Optimal Trajectory for Active Safe Falls in Humanoid Robots

Humanoid robots are being introduced in multiple environments, including houses, health care facilities or factories. Bipedal robots are, however, inherently unstable, and they might fall due to multiple reasons, including internal failures or external perturbations. In these situations, the robot should guarantee as much as possible the integrity of humans in the workspace, of the environment, and of the robot itself. When there is some control authority left on the robot, it can be actively commanded to follow a predefined trajectory that minimizes the consequences of the impact with the ground. This paper presents the computation of an optimal falling trajectory using a telescopic inverted pendulum model, which translates into squatting and stretching motions in the robot to dissipate as much energy as possible. The results show that the prescribed trajectory is effective for maximizing the dissipated energy before the actual impact.