結合視覺之人機協同系統研發

Abstract

機器人的應用已漸漸往人機協同作業的領域邁進，結合人類的技術以及機器人的高效率、重複性，不僅增加作業的彈性，更提升了工作的效率。在這種人機共工的合作下，人類與機器人間的距離將較於傳統產業大幅減小，由於工作環境中常會有不可預見的變化發生，若機器人不能即時地做出適當的反應，可能會傷害人類工作者，因此我們必須對於人機協同工作區進行檢測，使機器人認識周邊環境資訊以及未知障礙，以避免碰撞。本論文之目標是以達到使用6軸機器臂系統之製造單元(Work Cell)安全監控之要求，並使用3D感測器來建構一模擬驗證平台，建立人與機器人合作的安全環境，研究內容包括人物體態偵測、追蹤以及機器臂3D姿態資料融合快速運算技術，我們使用Kinect視覺系統即時偵測環境的突發狀況，設計機器人避障策略，實地的建立一套即時虛擬人機操作介面，並以實驗驗證之有效性。

Robot applications nowadays are also going to human-robot collaboration, which not only increases the flexibility, but also its efficiency. However, under the cellular manufacturing module, the distance between human and robot are much shorter than ever. Due to unexpected events that may happen in the working environment, it demands in-time response to avoid damages to the human operator. We thus propose developing an inspection system, so that the robot can recognize its nearby environment and also unknown obstacles. The main research topics include fast computing and fusion methods for the data from human skeleton tracking sensors and 3D robot trajectory. We use Kinect vision system to detect unexpected environmental changes, and also come up with a real-time collision avoidance approach for safe human-robot collaboration. A real-time virtual manipulation system is developed and experimentally validated.