Self-Generated Map Localization for GNSS-Limited Automated Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Self-driving systems face challenges in accurately estimating subject-vehicle positioning attitudes on high-precision maps without expensive GNSS receivers, especially in environments with satellite signal shielding or reflection, and existing solutions lack information like lane centerlines and traffic rules.
Innovation Solution
A self-position estimation device that uses sensors to measure the environment, detects characteristic information, and generates a self-generated map to improve estimation accuracy, including lane centerlines and traffic rules, without relying on preceding vehicles or expensive GNSS receivers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GNSS receivers are used to estimate subject-vehicle positioning attitudes, then positioning accuracy is improved, but device cost increases
Solution Approach 1:
The patent creates a virtual copy of the high-precision map by generating a self-generated map containing positioning information from multiple sources. This copied map data replaces the need for expensive GNSS receivers, allowing the system to achieve high positioning accuracy through software-based map matching rather than hardware-based satellite reception.
Solution Approach 2:
The system uses multiple low-cost positioning methods (odometry, visual odometry, inertial sensors) that can be easily replaced or updated, replacing the need for expensive, long-lived GNSS hardware. These cheaper positioning techniques are combined through fusion to achieve accuracy comparable to high-end GNSS receivers.
2Measurement precision
If GNSS receivers are used for positioning, then positioning accuracy is improved, but reliability deteriorates in shielded environments
Solution Approach 1:
The patent merges multiple independent positioning methods (odometry, visual odometry, inertial navigation) into a unified positioning system. This fusion approach ensures that when one method fails (such as GNSS in tunnels), other methods can compensate, maintaining positioning reliability in shielded environments where satellite signals are unavailable.
Solution Approach 2:
The system dynamically changes positioning parameters by switching between different positioning modes based on environmental conditions. In open areas, it may rely more on GNSS-like accuracy methods, while in shielded environments, it transitions to odometry and visual odometry-based positioning, adjusting the weighting and fusion parameters of different sensing modalities to maintain reliability.
3Measurement precision
If a complete high-precision map is generated and stored, then positioning accuracy is improved, but storage usage increases
Solution Approach 1:
The patent segments the high-precision map into multiple sections or layers, storing only the portions that are currently needed for positioning. The self-generated map is divided into road segment data, landmark data, and positioning information, allowing selective storage and loading of map portions based on the vehicle's current location and navigation needs.
Solution Approach 2:
The system applies local quality by storing high-precision map data only in areas where the vehicle is currently operating or is expected to operate soon. Map data quality and detail level are adjusted locally based on relevance, with higher precision stored for active areas and lower precision or compressed data for distant areas, reducing overall storage requirements while maintaining positioning accuracy in the current context.
4Measurement precision
If comprehensive map data is processed, then positioning accuracy is improved, but processing load increases
Solution Approach 1:
The patent extracts only the essential positioning-relevant features from comprehensive map data, such as road centerlines, intersection points, and key landmarks, rather than processing all map information. This extraction of critical positioning elements reduces the computational burden while maintaining the accuracy needed for vehicle localization.
Solution Approach 2:
The system performs partial processing by focusing computational resources on processing only the map portions and features that are currently relevant for positioning, rather than continuously processing the entire map database. This selective processing approach maintains positioning accuracy by updating only the necessary map segments while reducing overall processing load and power consumption.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An object is to provide a self-position estimation device capable of highly accurately estimating self-location and attitudes on a map containing information such as lane centerlines, stop lines, and traffic rules. A representative self-position estimation device according to the present invention includes a self-position estimation portion that estimates a self-location and attitude on a high-precision map from measurement results of a sensor to measure objects around a vehicle; a low-precision section detection portion that detects a low-precision section indicating low estimation accuracy based on the self-location and attitude estimated by the self-position estimation portion; and a self-map generation portion that generates a self-generated map saving a position and type of the object on the high-precision map in the low-precision section detected by the low-precision section detection portion.