本研究提出了一個基於自主感知和決策的無人機任務系統，使用六軸無人機搭配雷射測距和視訊鏡頭進行感知，並使用 Jetson Xavier NX 實現YOLOv5預訓練模型進行物體檢測。該系統可以自主感知周圍的環境，並檢測環境障礙物以實現避障。為了實現更好的任務執行，本研究引入了禁航區檢測技術。無人機可以透過檢測環境障礙物，並在任務執行過程中對障礙物劃定禁航區，將禁航區座標丟入 D*Lite 演算法進行路徑重劃待避讓完成後回到原始路徑上，從而實現高效且安全的任務執行。此外，該系統還建立了障礙物資訊通報系統使得不同機群間能進行障礙物資訊的流通。本研究的實驗結果顯示，該系統可以快速、準確地感知周圍的環境，並且能夠實現物體檢測和禁航區檢測。此外，禁航區檢測技術可以幫助無人機實現高效且安全的任務執行。

This research proposes an autonomous perception and decision-making unmanned aerial vehicle (UAV) mission system, utilizing a six-axis UAV equipped with laser rangefinders and video cameras for sensing. The system employs Jetson Xavier NX to implement the YOLOv5 pre-trained model for object detection. The UAV can autonomously perceive its surroundings and detect environmental obstacles to achieve obstacle avoidance.

To achieve improved task execution, this research introduces No Fly Zone detection technology. Unmanned aerial vehicles (UAVs) can detect environmental obstacles and designate No Fly Zones during mission execution. The coordinates of the No Fly Zones are then fed into the D*Lite algorithm for path replanning to avoid the obstacles. Once the avoidance is completed, the UAVs return to their original paths, thus achieving efficient and safe task execution. Additionally, the system establishes an obstacle information dissemination mechanism, enabling different UAV groups to exchange obstacle information.

The experimental results of this study show that the system can quickly and accurately perceive the surrounding environment, and is capable of object detection and no-navigation zone detection. In addition, no-fly zone detection technology can help UAVs achieve efficient and safe mission execution.

摘要.............................................................................................................................................i
Abstract.......................................................................................................................................ii
誌謝...........................................................................................................................................iii
目錄...........................................................................................................................................iv
表目錄.......................................................................................................................................vi
圖目錄......................................................................................................................................vii
符號說明...................................................................................................................................ix
第一章 緒論..............................................................................................................................1
1.1 研究背景 .........................................................................................................................1
1.2 研究動機 .........................................................................................................................2
1.3 研究目的 .........................................................................................................................3
1.4 論文架構 .........................................................................................................................4
第二章 文獻探討.......................................................................................................................5
2.1 YOLO 視覺深度學習演算法........................................................................................5
2.1.1 YOLO 歷史 ....................................................................................................................5
2.1.2 無人機 YOLO 辨識避障與感測器避障相關應用........................................................7
2.2 無人機之即時任務路徑重劃演算法.............................................................................7
2.3 無人機地面站系統開發.................................................................................................8
第三章 任務情境與系統介紹 .................................................................................................11
3.1 任務情境.......................................................................................................................11
3.2 YOLOV5介紹 ...............................................................................................................12
3.3 YOLOV5深度視覺模型訓練過程 ...............................................................................13
3.4 YOLOV5深度視覺模型預測 .......................................................................................15
3.5 TensorRT 介紹..............................................................................................................16
3.6 TensorRT 優化後實際使用比較..................................................................................17
3.7 先行者偵測無人機任務執行.......................................................................................18
3.7.1 測距感測器搭配 YOLOv5進行障礙物推算流程.......................................................18
3.7.2 先行者偵測機任務程式流程.......................................................................................19
3.7.3 上傳禁航區資訊...........................................................................................................20
3.7.4 後行者群組無人機自主飛行任務...............................................................................21
3.7.5 禁忌搜尋法...................................................................................................................21
3.7.6 A*搜尋演算法.............................................................................................................22
3.7.7 D*lite 避障演算法與 A*避障演算法比較 .................................................................24
第四章 研究內容與方法 .........................................................................................................25
4.1 群組無人機系統架構..................................................................................................25
4.2 硬體介紹與整合..........................................................................................................25
4.2.1 機架..............................................................................................................................27
4.2.2 Pixhawk 4 飛行控制器...............................................................................................27
4.2.3 Jetson Xavier NX 邊緣計算系統................................................................................29
4.2.4 XBee 無線通訊模組 ...................................................................................................30
4.2.5 無線控制系統..............................................................................................................30
v
4.2.6 全球定位系統..............................................................................................................31
4.2.7 電源系統......................................................................................................................32
4.2.8 感測器介紹..................................................................................................................33
4.3 軟體開發環境介紹......................................................................................................35
4.4 地面站監控系統..........................................................................................................35
4.5 Python 無人機控制函式庫–DroneKit.........................................................................36
4.6 命令式地面站–MAVProxy..........................................................................................36
第五章 實驗規劃與結果.........................................................................................................37
5.1 實驗規劃......................................................................................................................37
5.2 障礙物感測實驗..........................................................................................................39
5.2.1 測距感測器量測實驗..................................................................................................39
5.2.2 量測距離對畫面中框選對應實體長度......................................................................40
5.3 先行者偵測無人機......................................................................................................43
5.3.1 先行者偵測機模擬......................................................................................................43
5.3.2 先行者偵測機實際驗證..............................................................................................45
5.3.3 先行者偵查無人機與過去研究比較..........................................................................48
5.4 後行者機群無人機......................................................................................................48
5.4.1 後行者機群禁忌演算法及 A*避障模擬....................................................................48
5.4.2 後行者機群禁忌演算法及 A*避障實驗....................................................................51
5.4.3 後行者機群即時路徑重劃模擬..................................................................................53
5.4.4 後行者機群即時路徑重劃實驗..................................................................................54
第六章 結論與建議 .................................................................................................................56
6.1 結論 ...............................................................................................................................56
6.2 建議 ...............................................................................................................................57
參考文獻..................................................................................................................................58

