隨著自駕車發展愈發高階完整，對於高精地圖(High-Definition Maps, HD Maps)的需求更甚以往。高精地圖有別於提供大眾使用之傳統電子地 圖，不僅詳細描述道路資訊，包含道路形狀、拓樸關係、交通號誌等， 對精度更有著較高的限制，為自動駕駛不可或缺的輔助資訊。現今大多 使用移動式雷射掃描系統(Mobile Laser Scanner, MLS) 獲取環境資訊以產 製高精地圖，但目前多為人工對觀測資料進行數化、建置，此過程相當 耗時及費工，而減少生產成本為本研究主要目的。從車載行動測繪系統 (Mobile Mapping System, MMS)觀測資料中萃取出產製高精地圖所需道路 元素，以達到自動化產製及地圖格式轉換。
本研究可分為四個部份，第一部份改善車道線之點雲萃取技術，萃取出屬於車道線之點雲，而第二部分則以萃取成果對其自動化建模形成 完整車道線，利用 OpenDRIVE 格式裡參考線座標系統(Reference-Line Coordinate System)的概念，分類各車道點雲個別進行建模，以三次樣條 插植法(Cubic Spline Interpolation)擬合出完整車道線。在路口部分利用 B樣條插植法(B-Spline Interpolation)擬合虛擬路口車道線。在第三部份利 用先前部分成果參考 OpenDRIVE 官方指引轉換至 OpenDRIVE 格式，以曲率特徵產生相對應之幾何形狀表示各道路主要形狀。除了各路段及路口車道形狀，演算法在第三部份完成道路拓樸連結，獲得正確車道連接邏輯以確保在模擬器內虛擬車輛能夠自動駕駛，達到 OpenDRIVE 最初 核心目的。為了使高精地圖應用至自動駕駛中，需將 OpenDRIVE 轉至 其他高精地圖格式，Lanelet2、Aisan Vector Maps。在第四部份本研究選 取 Lanelet2 為轉出之高精地圖格式，將 OpenDRIVE成果轉換成座標點 (node)並得其相對應關係(way)、(lanelet)。
本研究成果能夠匯入常用檢視軟體，OpenDRIVE Viewer、Virtual Test Drive (VTD)，與 Car Learning to Act(CARLA)模擬器內運作，且產製 OpenDRIVE 成果和 Lanelet2 轉換成果皆利用 ASSURE Map 轉換成 Aisan Vector Map 與經驗證過之高精地圖進行成果比較。在二維成果均方根誤 差小於 20 公分、三維成果均方根誤差小於 30 公分，產製高精地圖總耗 時低於十分鐘，整體而言本研究減少相當的生產成本並且產製精度符合 高精地圖規範的成果。

With the development of autonomous vehicles (AVs) becoming more complete, the demand for High-Definition Maps (HD Maps) is even greater than before. HD maps are different from traditional electronic maps which are designed for people. HD maps not only present information about roads, like the shape of roads, topology relationship, and traffic signs but also are restricted by more severe specifications on accuracy. Owing to these characteristics, they are seen as one of the most significant auxiliary information for AVs. Nowadays, most of the production of HD maps utilize Mobile Laser Scanning systems (MLS) to obtain environmental information. However, most of the data processing including digitalization and construction is manual. The procedure costs lots of time and human resources, so the purpose of this research is to decline the production cost. The main methodology is extracting the road elements which are required for generating HD maps based on the measurements from Mobile Mapping System (MMS) to achieve automated generation and conversion of HD map formats.
This research can be divided into four parts. The first part improves lane lines point cloud extraction technology to obtained correct lane lines point cloud, and the algorithm automatically models complete lane lines based on Cubic Spline Interpolation from the extraction result. The algorithm refers the Reference-Line Coordinate System concept which is officially proposed by OpenDRIVE to cluster the extracted points into different lane lines and model them individually. As for lane lines in junctions, the algorithm adopts B-spline interpolation (Xu et al, 2009) to model due to lack of junction lane lines in the reality. In the third part of the research, the algorithm classifies the reference lines into different geometries based on the features of curvature (Pai et al, 2022). With the generation of main roads and roads in junction, the algorithm also completes the topological connection in the third part. In order to apply HD maps to autonomous driving, it needs to convert OpenDRIVE to other HD map formats, Lanelet2, or Aisan Vector Maps. The author selects Lanelet2 as the conversion format in the fourth part of the research. This research converts the reference lines of roads and other lane lines in the OpenDRIVE format into points with coordinates, defines the relationships, and presents them in the Open Street Map format.
