Mapless NOA as a New Standard for Autonomous Driving
“Mapless NOA” has become a key solution for modern autonomous driving systems. It reduces reliance on high-definition offline maps (HD maps), the development and maintenance of which are complex and resource-intensive. The “mapless” approach means moving from pre-prepared maps to real-time map building with subsequent evolution into “world models.” ADAS algorithms increasingly rely on data rather than rigid rules.
This mapless solution, similar to early SLAM technologies, dynamically creates a vector map and matches it with lightweight offline maps (LD maps) for simultaneous localization and navigation. While classical SLAM mostly relied on LiDAR, its role has diminished with the introduction of BEV. Nevertheless, SLAM is still used in specific conditions, such as underground parking lots.
Stages of Map Evolution for Autonomous Driving
Until 2022: focus on HD maps with geometric accuracy; ADAS worked based on rigid rules.
2023–2024: spread of LD maps focusing on topology, semantics, and data relevance; development of Mapless NOA.
After 2025: emergence of next-generation maps with 3D Gaussian blur and NeRF (neural radiance fields). Cartography begins to predict the future, not just reflect the past.
World models identify spatiotemporal patterns in large data from cameras, LiDAR, and crowdsourcing in real time. They dynamically update the environment representation, form road topology, interpret traffic rules and semantics, and most importantly — forecast future events and system behavior.
Key Trends in Cartography
Automated and low-cost mapping
Transition to vector HD maps in real time
Development of MapTR and VectorMapNet technologies as the foundation for future cartography.