Lane-Level Localization via Ground Imagery and Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current transportation management and autonomous driving systems face challenges in achieving accurate lane-level geographic localization due to noisy and erroneous GPS signals, which can lead to inaccurate estimated time of arrival and routing, and are hindered by the high cost and size limitations of high-end GPS equipment.
Innovation Solution
A localization system utilizing a geo-spatial deep convolutional neural network that processes ground images and raw GPS coordinates to determine accurate lane-level geographic locations without the need for 3D or HD maps, leveraging image data from dash cameras or phone cameras and integrating machine-learning models for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If raw GPS signals are used for geographic localization, then the system is simple and low-cost, but the localization accuracy deteriorates due to noise and errors from atmospheric uncertainty, building blockage, and multi-path signals
Solution Approach 1:
The patent introduces an intermediary processing system that receives raw GPS coordinates and refines them using multiple data sources including barometric pressure sensors, accelerometers, and map matching algorithms. This intermediary layer filters and corrects GPS errors without requiring expensive high-end GPS hardware, thereby improving localization accuracy while maintaining device simplicity.
Solution Approach 2:
The patent combines multiple sensing modalities (GPS, barometric pressure, accelerometer data, and map information) into a composite localization system. By fusing these different data types through algorithms, the system achieves high accuracy localization that would otherwise require expensive specialized GPS equipment, resolving the contradiction between accuracy and device complexity.
2Measurement precision
If high-end GPS equipment is used to improve localization accuracy, then measurement precision improves, but device size and cost increase making it impractical for mobile devices
Solution Approach 1:
The patent segments the localization function into multiple independent components: GPS receiver, barometric pressure sensor, accelerometer, and map matching processor. Each component is a standard off-the-shelf element that can be integrated into mobile devices without increasing size. The segmented architecture allows high accuracy localization through software processing rather than relying on a single large high-end GPS unit.
Solution Approach 2:
The patent makes the mobile device's existing sensors serve multiple functions: the barometric pressure sensor is used for both altitude measurement and GPS error correction, the accelerometer serves for navigation and GPS signal validation. This multi-functionality allows the system to achieve high localization accuracy using standard mobile device components, eliminating the need for specialized large-sized GPS equipment.
3Ease of operation
If standard mobile device GPS is used, then device size and cost are acceptable, but localization accuracy deteriorates leading to incorrect ETA and routing calculations
Solution Approach 1:
The patent implements feedback mechanisms where the system continuously monitors localization accuracy by comparing GPS-derived position with expected positions from map data and barometric altitude information. When GPS errors are detected (such as during building blockage or multi-path conditions), the system automatically adjusts by weighting alternative sensors more heavily, providing continuous correction that maintains accuracy while using standard mobile GPS hardware.
Data Source
AI summary
In one embodiment, a method includes receiving an image associated with an object in an environment, the image being captured by sensors associated with a vehicle, generating a feature representation of the image, determining a potential ground control point associated with the object based on the feature representation of the image, determining a predetermined location reading based on the potential ground control point, calculating a differential relative to the predetermined location reading based on the potential ground control point, and determining a location of the vehicle based on the differential and the predetermined location reading based on the potential ground control point.