基於雙四元數之B樣條曲線運動規劃方法

Abstract

擬人化已成為新一代機器人必備的條件，擬人化的範圍不再只是外型的模仿，更著重在於機器人行為擬人化的程度。而運動規劃方法的好壞，決定了機器人的行為與人類行為是否相似的關鍵。因此本篇論文提出了一個基於雙四元數的運動規劃器，導入了四元數指數和對數映射的方式去進行空間轉換，在三維的歐基里德空間中來描述末端效應器的姿態，並透過三階分段型B-樣條曲線的方法，同時針對末端效應器的位置與姿態進行軌跡曲線規劃，以達成平滑的目標。另外，在軌跡追蹤層面，建立了一個以雙四元數為基礎的追蹤控制架構，此架構採用模糊 PID控制器為主體，透過建立模糊規則庫的方式，來對PID參數進行動態調整，有效地處理複雜的動態軌跡追蹤控制的問題。最後實現在所設計的3D雙臂機器人(DAR-I)上，並利用兩組倒茶動作的模擬實驗，來驗證曲線的平滑性以及位置與姿態的相互配合度。並與其他軌跡規劃方法進行比較，皆得到不錯的效果。

The inclusion of anthropomorphic features in robots has become so commonplace that robots are now expected to behave like human beings. This study presents a motion planner based on dual quaternion with the import of exponential and logarithmic maps of the quaternion for space conversion in order to describe the attitude of the end-effector in a three-dimensional Euclidean space. Trajectory curves are planned for the position and attitude of the end-effector using the segmental cubic B-spline approach in order to provide unified treatment of smoothed trajectories of position as well as attitude. For trajectory tracking, we developed a dual-quaternion-based tracking control architecture based on a fuzzy PID controller. The PID parameters are dynamically adjustable using a fuzzy rule database, which was designed to handle complex trajectory tracking control problems. Finally, the proposed method was applied to a self-designed dual arm robot-I (DAR-I) through two tea-pouring experiments to evaluate the effects of treating of position and attitude in a unified manner. Our simulation results demonstrated better performance compared with those of other methods.