Map-Based Ground Truth Labeling for Fresh HD Autonomous Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional maps for autonomous vehicles lack precision and accuracy, and the process of creating and updating high-definition maps is expensive and time-consuming, making it difficult to provide up-to-date and safe navigation data.
Innovation Solution
A system that generates high-definition maps by combining sensor data from autonomous vehicles to create a point cloud, labels objects, and uses this data to train deep learning models for accurate navigation, allowing for real-time updates and efficient storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional maps are created using survey teams with high resolution sensors, then map accuracy and precision are improved, but the cost and time required to create and update maps increases significantly
Solution Approach 1:
The patent uses standard vehicles equipped with sensors to collect map data, creating copies of the mapping capability that don't require expensive specialized survey equipment. Multiple standard vehicles can simultaneously collect and contribute map data, enabling continuous updates without the bottlenecks of traditional survey teams.
Solution Approach 2:
The system enables vehicles to automatically contribute their sensor data to map creation and updates without requiring manual surveying. The automated pipeline processes sensor data from standard vehicles, extracts map features, and integrates them into the HD map, eliminating the need for human survey teams while maintaining high accuracy.
2Reliability
If survey teams frequently update maps to reflect road changes, then map freshness is improved, but the cost and complexity of maintenance increases
Solution Approach 1:
Standard vehicles serve multiple functions: they perform their primary transportation role while simultaneously collecting map data. This eliminates the need for dedicated survey vehicles and allows the same vehicle fleet to both transport passengers/goods and maintain map accuracy, reducing overall system complexity.
Solution Approach 2:
Vehicles automatically contribute their sensor data to map updates without requiring specialized survey equipment or manual intervention. The automated processing pipeline handles data collection, feature extraction, and map integration, enabling continuous map freshness with minimal operational complexity.
3Area of stationary object
If GPS systems are used for vehicle location, then coverage area is improved, but location accuracy deteriorates to over 100 meters error
Solution Approach 1:
The patent combines GPS data with visual data from cameras and processed map features to create a hybrid localization system. This fusion of multiple data sources maintains the broad coverage of GPS while achieving centimeter-level accuracy through visual odometry and feature matching, resolving the accuracy limitation of GPS alone.
Data Source
AI summary
Systems and methods related to ground truth data generation using maps and sensor data are disclosed. In some embodiments, a label corresponding to a feature included in a map may be assigned to an image based at least on the feature also being depicted in the image. In these and other embodiments, the labeled image may be used as training data (e.g., ground truth data) for one or more neural networks.