| Summary: | 碩士 === 國立臺灣科技大學 === 機械工程系 === 101 === Navigation is an important technique for autonomous robots to freely roam in desired environments and successfully perform assigned tasks. As the demands of complicated tasks in various environments keep increasing in service robots, recently cross-floor navigation using vision sensor has drawn many interests from researchers. However, in either elevator or stairway environments, the issue of image dead zones still limits most discussions in local area navigation and requires further investigation. This thesis aims to develop an autonomous cross-floor navigation system for stair-climbing mobile robot. Particularly, wireless sensor modules are installed in the image dead zones so that the robot can adjust and continue its motion accordingly, achieving such long-distance navigation task. By using upward and forward looking camera setups, imaged based navigation is conducted in most road sections. Image features including ceiling landmarks, stair lines, and skirting lines are adopted to detect robot guiding lanes with appropriate image-processing algorithms. To let the robot follow the detected image lanes correctly, a visual servo control driven by image errors and based on car-like mobile robot differential kinematics is also applied for motion control. A self-made stair-climbing robot is finally used to validate the effectiveness of the proposed navigation method.
|
|---|
