Cooperative Human-Robot Manipulation Tasks

Object manipulation tasks by humanoid robots often do not only involve the hands and upper limbs but require flexible control of whole body movements. Examples are cooperative transportation tasks where a human and a robot carry an object together. Here, the robot has to reactively adapt its walking motion in order to compensate for the pushing and pulling forces exerted by the human via the carried object. Our approach to this problem draws on a learned walking model that is able to predict the robot’s center of mass based on proprioceptive sensors and walking speed. Deviations between predicted and measured center of mass can then be used to adapt the robot’s walking style to the movements of the human cooperation partner.