This paper focuses on the developments of a generic gait synthesis for the humanoid robot COMAN. Relying on the essential Gait Pattern Generator （GPG）, the proposed synthesis offers enhanced versatilities for the lo...This paper focuses on the developments of a generic gait synthesis for the humanoid robot COMAN. Relying on the essential Gait Pattern Generator （GPG）, the proposed synthesis offers enhanced versatilities for the locomotion under different purposes, and also provides the data storage and communication mechanisms among different modules. As an outcome, we are able to augment new abilities for COMAN by integrating new control modules and software tools at a cost of very few modifications. Moreover, foot placement optimization is introduced to the GPG to optimize the gait parameter references in order to meet the robot＇s natural dynamics and kinematics, which enhances the synthesis＇s robustness while it＇s being implemented on real robots. We have also presented a practical approach to generate pelvis motion from CoM references using a simplified three-point-mass model, as well as a straightforward but effective idea for the state estimation using the sensory feedback. Three physical experiments were studied in an increasing complexity to demonstrate the effectiveness and successful implementation of the proposed gait synthesis on a real humanoid system.展开更多
文摘This paper focuses on the developments of a generic gait synthesis for the humanoid robot COMAN. Relying on the essential Gait Pattern Generator （GPG）, the proposed synthesis offers enhanced versatilities for the locomotion under different purposes, and also provides the data storage and communication mechanisms among different modules. As an outcome, we are able to augment new abilities for COMAN by integrating new control modules and software tools at a cost of very few modifications. Moreover, foot placement optimization is introduced to the GPG to optimize the gait parameter references in order to meet the robot＇s natural dynamics and kinematics, which enhances the synthesis＇s robustness while it＇s being implemented on real robots. We have also presented a practical approach to generate pelvis motion from CoM references using a simplified three-point-mass model, as well as a straightforward but effective idea for the state estimation using the sensory feedback. Three physical experiments were studied in an increasing complexity to demonstrate the effectiveness and successful implementation of the proposed gait synthesis on a real humanoid system.
