臺灣博碩士論文加值系統

本論文研究單眼視覺微型自動駕駛遙控車原型，該原型以python為基礎，在嵌入式系統Raspberry Pi 3 B+上使用機器學習和電腦視覺驅動，將深度學習應用於自駕遙控車。其使用1:12 RC遙控車為載台、樹莓派進行運算和前置攝像頭建構平台，先透過人為操作搖桿於車道上前進，蒐集影像及搖桿之資訊，將收集到的大量數據來訓練CNN模型，把訓練好的模型傳回樹莓派上，再將車子放入測試環境，計算出當下行進的方向及速度再轉換成PWM訊號，將方向訊號傳至伺服馬達、速度訊號給予電子變速器來控制馬達轉速。並持續做蒐集資料、測試及改善模型，以提升於車道上自主行進之成功率。

This paper study a monocular vision-based autopilot model-car prototype. It is based on python. Using machiane learning and computer vision on the embedded system, Raspberry Pi 3 B+ to apply deep learning to autopilot model-car. It uses a 1:12 RC model-car for the stage, Raspberry Pi calculation and front camera to construct platform. First, move on the driveway by operating the joystick. Collecting information on the image and the joystick. Take a large amount of data to train the CNN model. Transfer the model back to Raspberry Pi, and put the car into the test environment. Calculate the direction and speed of the current travel and then convert it into a PWM signal, send the direction signal to the servo motor, and send the speed signal to the electronic transmission to control the motor speed. Continuing to collect data, test and improve the model to improve the success rate.

摘要 i
ABSTRACT ii
誌 謝 iii
目 錄 iv
表目錄 vii
圖目錄 viii
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究目標 2
1.3 論文架構 2
第二章 文獻探討 3
2.1 微型自動駕駛車輛 3
2.1.1 課程 3
2.1.2 競賽 3
2.1.3 文獻 4
2.2 嵌入式系統 6
2.2.1 概述 6
2.2.2 特點 6
2.2.3 軟硬體架構 7
2.2.4 作業系統 9
2.3 機器學習 10
2.4 卷積神經網路(Convolutional Neural Network，CNN ) 11
2.4.1 卷積（Convolutional Layer) 13
2.4.2 ReLU層 13
2.4.3 池化層（Pooling Layer) 14
2.4.4 標準化層(Normalization Layer) 14
2.4.5 全連接層(Fully-Connected Layer) 15
第三章 系統與設備介紹 16
3.1 系統硬體 16
3.1.1 Raspberry Pi 3 Model B Plus 16
3.1.2 Pi Camera 19
3.1.3 PCA9685 20
3.1.4 Logitech F710 20
3.1.5 HoBao遙控車 21
3.2 系統軟體 22
3.2.1 Python 22
3.2.2 Colaboratory 22
3.2.3 TensorFlow 23
3.2.4 Keras 23
第四章 系統架構與步驟 25
4.1 硬體架構 25
4.1.1 車體組裝 25
4.1.2 車道設計 31
4.2 軟體架構 32
4.2.1 樹莓派初始設定 32
4.2.2 設置Wifi 34
4.2.3 樹莓派環境參數設定 36
4.2.4 連線樹莓派 37
4.2.5 創建項目 38
4.2.6 設定無線搖桿 38
4.3 校準轉向及油門 40
4.3.1 轉向校準 40
4.3.2 油門校準 41
4.4 蒐集數據 42
4.4.1 電腦收集 42
4.4.2 搖桿收集 43
4.5 訓練模型 44
4.5.1 模型訓練 44
4.5.2 轉移訓練 48
第五章 實驗結果與分析 49
5.1 訓練數據 49
5.2 訓練資料類別 50
5.3 CNN模型實驗結果 51
5.3.1 Lenet模型 51
5.3.2 基本模型 53
5.3.3 Maxpooling & BatchNormalization層 57
5.3.4 Dropout層 59
第六章 結論與未來展望 61
6.1 結論 61
6.2 未來展望 62
參考文獻 63
作者簡介 67
附錄 68

