| Summary: | 碩士 === 國立臺灣科技大學 === 自動化及控制研究所 === 101 === In recent years, Unmanned Aerial Vehicles (UAVs) have experienced a strong boost in performance, especially in environment monitoring, disaster surveillance and target search. However, the flat-wing aircraft cannot hover and provide enough resolution at high altitude because of the mechanical structure. Although ground mobile robots can navigate their way by using sonar and laser, they could not get the global position.
This thesis proposed joint operation architecture between aerial and ground mobile robots. The hovering aerial mobile robot provides visual sensing of the ground facts as the localization and mapping and followed by the collision free and shortest path generation to navigate the ground mobile robot to arrive the target in real time. The experiment includes modeling and analysis of the PID hovering attitude controller of the aerial mobile robot for achieve it will be stability hovering and has higher localization accuracy which is 23.87 cm in xy plane and 32.79 cm in z axis. The experiment in the visual sensing and perception system include foreground and background segmentation, and the detection of robot pose, target and obstacle position. The position accuracy error was 19.23 mm and the average computing time in the real-time system only spent 126.33 ms for each image frame. The visibility graph is implemented building the robot configuration space and followed by applying the Dijkstra’s algorithm to find the shortest and collision free path between the robot and target. Eventually, P3DX with visual servo control arrive the destination, and its average error was 85.89 cm.
|
|---|