The results of this research can be imported into the commonly used viewing software, OpenDRIVE Viewer (Pagel et al. 2021), and the simulator, Car Learning to Act (CARLA). The production and Lanelet2 conversion are converted into Aisan Vector Maps using ASSURE Map for the result analysis. With the conditions that the root mean square error in 2D and 3D are less than 20 cm and 30 cm, respectively, and the time cost is less than ten minutes, this research not only reduces a considerable production cost but also meets the accuracy specification of HD maps in Taiwan.

Alexey, D., German R., Felipe C., Antonio L., Vladlen K. CARLA: An Open Urban Driving Simulator, Proceedings of the 1st Annual Conference on Robot Learning, PMLR 78:1-16, (2017).
Alshawa, M., Boulaassal, H., Grussenmeyer, P. Acquisition and Automatic Extraction of Facade Elements on Large Sites from A Low Cost Laser Mobile Mapping System, (2009).
Althoff, M., Urban, S., & Koschi, M. Automatic conversion of road networks from opendrive to lanelets. In Proceedings of 2018 IEEE International Conference on Service Operations and Logistics, and Informatics (SOLI), pp. 157-162, (2018).
Autoware Community. Autoware Documentation Release 1.9.0, (2018). Retrieved from
https://buildmedia.readthedocs.org/media/pdf/autoware/feature-documentation_rtd/autoware.pdf
Basnayake, C., Williams, T., Alves, P., Lachapelle, G. Can GNSS Drive V2X. GPS World, (2010).
Bauer, S., Alkhorshid, Y., & Wanielik, G. Using high-definition maps for precise urban vehicle localization. In Proceedings of 2016 IEEE 19th International Conference on Intelligent Transportation Systems (ITSC), pp. 492-497, (2016).
Dai, J. M., Liu, T. A. J., & Lin, H. Y. Road surface detection and recognition for route recommendation. In Proceedings of 2017 IEEE Intelligent Vehicles Symposium (IV), pp. 121-126, (2017).
Daniel D., Filip C., Jakub D., Izzat I., Mateusz K., Mateusz W. Performance of LiDAR object detection deep learning architectures based on artificially generated point cloud data from CARLA simulator, 2019 24th International Conference on Methods and Models in Automation and Robotics (MMAR), (2019).
David, O. Road Data-Modelling for Autonomous Driving, Graz University of Technology, Master’s Thesis, pp. 47-59, (2019).
Demuynck V. What are the differences between the TomTom ADAS and HD Maps. (2019). Retrieved from https://www.tomtom.com/blog/autonomous-driving/adas-map-vs-hd-map/
Du, J. CAMP enhanced digital mapping project—Final report. U.S. Department of Transportation, Washington, (2004). Retrieved from
https://rosap.ntl.bts.gov/view/dot/3704
Gottschalk, S., Manocha, D., & Lin, M. C. Collision queries using oriented bounding boxes. Ph.D. Thesis, University of North Carolina at Chapel Hill, (2000).
HERE. HERE HD Live Map, (2017). Retrieved from
https://www.here.com/sites/g/files/odxslz166/files/2018-11/HERE_HD_Live_Map_one_pager_0.pdf
Hervieu, A., & Soheilian, B. Semi-automatic road/pavement modeling using mobile laser scanning. In Proceedings of ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2, 31-36, (2013).
High-Definition Maps Research Center. HD Maps data contents and formats standard v1.1, (2020).
Hsu, P.C. An Automotive-grade INS/HD Maps/Odometer/GNSS Integration Scheme for Lane-level Navigation Application in Urban Area. Master’s thesis, National Cheng Kung University, Tainan, Taiwan, (2019).
