A major challenge to developing neuroprostheses for walking and to widespread acceptance of these walking systems is the design of a robust control strategy that provides satisfactory tracking performance, to be robust against time-varying properties of neuromusculoskeletal dynamics, day-today variations, muscle fatigue, and external disturbances, and to be easy to apply without requiring offline identification during different experiment sessions. The lower extremities of human walking are a highly nonlinear, highly time-varying, multi-actuator, multi-segment with highly inter-segment coupling, and inherently unstable system. Moreover, there always exist severe structured and unstructured uncertainties such as spasticity, muscle fatigue, external disturbances, and unmodeled dynamics. Robust control design for such nonlinear uncertain multi-input multi-output system still remains as an open problem. In this paper we present a novel robust control strategy that is based on combination of adaptive fuzzy control with a new well-defined sliding-mode control (SMC) with strong reachability for control of walking in paraplegic subjects. Based on the universal approximation theorem, fuzzy logic systems are employed to approximate the neuromusculoskeletal dynamics and an adaptive fuzzy controller is designed by using Lyapunov stability theory to compensate for approximation errors. The proposed control strategy has been evaluated on a planar model of bipedal locomotion as a virtual patient. The results indicate that the proposed strategy provides accurate tracking control with fast convergence during different conditions of operation, and could generate control signals to compensate the effects of muscle fatigue, system parameter variations, and external disturbances. Interesting observation is that the controller generates muscle excitation that mimic those observed during normal walking.