機器手臂在農作採收之最佳姿態分析─數據驅動法

Abstract

本論文「機器手臂在農作採收之最佳姿態分析─數據驅動法」旨在利用數據驅動法(Data Driven Method)取代傳統的逆向運動學算法，並結合可操控度(Manipulability)來控制機器手臂迅速達到最佳採收姿態。文中以溫室農場為假想場地，以玉女番茄為採收目標，並將玉女番茄的生長方向納入採收的參數之中，以此來制定整套採收流程。由於玉女番茄的成串果實較多，因此本文自行設計的末端採收機構具有番茄套筒，可以將目標番茄與番茄串做區隔，同時避免在剪切根莖時誤傷其他果實。最後探討在單顆及多顆番茄的情況下，實際利用採收系統對各種生長方向的番茄進行採收。結果顯示本論文所使用的數據驅動法對於採收玉女番茄之成功率以及採收效率。

The thesis " Analysis of the Optimal Posture of a Manipulator for Agricul-tural Harvesting —Data Driven Method" aims to use Data Driven Method to replace traditional inverse kinematics algorithms and combine manipulability to control the robotic arm quickly reach the best harvest posture. In this paper, the greenhouse farm is used as the imaginary site, the cherry tomato is the harvesting target, and the growth direction of the cherry tomato is included in the harvesting parameters to plan the entire harvesting process. Since there are many bunches of fruit in cherry tomato, the self-designed end harvesting mechanism of this article has a tomato sleeve, which can separate the target tomato from the tomato bunch and avoid accidentally hurting other fruits when cutting the stem. Finally, it discusses the actual use of harvesting systems to harvest tomatoes of various growth directions in the case of single and multiple tomatoes. The results show the success rate and harvest efficiency of the data-driven method used in this thesis for the harvesting of cherry tomatoes.