Whole -body human -to -humanoid motion transfer with balance manipulability ellipsoid

This thesis aims at generating human-like, whole-body humanoid motion trajectories and introducing a new measure for measuring the ability of a humanoid robot to maintain its balance. The thesis addresses two important issues in the whole-body humanoid trajectory generation. First, we propose the whole-body trajectory generation based on motions captured from a human with trajectory adjustment based on double-layer-prioritized constraints. Several motion-capturing experiments were performed on a human to capture essential body movements, which were used to generate humanoid movements. The aim is to produce whole-body human-like humanoid movements that are compliant with different human-humanoid constraints, which are related to the problem of model and task discrepancies between humans and humanoid robots. Constraint prioritization is applied to eliminate any disagreement among conflicting constraints. Computer simulations based on a HOAP-2 humanoid-robot model were conducted to illustrate the performance and consistency of the proposed whole-body humanoid trajectory generation from the captured human motions. Second, we investigate the ability of a humanoid robot to maintain its balance by proposing the Zero Moment Point (ZMP) manipulability ellipsoid as a graphical representation of the ability of a humanoid robot to manipulate its ZMP. The size and shape of the ZMP manipulability ellipsoid are functions of the actuator limits and the current posture of a humanoid robot. The ZMP manipulability ellipsoid represents an area in which a humanoid robot can instantly move its ZMP based on the available actuating limits and posture to any location inside the area covered by the ellipsoid. We found that the existence of the overlapping area of the ellipsoid and the base of the support area can be used to predict whether the humanoid has an adequate actuating torque to maintain balance through the postural balance control. If the postural balance control is not possible to maintain its balance due to the lack of actuating torque, then it would be necessary to initiate a step to maintain its stability. The ability to predict the necessity of stepping based on actuator limits is the strength of the proposed ZMP manipulability ellipsoid