Robust gait generator based on the LIPDF model for biped robot with five anti-disturbance strategies

Robust control is crucial for bipedal locomotion. This paper proposes a novel model-based robust gait generator for biped robot. First, a linear inverted pendulum with distributed flywheels (LIPDF) model is introduced, where the actively regulated ankle joint serves as one of the gait anti-disturbance strategies. Building on this model, a robust gait generator is developed that integrates five linear anti-disturbance strategies to enhance locomotion stability on uneven terrain and adaptability to external disturbances. Then, a quadratic programming (QP) solver is embedded into the model predictive control (MPC) framework, enabling efficient and adaptive gait optimization under multiple constraints. Finally, the proposed method is systematically validated through simulations and experiments on the position-controlled biped robot NaoV50. The results demonstrate a 55.6% reduction in push recovery time and more than a threefold improvement in robustness on uneven terrain.