Lin, S. C., Zhang, Y., Hsu, C. H., Skach, M., Haque, M. E., Tang, L., & Mars, J. The architectural implications of autonomous driving: Constraints and acceleration. In Proceedings of 23th International Conference on Architectural Support for Programming Languages and Operating Systems, pp. 751-766, (2018).
Litman, T. Autonomous Vehicle Implementation Predictions Implications for Transport Planning, (2022).
Liu, R., Wang, J., & Zhang, B. High definition map for automated driving: Overview and analysis. The Journal of Navigation, 73(2), 324-341, (2020).
McCrae, J., & Singh, K. Sketching piecewise clothoid curves. Computers & Graphics, 33(4), 452-461, (2009).
McKinley, S., & Levine, M. Cubic spline interpolation. College of the Redwoods, 45(1), 1049-1060, (1998).
National Highway Traffic Safety Administration. Overview of Motor Vehicle Crashes in 2020, (2022).
Pagel, S., Flo, K., Kahraman, M. K., Egan, J. OpenDRIVE Viewer, (2021). Retrieved from https://odrviewer.io/
Pai, H., Y. Automated Modelling of Road for High-Definition Maps with OpenDRIVE Format Using Mobile Mapping Measurements, In Proceedings of ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLIII-B1-2022, 263–270, (2022).
Park, H., & Lee, J. H. B-spline curve fitting based on adaptive curve refinement using dominant points. Computer-Aided Design, 39(6), 439-451, (2007).
PCL. Downsampling a PointCloud using a VoxelGrid filter, (2020). Retrieved from
https://pcl-tutorials.readthedocs.io/en/latest/voxel_grid.html#voxelgrid
PCL. How to use KdTree to search, (2020). Retrieved from
https://pcl-tutorials.readthedocs.io/en/latest/kdtree_search.html#kdtree-search
Pettigrew, S., Talati, Z., Norman, R. The health benefits of autonomous vehicles: public awareness and receptivity in Australia, (2018).
Poggenhans, F., Pauls, J. H., Janosovits, J., Orf, S., Naumann, M., Kuhnt, F., Mayr, M. Lanelet2: A high-definition map framework for the future of automated driving, (2018).
Rehrl, K., Groechenig, S., Piribauer, T., Spielhofer, R., Weissensteiner, P. Towards a standardized workflow for creating high-definition maps for highly automated shuttles, (2022).
Santonato, S., F. A Complete End-to-End Simulation Flow for Autonomous Driving Frameworks, (2021).
Sepp, E. Creating High-Definition Vector Maps for Autonomous Driving, Masters'
Thesis, (2021).
Shi, H. Automatic generation of OpenDrive roads from road measurements. Master's thesis, Chalmers University of Technology, Gothenburg, Sweden, (2011).
TomTom. HD MAP WITH ROADDNA, (2018). Retrieved from
http://download.tomtom.com/open/banners/HD_Map_with_RoadDNA_Product_Info_Sheet.pdf
Wang, S., H., Tu, C., H., Juang, J., C. Automatic traffic modelling for creating digital twins to facilitate autonomous vehicle development, (2021).
Xu, H., Wang, X., Huang, H., Wu, K., & Fang, Q. A fast and stable lane detection method based on B-spline curve. In Proceedings of 2009 IEEE 10th International Conference on Computer-Aided Industrial Design & Conceptual Design, pp. 1036-1040, (2009).
Zeng, J., C. Automated Road-Elements Modelling and Centerline Generation for High-Definition Maps Utilizing 3D Point Cloud, (2020)
Zhang, P. Numerical Solution for the Curvature, (2019).
Zhang, W., Qi, J., Wan, P., Wang, H., Xie, D., Wang, X., and Yan, G. An easy-to-use airborne lidar data filtering method based on cloth simulation, pp. 1-22, (2016).
Zheng, L., Li, B., Yang, B., Song, H., & Lu, Z. Lane-Level Road Network Generation Techniques for Lane-Level Maps of Autonomous Vehicles: A Survey. Sustainability, 11(16), 4511, (2019).