A dynamic model for the bipedal robots under uncertain disturbances was established, aiming at the robust control problem of dynamic bipedal walking under uncertain disturbances. The specific Poincaré mapping method was extended to the stability analysis of bipedal robots under uncertain disturbances, by which the stability analysis of a robot random system was transformed into that of a deterministic and periodic system. An adaptive sliding-mode controller was proposed based on the sliding-mode control method. Compared with the traditional sliding-mode controllers, the proposed controller is applicable even when the magnitude information of the external disturbances is not accurately estimated. Considering that bipedal robots often encounter uneven terrains, the proposed sliding-mode controller was further extended to the robust control of dynamic walking on uneven terrains. For this goal, an invariance condition for impact velocity was presented and the trajectories were planned online based on landing-velocity control. Then, the presented controller was used to enforce feedback control on the robot. Numerical simulations were performed on a three-dimensional (3-D) five-link bipedal robot, and results show that the robust control of dynamic bipedal walking subject to uncertain disturbances can be realized by the proposed controller effectively.