Arm-SLAM:機器手臂與SLAM結合之應用

Abstract

本論文提出了Arm-SLAM 框架，它將同步定位與建圖 (SLAM)與機器手臂整合在一起。Arm-SLAM 框架使用 SLAM 創建為機器手臂提供的環境地圖。此外，Arm-SLAM框架包括改進的虛擬阻抗控制方法，並設計了無標記視覺伺服方法，以確保機械手臂的安全性和準確性。Arm-SLAM框架及其組件，包括改進的虛擬阻抗控制和無標記視覺伺服，透過三個實驗結果進行了演示。在第一個實驗中，改進的虛擬阻抗控制證明它可以成功地避開靜態和動態障礙物。在第二個實驗中，與使用ArUco標記的視覺伺服方法相比，無標記視覺伺服方法具有類似的重複性，但誤差為1.5公分，比基線方法低1.5公分。此外，無標記視覺伺服方法具有足夠的穩健性，可以應對手眼校準方法帶來的誤差。在最後的實驗中，ArmSLAM框架證明了它可以在複雜的環境下成功運作。

This thesis proposes the Arm-SLAM framework, which integrates Simultaneous Localization and Mapping (SLAM) with a robot manipulator. The Arm-SLAM framework uses the SLAM to create the map of the environment provided for the robot manipulator. Moreover, the Arm-SLAM framework includes a modified virtual impedance control method and designed a markerless visual servoing method to ensure the safety and accuracy of the robot manipulator. The Arm-SLAM framework and its components, including the modified virtual impedance control and markerless visual servoing, are demonstrated through three experimental results. In the first experiment, the modified virtual impedance control proved that it could successfully avoid both static and dynamic obstacles. In the second experiment, compared with the visual servoing method using the ArUco marker, the markerless visual servoing method has similar repeatability, but the error is 1.5 cm, which is 1.5 cm lower than the baseline method. Moreover, the markerless visual servoing method is robust enough to face the error caused by the hand-eye calibration method. In the final experiment, the ArmSLAM framework proved that it can successfully work under a complex environment.