人形機器人的手腳協調運動規劃與全身動量控制

Abstract

針對複合型的移動－操作任務，本論文提出了以優勢空間作為基底，結合人型機器人之動作意識的全身運動規劃。並透過浮體運動學與質心動量矩陣的運用，進一步推導出各連桿之轉動對機器人質心的影響。此外，根據不同之任務特性，末端執行器的軌跡則在優勢空間的不同區域中進行調整，使機器人能夠產生類似人類行為的動作。 然而在人形機器人生成所需軌跡後，仍需要實時進行控制以確保機器人在受到外力的干擾下，能夠保持自身之平衡。因此本論文提出了基於二次規劃(Quadratic Programming)的整合型控制器，藉由比例－微分控制與線性二次狀態增量(Linear Quadratic State-Incremental)控制等多種控制方法在機器人動量上的規劃，可確保機器人在執行任務期間的穩定性。並透過變動平衡穩定和任務執行之間的權重，使得機器人可以根據自身的狀態進行優先級調整，以達到穩定行走之目的。

Considering the problem of planning whole-body motions for humanoid robots that must execute loco-manipulation tasks, we propose a novel concept of dominant space to generate human-like motions with consciousness. Through the use of floating based kinematics and centroidal momentum matrix, the relation between the links and the COM can be derived. Moreover, according to the properties of different tasks, the trajectories of the end-effectors are modified in different zones of dominant space. After the generation of trajectories for a humanoid, it is essential to keep self-balance with a real-time controller to overcome the disturbance from external forces. An integrated controller based on quadratic programming method is designed to keep the balance of the robot during task period by combing several controllers, including PD control and LQSI (Linear Quadratic State-Incremental) control, to regulate the momentum of the robot. Furthermore, with the change on the weightings between balance keeping and motion execution, the robot can adjust the priorities to achieve stable walking according to its own situation.