Autonomous Driving Map Updates Using Exit-Event Data Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current high-precision maps used in autonomous driving are static and slow to update, leading to safety hazards as they do not adapt to real-world changes.
Innovation Solution
A method and device for updating a map by determining target vehicles exiting autonomous driving mode, acquiring their driving data in non-autonomous mode, clustering the data, and updating the autonomous driving enabling state of each location point on the map.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the high-precision map uses a static operational design domain (ODD), then the map data remains stable and consistent, but the map cannot adapt to real-world changes and update speed is very slow
Solution Approach 1:
The patent transforms the static ODD into a dynamic one by continuously collecting driving data from vehicles, clustering the data to identify new drivable scenarios, and automatically updating the map's operational design domain. This allows the map to adapt to real-world changes while maintaining timely updates through automated processing pipelines.
Solution Approach 2:
The system establishes a feedback loop where driving data from vehicles is continuously collected, analyzed, and used to update the map. The updated map is then redistributed to vehicles, creating a closed-loop system that continuously improves the map's adaptability based on real-world performance and emerging scenarios.
2Adaptability or versatility
If the map data is updated frequently to reflect real-world changes, then the adaptability improves, but the system complexity and data processing requirements increase
Solution Approach 1:
The system enables vehicles to automatically contribute their own driving data to the map update process. Each vehicle acts as both a consumer and contributor of map data, self-organizing the collection and transmission of driving information without requiring complex centralized data gathering infrastructure.
Solution Approach 2:
Instead of processing all raw driving data centrally, the system uses clustering algorithms to identify representative patterns and scenarios from the data. These clustered representations serve as simplified copies that capture the essential new drivable scenarios, reducing the complexity of map updates while maintaining accuracy.
3Adaptability or versatility
If the operational design domain is expanded to cover more scenarios, then the autonomous driving capability is enhanced, but the safety risks increase due to insufficient map coverage
Solution Approach 1:
The system performs preliminary validation by clustering driving data to identify and verify new drivable scenarios before incorporating them into the operational design domain. This pre-screening process ensures that only well-validated scenarios are added to the map, maintaining safety while expanding capability coverage.
Solution Approach 2:
The patent replaces manual map validation and update processes with automated clustering algorithms and machine learning systems. This substitution enables rapid, consistent, and scalable validation of new driving scenarios, improving both the speed and reliability of map updates while reducing human error.
Data Source
AI summary
A method for updating a map includes: determining target vehicles exiting an autonomous driving mode on a target road, acquiring driving data of the target vehicles in a non-autonomous driving mode, generating at least one cluster by clustering the driving data, and updating an autonomous driving enabling state of each location point of the target road on the map according to the at least one cluster. The autonomous driving enabling state is configured to indicate whether a passing vehicle is allowed to enter the autonomous driving mode at the location point.