This paper offers a new push recovery method to stabilize the quadruped robots walking on even terrains in the presence of external disturbance forces exerted on the body and legs of the robot. To avoid force sensors, a class of nonlinear observers is usually used to estimate all the forces that are applied to the leg joints. However, such estimations involve permanent errors in the case of fast varying external forces. A new sliding-mode control method is designed to enable the robot to track a desired gait with high precision despite the persisting estimation error of the original nonlinear disturbance observer. A new push recovery method that uses the estimations of the disturbance observer is proposed to help the robot maintain its balance in the presence of sudden external forces applied to the robot legs and body. To this end, an optimization problem is proposed to calculate the optimized accelerations of the leg joints of the robot. This controller was applied to the TMUBot quadruped robot as a case study to verify the effectiveness of the design. Moreover, unlike other similar works, which use massless model of the legs, the masses of the legs of the quadruped are considered in this work. As such, the dynamic equations of the legs have also been considered throughout the proposed design method.

中文翻译：

---

使用非线性干扰观测器作为虚拟力传感器的推力恢复算法的步行四足机器人的滑模跟踪控制

本文提供了一种新的推力恢复方法，以稳定四足机器人在平坦地形上行走，在存在施加在机器人身体和腿部上的外部干扰力的情况下。为了避免力传感器，通常使用一类非线性观测器来估计施加到腿关节的所有力。然而，在快速变化的外力的情况下，这种估计涉及永久误差。尽管原始非线性干扰观测器的估计误差持续存在，但设计了一种新的滑模控制方法，使机器人能够高精度地跟踪所需的步态。提出了一种新的推力恢复方法，该方法使用干扰观测器的估计，以帮助机器人在施加到机器人腿部和身体上的突然外力存在时保持其平衡。为此，提出了一个优化问题来计算机器人腿部关节的优化加速度。该控制器作为案例研究应用于 TMUBot 四足机器人，以验证设计的有效性。此外，与其他使用无质量腿模型的类似作品不同，本作品考虑了四足动物腿的质量。因此，在整个提议的设计方法中也考虑了腿的动力学方程。