The concept of HD maps was first introduced by Mercedes-Benz in 2010 and then contributed to the Bertha Drive project in 2013. In the Bertha Drive project, the Mercedes-Benz S500 completed the Bertha-Benz Memorial Route in a fully autonomous mode, using a highly accurate and information-rich 3D route map, which was later named by a map company called HERE that participated in the project as an HD real-time map. HD maps are highly valued in the field of autonomous driving because they can provide prior knowledge, reduce the difficulty of perception, and make planning targeted. HD maps are not new, and some map and autonomous driving companies have started to produce and use HD maps. However, the production of HD maps and the information and accuracy they contain still have no unified consensus. This chapter mainly introduces the content related to HD maps, where Sect. 6.1 introduces the basic types, levels, formats, and production methods of HD maps, Sect. 6.2 focuses on various generation methods of semantic maps, Sect. 6.3 emphasizes the production methods of lane-level HD maps, and Sect. 6.4 introduces SLAM methods based on deep learning.