HD Map Localization Using Predicted Node Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing localization techniques for objects, such as vehicles, face challenges in accurately determining the location of objects within environments due to limitations in positioning sensors and the lack of effective use of structured context modeling information.
Innovation Solution
The system generates a predicted map based on sensor data from the object and matches it with a high-definition (HD) map using a graph neural network, determining node scores to accurately localize the object within the environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional positioning sensor methods are used for localization, then the system is simple to implement, but the localization accuracy is insufficient
Solution Approach 1:
The localization system is segmented into multiple independent modules: sensor data acquisition module, predicted map generation module (using graph neural networks), HD map matching module (using SuperGlue algorithm), and localization determination module. Each module processes specific tasks independently, improving localization accuracy while maintaining manageable system complexity through modular architecture.
Solution Approach 2:
The patent introduces an intermediary predicted map generation step between sensor data and HD map matching. The graph neural network generates a predicted map that serves as an intermediary representation, bridging the gap between raw sensor data and the HD map, thereby improving localization accuracy without directly increasing the complexity of the core matching algorithm.
2Measurement precision
If sensor data alone is used for localization, then the processing is fast, but the localization accuracy is limited by sensor limitations
Solution Approach 1:
The system performs preliminary action by pre-generating a predicted map from sensor data using graph neural networks before the matching process. This predicted map contains pre-processed structural information that accelerates the subsequent HD map matching process, thereby improving localization accuracy without proportionally increasing processing time.
Solution Approach 2:
The patent replaces traditional mechanical positioning sensor methods with a machine learning-based approach using graph neural networks and the SuperGlue algorithm. This substitution transforms the localization process from direct sensor measurement to intelligent pattern recognition and matching, significantly improving accuracy while the optimized algorithm keeps processing time acceptable.
3Measurement precision
If structured context modeling information is not used, then the system is simpler, but the localization accuracy is reduced
Solution Approach 1:
The graph neural network implements local quality by modeling the local structural relationships between map features (nodes and edges) in the predicted map. Each node is processed with consideration of its local connectivity and structural context, improving localization precision through localized structural analysis without requiring complex global modeling of the entire environment.
Data Source
AI summary
Disclosed are techniques for localization of an object. For example, a device can generate, based on sensor data obtained from sensor(s) associated with an object, a predicted map comprising predicted nodes associated with a predicted location of the object within an environment. The device can receive a high definition (HD) map comprising HD nodes associated with a HD location of the object within the environment. The device can further match the predicted nodes with the HD nodes to determine pair(s) of matched nodes between the predicted map and the HD map. The device can determine, based on a comparison between nodes in each pair of the pair(s) of matched nodes, a respective node score for each pair of the pair(s) of matched nodes. The device can determine, based on the respective node score for each pair of the pair(s) of matched nodes, a location of the object within the environment.