| Summary: | 碩士 === 國立臺灣科技大學 === 機械工程系 === 106 === Nowadays, many researchers work on indoor localization techniques for an autonomous mobile robot. By means of different sensors, setups, and algorithms, they proposed a variety of solutions to reach diverse goals. Focusing on providing a simpler setup and general solution, this thesis proposed a ceiling-feature based navigation system by using a 2D camera to capture ceiling images and localize the robot.
The proposed localization method uses an overhead camera on the robot to capture the image features on the ceiling for localizing itself. The advantage of this method is unsophisticated preprocessing, not affected by tire-skidding and no accumulation errors after executing a period of time. The system could be easily and rapidly set up after a simple initialization process for any different environment and could localize and navigate the robot in the required workspace.
This thesis uses the stitching algorithm to build the panorama of the ceiling information and uses the improved matching algorithm to calculate the robot position and pose on the floor.
A laser projection device is implemented in this work to generate artificial landmarks on the ceiling which lacks features for recognition. With it, the robot can navigate itself in a designated floor space under a ceiling without features.
A number of experiments show the feasibility of the proposed navigation system.
|
|---|
