Functional electrical stimulation (FES) is used to excite paralyzed muscles that are no longer controlled by patients with spinal cord injuries (SCI). Appropriate stimulation patterns are chosen to stimulate intact muscles, in order to extend their overall performance, postponing thus the muscular fatigue during the daily activities such as standing, standing up, sitting down and walking. This paper presents the modeling and the control of a knee joint actuated by the quadriceps muscles. Appropriate stimulation patterns are computed as a function of the desired lower-limb knee joint movements. Parameters of the biomechanical model are identified based on experimental kinematic data. Model predictive control (MPC) is applied to the inputâ€"output feedback linearized (IOFL) system. IOFL allows linearization by inverting the system dynamics through a nonlinear feedback transformation. The described control approach is validated through different simulation scenarios for knee flexionâ€"extension. Internal dynamics stability is mathematically proved and performances are compared to those produced by classical pole placements method. The controller has shown satisfactory results in terms of regulation, stability and robustness with respect to external disturbances.