HD Map Updating with Heterogeneous Sensor Data Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in generating and maintaining accurate high-definition (HD) maps due to inconsistencies in data gathered from heterogeneous sensors and equipment, which can affect navigation and safety.
Innovation Solution
A machine-learning model is employed to transform sensor data from different vehicles into a common data space, using neural networks to encode and decode data, thereby minimizing discrepancies and enhancing map accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If data are collected by a fleet of vehicles equipped with sensors, then the coverage and quantity of map data are improved, but the accuracy and consistency of the gathered data deteriorate due to differences in data-collection equipment
Solution Approach 1:
The patent introduces a server as an intermediary that receives sensor data from multiple vehicles with different equipment configurations, processes this heterogeneous data through machine learning models, and generates standardized HD map data. This intermediary mediates between the diverse data sources and the final consistent map product, resolving the accuracy issues caused by equipment differences while maintaining the benefits of fleet-wide data collection
Solution Approach 2:
The system transforms sensor data from various vehicles by changing its parameters and representation through machine learning processing. The server converts raw sensor readings with different formats, resolutions, and coordinate systems into a unified parameter set that conforms to standardized HD map specifications, thereby harmonizing data from heterogeneous sources
2Ease of operation
If traditional maps are used for navigation, then the simplicity and ease of operation are improved, but the safety and autonomous driving capability deteriorate due to insufficient detail
Solution Approach 1:
The patent transitions from two-dimensional traditional maps to three-dimensional high-definition maps by adding vertical dimension data such as curb heights, road surface profiles, and overhead obstruction information. This dimensional enrichment provides autonomous vehicles with comprehensive spatial awareness needed for safe navigation while maintaining ease of operation through automated processing
3Measurement precision
If costly and laborious on-the-road data collection is performed, then the accuracy and detail of HD maps are improved, but the time and resources required deteriorate
Solution Approach 1:
The system performs preliminary data aggregation and preprocessing by collecting sensor data from multiple vehicles over time before generating the final HD map. Rather than requiring dedicated survey vehicles to collect all data at once, the system accumulates data from regular fleet operations and processes it systematically, significantly reducing the time and resources needed while maintaining high detail accuracy
Solution Approach 2:
The patent uses sensor data from multiple vehicles as copies of the same physical environment, captured from different perspectives and at different times. By fusing these multiple copies through machine learning, the system reconstructs a single high-accuracy HD map without requiring dedicated survey missions, thereby eliminating the time loss associated with traditional data collection methods
Data Source
AI summary
A method includes a computing system accessing a training sample that includes first sensor data obtained using a first sensor at a first geographic location, and first metadata comprising information relating to the first sensor. The system may train a machine-learning model by generating first map data by processing the training sample using the model and updating the model based on the generated first map data and target map data associated with the first geographic location. The system may then access second sensor data and second metadata, where the second sensor data is obtained using a second sensor. The system may generate second map data associated with a second geographic location by processing the second sensor data and the second metadata using the trained model. A high-definition map may be generated using the second map data.