=== How a robot can avoid to fall? ===
In this video we present the implementation on iCub of a so-called stepping-strategy, whose aim is to recover the robot balance by taking a step .
This approach mimics humans reactions to pushes, but its development resorts to simple models like the Linear Inverted Pendulum.
After the push, the robot understands it is the right moment to start moving, then it places the swing foot in a position computed depending on the perturbation, ensuring the robot won’t fall once in double support.
These two informations (the trigger and the foot target position), together with the definition of a Center of Mass trajectory, constitute a planner able to provide references to the Momentum-Based Whole-Body Controller implemented on iCub . Thus, a peculiarity of this approach is the application on a torque controlled robot, providing additional robustness to impacts and foot placement errors, for example.
The one shown in this video is a benchmark test, where the robot is pushed while performing a part of the “Tai Chi” movements (https://youtu.be/9XRI4BeXN78).
The push causes the robot to lose its balance, forcing it to perform a side step.
=== Associated Publication: ===
 Dafarra, S. and Romano, F. and Nori, F. “Torque-Controlled Stepping-Strategy Push Recovery: Design and Implementation on the iCub Humanoid Robot”. In IEEE/RAS International Conference on Humanoid Robots, 2016. https://drive.google.com/open?id=0BzIrhcovQt-UZXdodTRaNGJhWTg
 Nori, F. et al. “iCub whole-body control through force regulation on rigid non-coplanar contacts”. In Frontiers in Robotics and AI, 2015. http://journal.frontiersin.org/article/10.3389/frobt.2015.00006/full
=== Useful links ===
Dynamic Interaction Control
This work was supported by the
FP7 EU Project Koroibot (No. 61909 ICT-2013.2.1)
IIT Video Production
Video editing: D.Farina