Abstract
In 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.
Highlights
H UMANOID robots have a wide span of possible applications, potentially becoming universal assistants
We focus on providing a real-time solution to the self-collision avoidance (SCA) problem for highdimensional position-controlled humanoid robots
The SCA boundary functions and gradients were re-formulated as inequality constraints in a quadratic program (QP)-based optimization problem that is solved in real-time
Summary
H UMANOID robots have a wide span of possible applications, potentially becoming universal assistants. As the robot’s joint space dimensionality increases, so does the computational and geometrical complexity of the problem While these specific methods have been capable of generating collision-free trajectories for high-DoF humanoid robots that can be dynamically-stable, efficient and optimal [8], [9], [10], their main shortcoming is computation time. The SCA boundary functions and gradients were re-formulated as inequality constraints in a quadratic program (QP)-based optimization problem that is solved in real-time This approach was validated on a 14-DoF dual-arm setup, it’s scalability to higherdimensional humanoid robots was not explored. We propose to learn continuously differentiable boundary functions that can be approximated in real-time to prevent self-collisions while controlling the 29 DoF iCub humanoid.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
