Koptev, MikhailFigueroa, NadiaBillard, Aude2021-03-262021-03-262021-03-262021-04-0110.1109/LRA.2021.3057024https://infoscience.epfl.ch/handle/20.500.14299/176410WOS:000621399900004In this letter, we propose a real-time self-collision avoidance approach for whole-body humanoid robot control. To achieve this, we learn the feasible regions of control in the humanoid's joint space as smooth self-collision boundary functions. Collision-free motions are generated online by treating the learned boundary functions as constraints in a Quadratic Program based Inverse Kinematic solver. As the geometrical complexity of a humanoid robot joint space grows with the number of degrees-of-freedom (DoF), learning computationally efficient and accurate boundary functions is challenging. We address this by partitioning the robot model into multiple lower-dimensional submodels. We compare performance of several state-of-the-art machine learning techniques to learn such boundary functions. Our approach is validated on the 29-DoF iCub humanoid robot, demonstrating highly accurate real-time self-collision avoidance.Roboticsrobotshumanoid robotsmanipulatorslegged locomotioncollision avoidancesupport vector machinestorsohumanoid robot systemsmachine learning for robot controltext::journal::journal article::research article