Removing dynamic objects in HD maps for autonomous driving
High-Definition (HD) Maps serve as a critical navigation tool for autonomous driving systems, providing detailed environmental context derived from comprehensive 3D sensor data, including lidars and multi-camera arrays. Despite their invaluable contribution to enhancing vehicle perception capabi...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/177125 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | High-Definition (HD) Maps serve as a critical navigation tool for autonomous driving systems, providing detailed environmental context derived from comprehensive 3D sensor data, including lidars and multi-camera arrays.
Despite their invaluable contribution to enhancing vehicle perception capabilities, HD Maps often encapsulate transient, dynamic objects (e.g., pedestrians, vehicles, and reflective surfaces) that introduce potential inaccuracies in essential autonomous driving functions such as localization and trajectory planning.
The presence of these dynamic entities in HD Maps can significantly compromise the reliability and safety of autonomous navigation systems.
This project proposes the development of an innovative automated pipeline designed to efficiently identify and eliminate such dynamic objects from HD Maps.
By integrating advanced spatial-temporal occupancy analysis, cutting-edge learning-based semantic prediction algorithms, and robust multi-modal sensory data processing, this pipeline aims to refine the accuracy and utility of HD Maps. |
|---|