[1] MIT RACECAR Website
https://mit-racecar.github.io/?fbclid=IwAR2GE9MEWj48I8j6HlGlZxDHOP_PxpLxUItkKSDYjih9UXTRK9GmXMFVvsI
[2] BARC Website
http://www.barc-project.com/?fbclid=IwAR044xFW2CSUQUAxDIpI7VUWk1ymsfqfiXO4c9tk42UaKA8JSXE_3F6z7_I
[3] F1TENTH Website
http://f1tenth.org/
[4] Vision-Based High Speed Driving with a Deep Dynamic Observer Website
https://arxiv.org/abs/1812.02071
[5] Audi Autonomous Driving Cup Website
https://www.audi-autonomous-driving-cup.com/?fbclid=IwAR0FVFmOomYpJ_apc6duHsN9wwJfH08QuScWOvfzqSQEgcWzGLWbNFMkv3A
[6] Carolo Cup Website
https://wiki.ifr.ing.tu-bs.de/carolocup/en/carolo-cup
[7] University Rover Challenge Website
http://urc.marssociety.org/?fbclid=IwAR3ZE1OTACZyMjpwRhV5H1jrWwXWpSv5rbhnHoJXDdfVv6NYUnNK9T7hclA
[8] T. Do, M. Duong, Q. Dang and M. Le, “Real-Time Self-Driving Car Navigation Using Deep Neural Network,” 2018 4th International Conference on Green Technology and Sustainable Development (GTSD), Ho Chi Minh City, Vietnam, 2018.
[9] A. Jain, “Working model of Self-driving car using Convolutional Neural Network, Raspberry Pi and Arduino,” 2018 Second International Conference on Electronics, Communication and Aerospace Technology (ICECA), Coimbatore, India, 2018.
[10] B. Blaga, M. Deac, R. Al-doori, M. Negru, R. Dǎnescu, “Miniature Autonomous Vehicle Development on Raspberry Pi,” 2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP), Cluj-Napoca, Romania, 2018.
[11] C. Berger, A. M. Md Abdullah, J. Hansson, “COTS-Architecture with a Real-Time OS for a Self-Driving Miniature Vehicle,” SAFECOMP 2013 - Workshop ASCoMS (Architecting Safety in Collaborative Mobile Systems) of the 32nd International Conference on Computer Safety Reliability and Security, 2013.
[12] O. Oyetoke, “Embedded Systems Engineering, the Future of Our Technology World; A Look Into the Design of Optimized Energy Metering Devices,” International Journal of Recent Engineering Science (IJRES), vol. 18, pp. 17-25, 2015.
[13] 嵌入式系統概述
http://www.iot-online.com/art/2012/061420671.html
[14] 嵌入式系統分類
http://www.elecfans.com/emb/xitong/2009061769440.html
[15] 李青旻，基於影像辨識之嵌入式系統設計與實現，碩士論文，義守大學電機工程研究所，高雄，2015。
[16] 嵌入式系統的架構
http://embedhq.org/html/qrslinux/2012/0328/3330.html
[17] 常見的嵌入式系統有哪些
http://www.study-area.org/linux/embedded/articles/Embedded_Linux_Introduction/Embedded_System_Introduction-2.html
[18] 李毅信，機器學習方法在蝴蝶辨識中之比較，碩士論文，國立中央大學數學研究所，桃園，2019
[19] M. Leshno, V. Y. Lin, A. Pinkus, S. Schocken, “Multilayer feedforward networks with a nonpolynomial activation function can approximate any function,” Neural Networks, vol. 6, no. 6, pp. 861-867, 1993.
[20] Y. LeCun, L. Bottou, Y. Bengio, P. Haffner, “ Gradient-based learning applied to document recognition,” Proceedings of the IEEE, vol. 86, pp. 2278-2324, 1998.
[21] Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton, “Imagenet classification with deep convolutional neural networks,” Advances in neural information processing systems, pp. 1097–1105, 2012.
[22] 斎藤康毅。Deep Learning –用Python進行深度學習的基礎理論實作。碁峰資訊股份有限公司，2017。
[23] Y. LeCun, Y. Bengio, G. Hinton, “Deep Learning,” Nature, no. 521, pp. 436-444, 2015.
[24] 台灣樹莓派Raspberry Pi 3 Model B+ 英國UK製產品
https://www.raspberrypi.com.tw/19429/57/
[25] Logitech G F710介紹
https://www.logitechg.com/zh-tw/products/gamepads/f710-wireless-gamepad.html
[26] 介紹Python
https://www.w3schools.com/python/python_intro.asp
[27] TensorFlow
https://www.tensorflow.org/
[28] Keras
https://keras.io/
[29] Donkey car 實作模型設計
https://medium.com/ljlstyle/donkey-car%E5%AF%A6%E5%81%9A-3-%E6%A8%A1%E5%9E%8B%E8%A8%AD%E8%A8%88-3c4281ab4d47