High-Definition Map (HD Map)

A High-Definition Map (HD Map), also known as a high-precision map, is a dedicated map designed for autonomous driving systems. It reconstructs the road environment in 3D digital form with centimetre-level accuracy, providing essential data support for beyond-visual-range sensing, precise localisation, and path planning for autonomous vehicles.

Definition and Core Characteristics

There is a fundamental difference between HD maps and conventional navigation maps designed for human drivers. Traditional maps offer metre-level accuracy, provide "road-level" navigation information, and cater to human route planning needs; whereas HD maps achieve centimetre-level precision (typically 10–20 cm), deliver "lane-level" navigation information, and serve as structured data for autonomous driving systems to "read" and "interpret".

In terms of data dimensions, HD maps include not only fundamental road geometry such as road shape, gradients, and curvature, but also detailed semantic information such as lane marking types, lane widths, traffic sign locations, and traffic light timing, alongside dynamic variables such as active roadwork zones. This "quantitative" descriptive capability allows autonomous driving systems to obtain prior information in advance—such as slip road locations, corner curvatures, and inclines that onboard sensors cannot detect beyond the line of sight—thereby securing more time for perception and decision-making, and compensating for the limits of the system's real-time sensing capabilities.

Production and Application

The production of HD maps relies on professional surveying and mapping equipment, typically carried out by survey vehicles equipped with multi-beam LiDAR, RTK-level GNSS, Inertial Measurement Units (IMU), and high-resolution cameras. The collected data undergoes post-processing procedures, including point cloud registration, semantic annotation, and road topology generation, to form structured map data. Given that mapping involves sensitive geospatial information, producing HD maps in China requires obtaining a Class-A Navigation Electronic Map Surveying and Mapping Qualification.

At the application level, HD maps serve as a key technical pillar for Level 3 (L3) and higher autonomous driving. Vehicles achieve precise lane-level localisation by matching onboard GPS, IMU, LiDAR, or camera data with the map database; simultaneously, the map provides prior road information for path planning and vehicle control, complementing real-time sensors to enhance safety redundancy in challenging conditions such as adverse weather.

Industry Landscape and Evolution Trends

In 2021, China's HD map market reached RM646,000,000, characterised by high market concentration with major players such as Baidu, NavInfo, and AutoNavi dominating the sector. However, HD maps face significant hurdles, including high data collection costs, challenging updates and maintenance, and strict regulatory restrictions. High equipment costs and labour-intensive manual annotation, combined with frequent road network changes that make it difficult to guarantee data freshness, render a business model entirely reliant on HD maps unsustainable across wide-area urban road networks.

In recent years, the industry has been transitioning from early "map-heavy, perception-light" solutions towards "map-light, perception-heavy" approaches. Systems such as Tesla FSD and Huawei ADS are reducing their reliance on pre-mapped HD data, turning instead to more powerful onboard perception capabilities to achieve real-time mapping. Lightweight map (LD Map) solutions—which balance topology, semantics, and data freshness—continue to play a critical role in structured scenarios such as highways, whilst effectively complementing vision-centric perception systems across complex urban road networks.