近年來，自動駕駛技術的研究逐漸興起，車輛定位被視為該領域的重點研究項目之一。全球定位系統（Global Positioning System, GPS）為目前最普及的車輛定位系統，其優勢在於可以快速地透過衛星定位取得車輛的絕對座標，且價格便宜容易取得。但由於環境的差異，GPS訊號會受到環境地形等影響從而導致定位誤差的不同，若定位誤差較大可能導致自動駕駛車輛有安全上的疑慮。為了解決此問題，現有許多合作式鄰近車輛定位系統，該系統透過與相鄰車輛交換訊息，或裝載額外的傳感裝置輔助並提升其GPS精準度。然而，現有方法大多只針對單一環境進行驗證以提升其車輛定位精準度。在這篇論文中，我們引入了轉移學習（Transfer Learning, TL）的核心概念，透過其架構來定義不同環境間之關係，並找出多個環境共同的特徵以同時優化多個環境的車輛定位效能。此外，在現實中我們無法保證可以索取各種環境之資料數據，因此透過TL可以使我們在保持一定效能的標準下有效的減少對車輛資料數量的需求。最後模擬結果表明所提出之系統架構與傳統方法相比，不僅效能提升，所需的資料量也大幅減少。

In recent years, research on autonomous driving technology has gradually emerged, and vehicle positioning has been regarded as one of the key research projects in this field. The Global Positioning System (GPS) is currently the most popular vehicle positioning system. Its advantage is that it can quickly collect the absolute coordinates of the vehicle through satellite positioning, and its price is cheap and easy to obtain. However, due to environmental differences, environmental terrain will affect GPS signals, resulting in different positioning errors. If the positioning error is large, it may cause safety concerns for autonomous vehicles. To solve this problem, there are many cooperative vehicle positioning systems, which exchange information with neighboring vehicles or install additional sensing devices to assist and improve their GPS accuracy. However, most of the existing methods only perform verification in a single environment to improve the accuracy of vehicle positioning. In this paper, we introduce the concept of transfer learning（TL）, define the relationship between different environments through its architecture, and find the common features of multiple environments to optimize the performance of vehicle positioning in multiple environments at the same time. In addition, we cannot guarantee that we can obtain various environmental data in reality. Therefore, through TL, we can effectively reduce the demand for vehicle data while maintaining a certain performance standard. Finally, the simulation results show that compared with the traditional method, the proposed system architecture not only improves the performance but also greatly reduces the amount of data required.

論文摘要..........................................i
Abstract.........................................ii
謝誌..............................................iv
目錄..............................................v
圖目錄............................................vii
表目錄............................................ix
一、 緒論.........................................1
1.1 研究背景......................................1
1.2 研究動機與目的.................................2
1.3 論文架構......................................3
二、文獻探討.......................................4
2.1 基於濾波器之車輛定位技術........................4
2.2 基於機器學習之車輛定位技術......................5
2.3 綜合觀點......................................5
三、系統模型.......................................7
3.1 合作式車輛定位系統.............................9
3.2 環境模型......................................10
3.3 問題闡述......................................10
四、本論文所提出之演算法...........................12
4.1 圖卷積網路....................................12
4.2 轉移學習......................................13
4.2.1 轉移學習之方法..............................14
4.2.2 域自適應....................................15
4.3 DANN架構......................................17
4.3.1 DANN總覽....................................17
4.3.2 特徵提取器..................................18
4.3.3 座標預測器..................................19
4.3.4 域分類器....................................19
4.3.5 對抗機制....................................20
4.3.6 方法總結....................................21
五、模擬結果與分析.................................22
5.1 模擬設置......................................22
5.2 模擬性能結果..................................24
5.2.1 車輛直行....................................25
5.2.2 車輛變換車道................................32
六、 結論與貢獻....................................42
參考文獻..........................................43

