Autonomous Vehicle Sparse Mapping for Low-Data Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in navigating due to the vast amounts of data they need to process and store, including visual information, GPS data, and sensor data, which can lead to limitations and adverse effects on navigation, especially with traditional mapping technologies requiring extensive data storage and updates.
Innovation Solution
A system utilizing a sparse map that includes polynomial representations of road features and landmarks, allowing for efficient navigation with reduced data storage and transfer, using image analysis and sensor data from multiple vehicles to generate and update the map, enabling autonomous vehicle navigation with minimal data requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional mapping technology is used to navigate autonomous vehicles, then navigation accuracy can be maintained, but the volume of data needed to store and update the map becomes excessively large
Solution Approach 1:
The patent extracts only the essential navigational elements from complete map data, creating a sparse map that contains only road centerlines and necessary landmarks. This extraction principle removes redundant information while preserving navigation functionality, directly resolving the contradiction between maintaining navigation accuracy and reducing data volume.
Solution Approach 2:
The patent transforms map representation from detailed geometric models to simplified polynomial curves (e.g., third-order polynomials) that describe road centerlines. This parameter change compresses the data representation while maintaining sufficient accuracy for navigation purposes, addressing the data volume issue without sacrificing reliability.
2Reliability
If complete map data is stored and updated continuously, then navigation reliability is improved, but system complexity and computational burden increase
Solution Approach 1:
The system extracts only critical navigational features (road centerlines, key landmarks) from complete map data, creating a simplified sparse map representation. This reduction in data complexity directly lowers system complexity while maintaining navigation reliability through essential information retention.
Solution Approach 2:
Instead of storing complete map data and filtering it during navigation, the patent inverts the approach by storing only the essential sparse representation from the beginning. This inversion simplifies the system architecture by eliminating the need for complex data filtering and processing pipelines.
3Measurement precision
If high-resolution map data is used for navigation, then positioning accuracy is improved, but data transfer requirements and storage needs increase significantly
Solution Approach 1:
The patent represents road geometry using polynomial parameters (coefficients of third-order polynomials) instead of storing detailed geometric point clouds or high-resolution images. This parameter transformation maintains positioning accuracy by preserving essential road shape information while dramatically reducing storage requirements.
Solution Approach 2:
The system creates simplified polynomial curve copies of road centerlines that approximate the actual road geometry. These polynomial copies serve as compact representations that maintain positioning accuracy without requiring storage of the original high-resolution map data.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Systems and methods are provided for autonomous vehicle navigation. The systems and methods may map a lane mark, may map a directional arrow, selectively harvest road information based on data quality, map road segment free spaces, map traffic lights and determine traffic light relevancy, and map traffic lights and associated traffic light cycle times.