Road Network Validation via Sensor Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing road maps lack the detailed and accurate information required by autonomous and semi-autonomous vehicles, as they often rely on high-level data that may not suffice for safe navigation, and validating highly detailed map data is challenging due to tracing inaccuracies and scaling issues from satellite imagery.
Innovation Solution
The technique involves generating and validating semantic map data using sensor data captured by vehicles, where reference marks are projected into actual image data to align with road network features, allowing for accurate validation and correction of map data without requiring vehicles to navigate through safety-critical environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If highly detailed map data is generated using satellite imagery and tracing methods, then the detail level and information content improve, but the accuracy and reliability deteriorate due to tracing inaccuracies and scaling issues
Solution Approach 1:
The patent creates virtual images by projecting map data into sensor image data, forming a virtual copy that can be compared with actual sensor observations. This virtual copy approach allows validation without physical vehicle traversal, resolving the contradiction between detailed mapping and accuracy verification.
Solution Approach 2:
The patent introduces sensor image data as an intermediary medium between map generation and validation. By projecting map references into sensor images and comparing with actual sensor observations, it creates a validation pathway that maintains both detail and accuracy without requiring risky on-vehicle validation.
2Reliability
If map validation is performed by having vehicles navigate through safety-critical environments, then the accuracy and reliability of map data improve, but the safety risks and resource costs worsen
Solution Approach 1:
The patent performs map validation in advance by comparing projected map references with sensor observations collected during normal operation, before deploying vehicles for safety-critical validation tasks. This preliminary validation ensures map reliability without exposing vehicles to unnecessary risks.
Solution Approach 2:
The patent creates virtual validation scenarios by projecting map data into sensor images and comparing with actual observations, eliminating the need for physical validation trips. This virtual copying approach maintains reliability assessment while removing safety hazards associated with on-vehicle validation.
3Ease of operation
If traditional road maps are used for autonomous vehicle navigation, then the ease of operation and simplicity are maintained, but the manufacturing precision and detail level worsen due to lack of high-detail information
Solution Approach 1:
The patent merges traditional map data structures with detailed sensor observation data, combining the simplicity of conventional maps with the precision of sensor-level detail. This integration maintains ease of operation while achieving manufacturing precision through fused information from multiple sources.
Solution Approach 2:
The patent adds a new dimension of validation by projecting map data into the image space dimension, allowing comparison with sensor observations. This dimensional transformation enables detailed validation without complicating the operational interface, maintaining simplicity while achieving precision.
Data Source
AI summary
Techniques for generating and validating map data that may be used by a vehicle to traverse an environment are described herein. The techniques may include receiving sensor data representing an environment and receiving map data indicating a traffic control annotation. The traffic control annotation may be associated, as projected data, with the sensor data based at least in part on a position or orientation associated with a vehicle. Based at least in part on the association, the map data may be updated and sent to a fleet of vehicles. Additionally, based at least in part on the association the vehicle may determine to trust the sensor data more than the map data while traversing the environment.