二維機械手臂之避碰路徑規劃

Abstract

本論文提出了一個機械手臂路徑規劃的演算法，以實體工作空間中的人工位能場做避碰的規劃，且能夠在有限的時間中，為機械手臂規劃出一系列無碰撞的動作組態，以到達指定的目的地。本論文利用二維的人工位能場模型，讓機械手臂和障礙物之間產生推斥力作用，我們就利用力和轉動力矩兩種推斥形式，在機械手臂的各節上討論此推斥力的作用，並訂定一些規則調整機械手臂的組態，在自由空間中找到使其位能極小的組態，而達到避開障礙物的目的。 此路徑規劃演算法可應用到高自由度的機械手臂，包括自由度在8以內的機械手臂均可以達到即時運算的要求，整體效能相當令人滿意。本論文並與其他作品做比較，討論並分析做法與效能的差異。本論文所提的方法有非常廣泛的應用空間，例如此演算法也可應用到非靜態工作空間的狀況。

A physically based path-planning algorithm is presented in this thesis. The algorithm utilizes a physically based artificial potential field to model the workspace to avoid collision. It is shown that a path of a series of collision-free configurations can be obtained efficiently in very short time intervals. By analytically calculating the gradient of the Newtonian potential function, the repulsion between manipulator and obstacle is determined in forms of repulsion force and torque. According to the equilibrium condition at each joint on the manipulator, rules of configuration adjustment for minimum potential are developed. Our method is applicable to very high-DOF redundant manipulators, and the computation time for manipulators of less than 8 DOFs will satisfy the requirement of real-time tasks. In this thesis, several simulation results are compared with those based on other approaches. An example of moving obstacles shows that our method may also apply to more general path planning problems.