High Definition Map
High-definition (HD) maps provide precise, detailed representations of road environments, crucial for autonomous vehicle navigation and safety. Current research focuses on automating HD map creation and maintenance, employing various techniques including deep learning models (e.g., transformers, autoencoders) to process data from diverse sources like aerial imagery, LiDAR, and camera sensors, often incorporating prior map information (e.g., OpenStreetMap) to improve accuracy and efficiency. This work is significant because reliable, up-to-date HD maps are essential for the safe and widespread deployment of autonomous vehicles, impacting both the scientific understanding of automated mapping and the practical development of self-driving technology.
Papers
Towards High-Definition Maps: a Framework Leveraging Semantic Segmentation to Improve NDT Map Compression and Descriptivity
Petri Manninen, Heikki Hyyti, Ville Kyrki, Jyri Maanpää, Josef Taher, Juha Hyyppä
InstaGraM: Instance-level Graph Modeling for Vectorized HD Map Learning
Juyeb Shin, Hyeonjun Jeong, Francois Rameau, Dongsuk Kum