[1] K. Liu, H. B. Lim, E. Frazzoli, H. Ji, and V. C. S. Lee, “Improvingpositioning accuracy using gps pseudorange measurements for cooperative vehicular localization,” IEEE Transactions on Vehicular Technology, vol. 63, no. 6, pp. 2544–2556, 2014.
[2] S. Nam, D. Lee, J. Lee, and S. Park, “Cnvps: Cooperative neighbor-ing vehicle positioning system based on vehicle-to-vehicle communication,” IEEE Access, vol. 7, pp. 16 847–16 857, 2019.
[3] M. A. Hossain, I. Elshafiey, and A. Al-Sanie, “High accuracy gps-free vehicular positioning based on v2v communications and rsu-assisted doa estimation,” in 2017 9th IEEE-GCC Conference and Exhibition(GCCCE), 2017, pp. 1–5.
[4] P.-P. Mao, X. Ji, X. Qu, Z. Yi, and B. Ran, “Deep learning based vehicle position estimation for human drive vehicle at connected freeway,” in 2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC), 2020, pp. 1–6.
[5] S. J. Pan and Q. Yang, “A survey on transfer learning,” IEEE Transactions on Knowledge and Data Engineering, vol. 22, no. 10, pp. 1345–1359, 2010.
[6] H. Guan, L. Li, and X. Jia, “Multi-sensor fusion vehicle positioning based on kalman filter,” in2013 IEEE Third International Conference on Information Science and Technology (ICIST), 2013, pp. 296–299.
[7] S. Rezaei and R. Sengupta, “Kalman filter-based integration of dgps and vehicle sensors for localization,” IEEE Transactions on Control SystemsTechnology, vol. 15, no. 6, pp. 1080–1088, 2007.
[8] M. Rohani, D. Gingras, V. Vigneron, and D. Gruyer, “A new decentralized bayesian approach for cooperative vehicle localization based on fusion of gps and inter-vehicle distance measurements,” in2013 Inter-national Conference on Connected Vehicles and Expo (ICCVE), 2013, pp. 473–479.
[9] K. Jo, K. Chu, and M. Sunwoo, “Interacting multiple model filter-based sensor fusion of gps with in-vehicle sensors for real-time vehicle positioning,” IEEE Transactions on Intelligent Transportation Systems, vol. 13, no. 1, pp. 329–343, 2012.
[10] E. Wang, W. Zhao, and M. Cai, “Research on improving accuracy of gps positioning based on particle filter,” in2013 IEEE 8th Conference on Industrial Electronics and Applications (ICIEA), 2013, pp. 1167–1171.
[11] B. Zhao, X. Gao, J. Liu, J. Zhao, and C. Xu, “Spatiotemporal data fusion in graph convolutional networks for traffic prediction, ”IEEE Access, vol. 8, pp. 76 632–76 641, 2020.
[12] S. Carrasco, D. F. Llorca, and M. Ángel Sotelo, “Scout: Socially-consistent and understandable graph attention network for trajectory prediction of vehicles and vrus,” 2021.
[13] B. Jiang, Z. Zhang, D. Lin, and J. Tang, “Graph learning-convolutional networks,” 2018.
[14] J. Yao, A. T. Balaei, M. Hassan, N. Alam, and A. G. Dempster, “Improving cooperative positioning for vehicular networks,” IEEE Transactions on Vehicular Technology, vol. 60, no. 6, pp. 2810–2823, 2011.
[15] F. de Ponte Müller, E. M. Diaz, and I. Rashdan, “Cooperative position-ing and radar sensor fusion for relative localization of vehicles,” in 2016 IEEE Intelligent Vehicles Symposium (IV), 2016, pp. 1060–1065.
[16] S. B. Cruz, T. E. Abrudan, Z. Xiao, N. Trigoni, and J. Barros, “Neighbor-aided localization in vehicular networks,” IEEE Transactions on Intelligent Transportation Systems, vol. 18, no. 10, pp. 2693–2702, 2017.
[17] Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and P. S. Yu, “A comprehensive survey on graph neural networks,” IEEE Transactions on Neural Networks and Learning Systems, vol. 32, no. 1, pp. 4–24, 2021.
[18] Y. Ma, J. Hao, Y. Yang, H. Li, J. Jin, and G. Chen, “Spectral-based graph convolutional network for directed graphs,” arXiv, 2019.
[19] V. M. Patel, R. Gopalan, R. Li, and R. Chellappa, “Visual domain adaptation: A survey of recent advances,” IEEE Signal Processing Magazine, vol. 32, no. 3, pp. 53–69, 2015.
[20] Y. Ganin, E. Ustinova, H. Ajakan, P. Germain, H. Larochelle, F. Laviolette, M. Marchand, and V. Lempitsky, “Domain-adversarial neural net-works,” Journal of Machine Learning Research, 2015.