[1] 陳柏宇，“群組機器人路徑規劃及失效後之路徑補償研究”，國立虎尾
科技大學自動化工程系碩士論文，2017。
[2] 陳彥儒，“群組無人機於非確定環境下之及時任務路徑重劃”，國立虎
尾科技大學自動化工程系碩士論文，2021。
[3] 無人機物聯網——分佈式無人機交通管理的多層低空空域模型，
2023.4
https://www.researchgate.net/publication/336995119_Internet_of_Unmanned_Aerial_Vehicles-A_Multilayer_Low-Altitude_Airspace_Model_for_Distributed_UAV_Traffic_Management
[4] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, "You only look once:
Unified, real-time object detection," IEEE Conference on Computer Vision
and Pattern Recognition, pp. 778–789, 2016.
[5] J. Redmon and A. Farhadi, " YOLOv3: An Incremental Improvement", 2018.
[6] A.Bochkovskiy, Chien-Yao Wang, Hong-Yuan Mark Liao" YOLOv4: Optimal Speed and Accuracy of Object Detection", 2020.
[7] 楊忠道，“基於物體偵測實現之無人機避障系統”，國立陽明交通大學
2017。
[8] 林祐弘，“無人飛行系統基於機器學習於避障與目標追隨之研究”，國
立中興大學，2017。
[9] 盧柏瑋，“多軸飛行器的障礙物偵測與避障系統”，國立中央大學，
2017。
[10]陳昭宇，“基於即時深度預估使無人機在未知環境中自主避障”，國立
陽明交通大學，2021。
[11]李聞撰，“應用 YOLO 深度學習影像辨識技術之自動導引車系統”，國
立彰化師範大學，2021。
[12]劉晏呈，“機器學習應用於割草機避障”，國立雲林科技大學 2022。
[13]Sven Koenig, Maxim Likhachev, David Furcy, “Lifelong Planning A*”,
Computer Science Department, University of Southern California, Los Angeles, 2001.
[14]Anthony Stentz, “Optimal and Efficient Path Planning for Partially-known
Environments”, IEEE International Conference on Robotics and Automation,
1994.
[15]Sven Koenig, Maxim Likhachev, “D* Lite”, College of Computing, Georgia
59
Institute of Technology, Atlanta, 2002.
[16]莊明龍，“4G-LTE 群組無人機系統建立”，國立虎尾科技大學自動化工
程系碩士論文，2018。
[17]吳盛佑，“群組無人機4維任務規劃系統建立”，國立虎尾科技大學自動
化工程系碩士論文，2019。
[18]李俊穎，“無人機視覺之即時路徑重劃及障礙物地理資訊建立”，國立
虎尾科技大學自動化工程系碩士論文，2022。
[19]TensorRT 官方網站，2023.2
https://developer.nvidia.com/tensorrt
[20]Quadcopter Garage，DJI F450 機架圖片，2022.5
http://quadcoptergarage.com/dji-flame-wheel-f450-review/
[21]PX4 官方網站，Pixhawk 4 飛控制器接口說明圖片，2022.7
https://docs.px4.io/v1.11/zh/flight_controller/pixhawk4.html
[22]U-blox 官方網站，NEO-M8N 全球定位模組規格，2022.5
https://www.u-blox.com/en/product/neo-m8-series