Vehicle Navigation System Using AI Lane Detection for GPS-Blocked Areas
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing navigation systems for autonomous vehicles rely solely on satellite-based GPS, which becomes unreliable in obstructed areas, posing a risk for accurate location determination and safe maneuvering, especially in tight spaces.
Innovation Solution
A navigation system that uses in-vehicle location sensors and artificial intelligence to determine a vehicle's geographic location by extracting image lines from surrounding images, transforming them into world coordinates, and matching them with high-definition map data to generate a map-related location, even when GPS is blocked.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If satellite-based GPS is used for location determination, then the system is simple and cost-effective, but the reliability deteriorates when GPS signals are blocked by obstructions
Solution Approach 1:
The positioning function is segmented into multiple independent components: GPS receiver for satellite-based positioning, inertial sensors (accelerometers, gyroscopes) for inertial navigation, and vision-based lane detection system. Each component operates independently and can provide positioning information when others fail, particularly the vision system that can determine location by matching detected lane markings with map data when GPS is blocked.
Solution Approach 2:
Multiple positioning methodologies are merged into a unified navigation system that combines satellite-based GPS, inertial navigation system (INS), and vision-based lane detection. The system integrates data from all three sources through sensor fusion algorithms, creating a redundant positioning architecture where if one method fails (e.g., GPS blocked), the others can maintain accurate positioning.
2Reliability
If multiple positioning systems are integrated to improve reliability, then the reliability improves, but the device complexity increases
Solution Approach 1:
The vision-based positioning system serves multiple functions simultaneously: it detects lane markings for navigation, identifies road geometry for route planning, and provides visual confirmation for driver awareness. The same image processing algorithms are used for both lane detection and location verification, reducing the need for separate dedicated systems and lowering overall complexity.
Solution Approach 2:
The system uses the vehicle's existing camera infrastructure (originally intended for other purposes) to capture images for lane detection and positioning. The processor leverages the captured images to automatically determine vehicle location by comparing detected features with pre-stored map data, making the system self-sufficient without requiring additional specialized hardware.
3Measurement precision
If image processing is used to extract lane information, then the positioning accuracy improves in obstructed areas, but the processing time and computational load increase
Solution Approach 1:
High-definition map data is pre-acquired and stored in the system before the vehicle reaches those locations. The map contains pre-processed information about lane markings, road geometry, and distinctive features. When the vehicle needs positioning, the system only needs to compare real-time sensor data against this pre-prepared reference data, significantly reducing processing time compared to generating and analyzing complete maps in real-time.
Solution Approach 2:
The image processing focuses only on specific regions of interest within the captured images, such as lane markings directly ahead and distinctive road features. The system extracts only the necessary geometric and positional information from these localized areas rather than processing the entire image, reducing computational load while maintaining positioning accuracy.
Data Source
AI summary
A method of operation of a navigation system includes: determining a geographic location, of a vehicle, while a global positioning location is blocked by an obstruction and the geographic location is from an in-vehicle location sensor in the vehicle; extracting an image line from a surrounding image with an artificial intelligence model and the surrounding image is from an in-vehicle image sensor in the vehicle; transforming an image coordinate of the image line to a world coordinate of a HD map local storage based on the geographic location; extracting a map line from the HD map local storage based on the world coordinate; determining a pose relationship between the image line and the map line paired as either horizontal or vertical; generating a map-related location for the vehicle based the geographic location and the pose relationship; and communicating the map-related location for displaying on a user interface.