IEEE Robotics and Automation Letters | 2021
Immediate Generation of Jump-and-Hit Motions by a Pneumatic Humanoid Robot Using a Lookup Table of Learned Dynamics
Abstract
This letter focuses on the jump-and-hit motion of a humanoid robot, wherein a robot instantaneously jumps forward and hits a flying ball in the air, similar to how human players behave in volleyball games. We propose a Immediate Motion generation using a Lookup table of learned dynamics (IMoLo) for generating the motions of a pneumatic humanoid robot. To test this method, we developed a humanoid robot called “Liberobot” with eight joints applying structure-integrated pneumatic cable cylinders. Using simulations, the prediction errors of the robot hand positions during the jump-and-hit motions measured via nonlinear interpolation when using IMoLo was smaller than without it in cases having a small number of training trials. In the experiments, the robot jumped and hit the flying ball 16\xa0times out of 20\xa0trials using the proposed motion generation method. The results indicate that a pneumatic humanoid robot using IMoLo can instantaneously perform dynamic whole-body motions, such as jump-and-hit motions, with a changing target within a specified time. Our humanoid robot is the first pneumatic humanoid robot capable of executing such dynamic motions.