Local Map Construction with SDMap: A Comprehensive Survey

TL;DR

The paper addresses the challenge of local map perception for intelligent driving by examining how Standard Definition Maps (SDMaps) can serve as low-cost priors to augment perception using multimodal data. It surveys data representations (raster and vector SDMap), encoding strategies, and fusion methods (alignment and cross-attention-based fusion) that integrate SDMap with image and lidar signals to produce BEV maps. The review catalogs public datasets and evaluation metrics for lane extraction and topology reasoning, highlighting key methods such as SMERF, P-MapNet, MapLite 2.0, and MapVision, and discusses challenges in SDMap processing, BEV alignment, and road topology inference. The findings emphasize that SDMap priors can improve long-range perception, occlusion handling, and topology reasoning, driving opportunities for online HD map generation and robust perception under diverse conditions. Overall, the work guides future research toward richer SDMap priors, temporal fusion, and graph-based topology models to enhance robustness and scalability in local map perception.

Abstract

Local map construction is a vital component of intelligent driving perception, offering necessary reference for vehicle positioning and planning. Standard Definition map (SDMap), known for its low cost, accessibility, and versatility, has significant potential as prior information for local map perception. This paper mainly reviews the local map construction methods with SDMap, including definitions, general processing flow, and datasets. Besides, this paper analyzes multimodal data representation and fusion methods in SDMap-based local map construction. This paper also discusses key challenges and future directions, such as optimizing SDMap processing, enhancing spatial alignment with real-time data, and incorporating richer environmental information. At last, the review looks forward to future research focusing on enhancing road topology inference and multimodal data fusion to improve the robustness and scalability of local map perception.

Figures (26)

• Figure 1: The overview of local map construction with SDMap.
• Figure 2: General pipeline of local map construction with SDMap, which contains the processing and fusion flow of the three modal data.
• Figure 3: Sample graph of SDMap from nuScenes dataset.
• Figure 4: This diagram chronologically presents some of the classic local map construction methods using SDMap in recent years, as well as other methods.
• Figure 5: The pipeline of MapTR using self-attention variants and cross-attention variants.