This paper presents a Functional Electro-Stimulation distributed architecture based on a wireless network, for therapeutic training of disabled patients. On this distributed architecture, the movement (of disabled members) is artificially controlled by means of a global controller which pilots a set of stimulation units. The closed loop control system we developed for controlling muscle is based on a high order sliding mode method. In such wireless network-based control, the variable delay introduced by the network must be taken into account to ensure the stability of the closed loop. Thus, in order to characterize the medium on which the control is performed, we carried out accurate measurements of the architecture performances (stack-crossing, round-trip time, etc.). We then propose the use of a Kalman filter to predict the communication delay evolution, with the aim to exploit it within the closed loop control.