Lane-Topology Reasoning Using Standard-Definition Navigation Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving systems face challenges in accurately recognizing lane topologies due to inadequate sensor inputs and the high cost and complexity of high-definition (HD) maps.
Innovation Solution
The method involves encoding a standard definition (SD) map representation with road-level topology into an encoded format, which is then used by a lane-topology model for reasoning, enhancing the accuracy of lane detection and connectivity inference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If HD maps are used for lane topology recognition, then measurement precision is improved, but cost increases significantly
Solution Approach 1:
The patent substitutes expensive HD maps with cheaper SD maps that can be more easily updated and maintained. The system uses multiple inexpensive sensor inputs (cameras, LiDAR) to compensate for the lower precision of SD maps, achieving acceptable lane topology recognition without the high cost of HD map acquisition and continuous updates.
Solution Approach 2:
The patent combines multiple sensor inputs (camera images, LiDAR data, GPS information) with SD map data to achieve accurate lane topology recognition. By fusing these diverse data sources, the system compensates for the lower precision of SD maps and achieves results comparable to HD maps without the associated costs.
2Measurement precision
If HD maps are used for lane topology recognition, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent adopts SD maps which have simpler annotation requirements and lower update complexity compared to HD maps. The system processes multiple sensor inputs to achieve the necessary precision, avoiding the complex manual annotation and continuous update processes required for HD maps.
Solution Approach 2:
The system uses automated sensor-based lane detection and map matching to maintain accuracy without requiring manual HD map updates. The multiple sensors continuously capture road conditions and automatically adjust lane topology recognition, eliminating the need for complex human-in-the-loop map maintenance.
3Device complexity
If sensor inputs alone are used for lane topology recognition, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent combines multiple sensor inputs (cameras, LiDAR, GPS) with SD map data to achieve accurate lane topology recognition. This fusion of data sources compensates for the limitations of individual sensors and SD maps, providing robust lane topology recognition without requiring complex HD maps.
Solution Approach 2:
The system uses a multi-functional sensor suite where cameras provide visual lane markings, LiDAR provides 3D spatial information, and GPS provides location context. Each sensor serves multiple functions and compensates for the weaknesses of others, achieving high precision lane topology recognition with a relatively simple integrated system.
Data Source
AI summary
In the context of autonomous driving, the recognition of lane topologies is required for the vehicle to make well-informed and prudent decisions such as lane changes, navigation through intricate intersections, and smooth merging. Current autonomous driving systems rely solely on sensor (e.g. camera) inputs to recognize lane topology. As a result, poor sensor data will have a direct negative impact on lane topology recognition. The present disclosure augments lane topology reasoning with a standard definition navigation map for use in autonomous driving applications.