technique for recording and assessing the electrical activity produced by skeletal muscles. The main advantage of EMG based control is the ability to forecast intended motion, even if the user is unable to generate it. This work aims to define strategies for controlling the exoskeleton of the upper limb in children suffering from neuromuscular diseases. Such diseases gradually reduce the mobility of the lower and upper limbs. These children are wheelchair bound, so it was assumed that the upper limb exoskeleton could be attached to a wheelchair. EMG signals are recorded, amplified and filtered. An artificial neural network using fuzzy logic to process EMG was used. This network predicts movement trajectories. Using this forecast and taking into account the feedback information, the control system generates the appropriate drive torques.