Balance recovery of a quadruped robot

Maintaining the balance or stability of legged robots in natural terrains is a challenging problem. Besides the inherent unstable characteristics of legged robots, the sources of instability are the irregularities of the ground surface and also the external pushes. In this paper, a push recovery framework for restoring the robot balance against external unknown disturbances will be demonstrated. It is assumed that the magnitude of exerted pushes is not large enough to use a reactive stepping strategy. In the comparison with previous methods, which a simplified model such as point mass model is used as the model of the robot for studying the push recovery problem, the whole body dynamic model will be utilized in present work. This enhances the capability of the robot to exploit all of the DOFs to recover its balance. To do so, an explicit dynamic model of a quadruped robot will be derived. The balance controller is based on the computation of the appropriate acceleration of the main body. It is calculated to return the robot to its desired position after the perturbation. This acceleration should be chosen under the stability and friction conditions. To calculate main body acceleration, an optimization problem is defined so that the stability, friction condition considered as its constraints. The simulation results show the effectiveness of the proposed algorithm. The robot can restore its balance against the large disturbance solely through the adjustment of the position and orientation of main body.