臺灣博碩士論文加值系統

本論文研究目的建立一套無人機電力續航力預估系統，先將預規劃好的路徑任務上傳至雲端，使無人機出發前至雲端選一條任務路徑，任務完成後將此趟飛行電壓數據上傳至雲端，進行深度學習回歸預測出剩餘電壓可完成任務之路徑長，再由剩下的任務路徑選出一條可完成的路徑，並能在安全的電壓範圍內完成任務。因此無人機在未來進行任務規劃時可以考慮到電力續航的問題，使無人機用最節省成本及能源的方式來完成任務。
為了達到將電量消耗達到最佳化，本文分別使用NN神經網路、K-鄰近演算法(K-Nearest Neighbor)及隨機森林(random decision forests)來對電量進行回歸預測，同時將飛行時間、風速及飛行距離加入，利用大量歷史飛行數據來進行訓練，並比較各深度學習的準確性及優缺點。

Create a set of drone electrical renewal estimation systems for the research purpose of this thesis. First upload the preserved path task to the cloud, so that the drone will choose a task path before the departure of the drone. After the task is completed, the flight voltage data is completed Upload to the cloud, then return to the deep learning predicts that remaining voltage can complete the length of the task, then select a complete path from the remaining task path, and can complete the task within a safe voltage range. Therefore, the drone can consider the problem of power battery life when planning the mission in the future, so that the drone can use the most cost -saving and most effective way to complete the task.
In order to achieve the optimization of the consumption electricity, this article uses NN neural networks, K-Nearest Neighbor and Random Decision Forests to reward the power to predict, and at the same time, the flight time, wind speed and flight will be Join the distance, use a large amount of historical flight data for training, and compare the accuracy and advantages and disadvantages of deep learning.

摘要......i
Abstract......ii
誌謝......iii
目錄......iv
表目錄......vi
圖目錄......vii
符號說明 ......x
第一章 緒論......1
1.1 研究背景......1
1.2 研究動機......1
1.3 研究目的......2
1.4 論文架構......2
第二章 文獻探討......4
2.1 民航機航行燃油預測......4
2.2 電動車續航里程預測......6
第三章 續航力估測......10
3.1 電壓推導數學式分析......10
3.2 人工神經網路......11
3.2.1 人工神經網路原理......11
3.2.2 人工神經網路訓練過程......15
3.2.3 人工神經網路模型參數比較......17
3.3 K-近鄰演算法......25
3.3.1 K-近鄰演算法原理......25
3.3.2 K-近鄰演算法參數說明......26
3.4 隨機森林......28
3.4.1 隨機森林原理......28
3.4.2 隨機森林參數說明......29
3.5 三種深度學習模型比較......30
第四章 實驗用無人載具......32
4.1 無人機系統架構......32
4.2 硬體元件架構與整合......32
4.2.1 Pixhawk 4飛行控制器......33
4.2.2 DJI F450機架......36
4.2.3 XBee無線數據通訊模組......36
4.2.4 Jetson Xavier NX邊緣計算系統......37
4.2.5 GPS全球定位系統......38
4.2.6 無線控制系統......38
4.2.7 電力系統......39
4.2.8 動力系統......40
4.3 開發環境介紹......41
4.4 地面監控系統......42
4.5 程式架構流程圖......43
第五章 實驗與結果......45
5.1 實驗流程與規劃......45
5.2 A起B降測試風向之影響......46
5.3 地面站監視模擬測試......47
5.3.1 路線A自主飛行模擬......48
5.3.2 路線B自主飛行模擬......49
5.3.3 路線C自主飛行模擬......49
5.3.4 路線D自主飛行模擬......50
5.4 全系統實地驗證......50
5.4.1 實驗(一)路徑A自主飛行......52
5.4.2 實驗(一)路徑B自主飛行......53
5.4.3 實驗(一)路徑C自主飛行......54
5.4.4 實驗(一)路徑D自主飛行......55
5.4.5 實驗(二)路徑A自主飛行......56
5.4.6 實驗(二)路徑B自主飛行......58
第六章 結論與未來展望......60
參考文獻......61

[1] 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。
[2] Dmytro S. Komarchuk, et al. “Forecasting the State of Charging Batteries on Board the UAV on the Basis of Neuro-Fuzzy Network Using”, IEEE,2021.
[3] 大疆社區，MINI2操作介面。https://bbs.dji.com/forum.php?mod=viewthread&tid=256571
[4] Leonard Bauersfeld, Davide Scaramuzza, “Range, Endurance, and Optimal Speed Estimates for Multicopters”, IEEE,2022.
[5] Zhaoyue Zhang, et al. “Research on Flight Fuel Prediction based on Historical Data Mining”, IEEE,2019.
[6] 截自Youtube Choyce TW的影片，https://www.youtube.com/watch?v=6N6kMQ1yLos&ab_channel=ChoyceTW
[7] Laurent Thibault, et al. “A Unified Approach for Electric Vehicles Range Maximization via Eco-Routing, Eco-Driving, and Energy Consumption Prediction”, IEEE,2018.
[8] JUAN P. ORTIZ, et al. “Continual Refoircement Learning Using Real-World Data for Intelligent Prediction of SOC Consumption in Electric Vehicles”, IEEE,2022.
[9] Qiyan Yan, Yanning Wang, “Predicting for power battery SOC based on neural network”, IEEE,2017.
[10] Li Wei, et al. “Method and system for estimating remaining range of electric car, and electric car”, Google Patents,2017.
[11] IBM Cloud Education，人工神經網路基本架構，
https://www.ibm.com/tw-zh/cloud/learn/neural-networks
[12] Ken Huang，人工神經網路 Artificial Neural Network之圖片，https://medium.com/%E4%BA%BA%E5%B7%A5%E6%99%BA%E6%85%A7-%E5%80%92%E5%BA%95%E6%9C%89%E5%A4%9A%E6%99%BA%E6%85%A7/%E4%BA%BA%E5%B7%A5%E7%A5%9E%E7%B6%93%E7%B6%B2%E8%B7%AF-artificial-neural-network-38bca19f189b
[13] 深度學習筆記，激活函數圖示，https://codingnote.cc/zh-tw/p/176736/
[14] PX4官方網站，Pixhawk 4控制器接口說明圖片，
https://docs.px4.io/v1.11/zh/flight_controller/pixhawk4.html
[15] TE官方網站，MS5611氣壓高度計圖片，
https://www.te.com/usa-en/product-CAT-BLPS0036.html
[16] Bosch Sensortec官方網站，BMI055六軸姿態感測器圖片和規格，
https://www.bosch-sensortec.com/products/motion-sensors/imus/bmi055/
[17] 意法半導體，LSM303D電子羅盤圖片，
https://www.st.com/en/mems-and-sensors/lsm303d.html
[18] Quadcopter Garage，DJI F450機架圖片，
http://quadcoptergarage.com/dji-flame-wheel-f450-review/
[19] U-blox官方網站，NEO-M8N 全球定位模組規格，
https://www.u-blox.com/en/product/neo-m8-series
[20] T-motor官方網站，T-motor Air 2216 KV880馬達規格，
https://store.tmotor.com/goods.php?id=727