Applying Niche Genetic Algorithms to Optimizing Motion Control of Multi-DOF Robots

[[abstract]]本研究主要基於遺傳基因演算法(Genetic Algorithms)，以最佳化方式結合順向運動學(Direct Kinematics)，求解機械手臂之逆向運動解(Inverse Kinematics)，並達成機械手臂操作時，軸關節間之平滑性。 多自由度機械手臂之逆向運動解，本質上屬於多模態(Multimodal)問題，即同時可能有多組軸關節模式(Configuration)滿足一給定的操作點。當進一步考慮運動中的軸關節平滑性及相關參數設定時，其解空間將為一多峰值分佈，同時存在多個可能的區域最佳解(Local Minima)。 因此，本研究利用小生境(Niche)法來保持演化算法之多樣性，以求得可能的全域最佳解，並以六軸關節機械手臂PUMA 560、五軸關節機械手臂SCORBOT-ERVII及五軸關節機械手臂JM-ARM-001，作為研究載具，依順向運動學中之Denavit -Hartenbreg (D-H)法則，推導其空間幾何轉換矩陣，藉以推論並驗證所得軸關節解其平滑特性。[[abstract]]This research uses genetic algorithms with the direct kinematics equations to explore the inverse kinematics solutions of multiple DOFs (Degree of Freedoms) robot arms. And the motion smoothness in the joint spaces would be guaranteed simultaneously in the inverse process. The inverse kinematics solutions of a multiple DOFs robot arm are multimodal in nature and usually multiple possible configurations may exist to achieve the expected manipulator position. While considering the parameterized motion smoothness of a robot arm, several possible peak values would distribute in the search space and form the local minima set to be overcome. Hence, the niche method is applied here to maintain the diversity of genetic algorithms and serve as the global optimum explorer. In the article, an experimental robot arm is modeled by Denavit-Hartenbreg (D-H) formulations according to the direct kinematics and used to validate the joint-space smoothness for the proposed inverse solution