<p>With the development of the quasi-passive biped robot, chaotic gait suppression serves as the research focus in recent years since it has a considerable influence on gait stability in bipedal walking. Aiming at limited performance of traditional control methods that yields unsatisfactory chaotic gait suppression in bipedal walking, in this paper, a novel chaotic gait suppression method called adaptive predictive feedback control based on dynamic fuzzy neural network is proposed for suppressing chaotic gait of the biped robot. To model the controller, first, the predictable condition of the biped chaotic gait is analyzed and derived based on the asymptotic stability of the hybrid dynamical system. Then, a multi-step prediction model based on a dynamic fuzzy neural network was established to perform the chaotic gait prediction. Furthermore, an adaptive predictive feedback controller derived from the conventional delayed feedback control was designed to gradually adjust the chaotic gait based on current state and the future state predicted by neural network. The performance of the proposed framework is validated by numerical simulations, the results demonstrate the effectiveness and superiority of the proposed approach.</p>