自駕車用高精地圖快速製圖技術

在自駕車、智慧城市應用中，高精地圖扮演相當關鍵的角色，它負責提供精確、豐富環境資訊。在實務應用上，高精地圖需要適時的更新及擴張以提供最新的環境資訊和應用範圍，但受限於高價的硬體設備及大量的人工作業流程，國內既有的圖資產製及更新流程速度緩慢，難以規模化。為達到快速更新及擴張需求，本團隊試圖評估低成本設備與半自動化的流程的綜合成效。因此，本團隊通過工研院機械所自製的低成本光達測繪設備來對桃園國際機場場域進行案例分析，並透過本文展示只使用國內測繪業者10%成本的硬體設備及20%的向量地圖產製時間，來建置自駕車用的高精地圖，並驗證本系統產生的圖資更新能符合內政部對於高精地圖的精度規範。透過這些成果，本團隊描繪一個建置臺灣高精地圖的願景，結合既有高成本且高精度的地圖產製流程與本文提出低成本快速建圖流程，一起協作創建出兼具精度與新鮮度並包覆全臺灣的高精地圖。

關鍵字

高精地圖；光達點雲；自駕車

並列摘要

HD maps provide rich, cm-accurate information about an environment that is crucial for self driving cars and other smart city applications. But efforts in Taiwan to expand HD map coverage and provide timely updates are severely limited by current map-making methods, due to reliance on high-cost equipments and manual workflows. This work evaluates the use of ITRI's low-cost equipment and semi-automated workflows for accelerating HD map-making. We perform a case study, using ITRI's low-cost Lidar mapping system to create an HD map of Taoyuan International Airport. We show that at just 10% the equipment cost and 20% the vector map production time of current Taiwan mapping companies, ITRI's system can create high quality HD maps proven to work on self-driving cars and can also be used to provide highly accurate updates to existing HD maps (i.e. base map). Based on these results, we articulate a vision for HD mapping in Taiwan, where emerging low-cost mapping solutions collaborate alongside existing processes (high-cost but more accurate) to create an island-wide HD map that is both fresh and accurate.