Research > Standing push recovery
Developing a robust control to maintain balance in the presence of external disturbances has remained one of the primary challenges of humanoid robots. These robots must be able to perform multiple tasks in highly complex environments. If the robot loses its equilibrium, it might injure nearby people or malfunction, which could be disastrous. Therefore, designing a reliable control method that preserves the robot’s stability is crucial.
This research is specifically concerned with the issue of maintaining balance while in an upright stance. Experiments on human balance have shown that people react to disturbance forces by bending their ankles, hips, and arms or taking steps in the case of excessive forces. The "ankle strategy" locks all joints except the ankle and balances the body like an inverted pendulum. When the pushing force is substantially larger, humans utilize a mechanism known as the "hip strategy" to generate momentum by rotating the torso at the hips. By bending the body in the direction of the disturbance force, angular momentum can be immediately generated around the center of mass (CoM), allowing the force of the disturbance to be quickly absorbed. As the disruptive force diminishes, the body attempts to regain its original position.
We propose a new controller based on momentum concepts and a new reaching law for sliding mode control that can help the robot withstand higher forces while standing. A demo of the controller in Gazebo simulator can be seen below, where the robot is pushed with a 9 Ns impulse (9000 N in 1 ms). The robot is able to balance and move back to the upright posture after the impact.
The next step is to implement this control on the real physical robot.