基於 ROS 之整合 Pixhawk 飛控板與 Jetson TX2 模組之無人機自主飛行控制系統設計

本論文主旨為，建立一雙層硬體整合系統架構，頂層為機載微型電腦，其負責對於飛行任務之決策與命令下達，而底層為飛行控制電腦，其負責對於多旋翼無人機飛行姿態控制與即時狀態資訊回傳，硬體間數據可雙向傳遞，實現兩硬體之完整整合。系統架構方面，兩硬體透過 UART 通訊協議以線傳方式連接，使用 ROS 系統架構，並配合具 Mavlink 通訊協議之 MAVROS 軟體包，搭建起 NVIDIA Jetson TX2 機載微型電腦與 Pixhawk 飛行控制電腦間通訊橋梁。應用方面，撰寫離線控制程式並運行於機載微型電腦，藉由機載微型電腦具備高效能數據運算能力優勢，分擔影像辨識之資訊處理工作，再透過離線控制程式，將影像辨識結果轉換為控制指令下達飛行控制電腦，由飛行控制電腦進行無人機姿態控制，並回傳當前無人機運作狀態，供機載微型電腦進行控制指令修正等操作，利用本研究之整合系統架構，實現無人機自主飛行控制。驗證方面，以離線控制程式運行軟體在環模擬，與軟硬體整合測試驗證，進行本研究整合系統驗證。軟體在環模擬使用 Gazebo 模擬器進行操作。軟硬體整合測試驗證使用六旋翼無人機作為實驗平台，並將其固定至調機架，透過觀察無人機姿態變化，確定是否依照離線控制程式設計流程運作，驗證本研究之雙層硬體整合系統架構可行性。

關鍵字

ROS ； MAVROS ； Gazebo ；離線控制； NVIDIA Jetson TX2

並列摘要

The thesis focuses on the development of a ROS based integration of a flight control module and an embedded AI computing device for UAV application. The integrated system is arranged in a two-layer architecture when installed on a multi-rotor UAV. The top layer is the NVIDIA Jetson TX2 embedded AI computing module for decision making and command generation for the flight mission. The bottom layer is the Pixhawk flight control board for controlling the motion of the UAV. With real-time status information feedback, data can be bidirectionally tramsmitted between the two computation modules. The two modules are connected by wire transmission using the UART communication protocol. Based upon the ROS system architecture and the MAVROS package, a communication bridge between the NVIDIA Jetson TX2 module and the Pixhawk flight control board is established to ensure that the two computation modules are completely integrated. In the thesis, offboard control programs are designed to perform an image recognition algorithm on the Jetson TX2 module. Depend on the result of the image recognition, flight control command is generated and transmit to the Pixhawk flight control board to control the motion of the UAV so as to complete the flight mission. The software-in-the-loop simulation and the hardware/software integrated system were tested utilizing the designed offboard control programs to verify the proposed system architecture. The software-in-the-loop simulation was performed using the Gazebo simulator. The hardware/software integrated system was verified by using a Hexacopter mounted on a ground-fixed attitude-movable frame as the experimental platform. Both results of the simulation and hardware/software integration tests show the success of the design and demonstrate the usefulness of the proposed system architecture.