To compensate for missing functions of lower extremity amputees, this paper proposes a model of an intelligent bionic leg semi-actively controlled by a magnetorheological damper (MRD). A mechanical structure of the leg is designed, and with the help of Bouc-Wen model, a MRD forward dynamics model is constructed for simulation. On the basis of data obtained from simulations, the MRD inverse dynamics model is constructed by BP neural network. An integrated control platform is constructed for the intelligent bionic leg, on which the simulation for knee joint control is conducted. Results of the research show that the MRD forward dynamics model can precisely express relationships among damping force, velocity, and displacement, while the inverse forward dynamics model can accurately predict MRD control currents, thus obtaining precise trajectory tracking effects of knee joint movement.