GNSS Data Validation Using Multi-Sensor Confidence Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurate navigation due to the limitations of Global Navigation Satellite System (GNSS) accuracy and sensor data, particularly in environments where GNSS signals are obscured or unreliable, leading to potential safety issues and inefficiencies in map data usage.
Innovation Solution
A system that combines GNSS location data with additional sensor data such as acceleration, image, and LIDAR data to generate a probabilistic model and confidence scores, validating location data for improved navigation and HD map updates using machine learning models and sensor fusion techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GNSS location data is used for navigation, then vehicle location can be determined globally, but the accuracy is insufficient (3-5 meters typical, over 100m with errors) for safe autonomous navigation requiring 10 cm or less accuracy
Solution Approach 1:
The patent combines GNSS location data with multiple sensor data sources (IMU, cameras, LIDAR, radar) to create a fused location estimate. The system integrates these diverse data sources through sensor fusion algorithms to achieve the required 10 cm or less accuracy for safe autonomous navigation, overcoming the limitations of GNSS alone.
Solution Approach 2:
The patent introduces map data and sensor fusion algorithms as intermediary processing layers between raw GNSS data and final navigation decisions. These intermediaries refine the coarse GNSS location data into precise location estimates by comparing with map features and cross-validating with sensor measurements.
2Loss of information
If conventional maps are used for navigation, then some environmental information can be obtained, but the accuracy level (typically meter-level) is insufficient for safe autonomous vehicle operation requiring 10 cm or better precision
Solution Approach 1:
The patent uses preliminary HD map data as a reference framework before actual navigation. The system pre-processes and stores high-precision map information including lane geometries, traffic signs, and road features, which are then used to validate and refine real-time sensor measurements to achieve 10 cm or better location accuracy.
Solution Approach 2:
The system implements feedback loops where sensor measurements are continuously compared against map data, and location estimates are refined based on discrepancies. This feedback mechanism allows the vehicle to correct its position relative to map features, achieving the required precision for safe navigation.
3Measurement precision
If multiple sensor data sources are integrated to improve location accuracy, then navigation precision can be enhanced to 10 cm or less, but the system complexity increases significantly
Solution Approach 1:
The patent segments the complex sensor fusion system into modular functional components: data acquisition modules for each sensor type, preprocessing modules for individual sensor data, fusion algorithms for integrating multiple sources, and validation modules for quality control. This segmentation manages complexity while achieving 10 cm or better location accuracy.
4Reliability
If real-time validation of GNSS data is performed using sensor fusion, then location reliability improves for safe navigation, but processing time and computational load increase
Solution Approach 1:
The patent applies partial validation by focusing computational resources on validating only the most critical location data points and using selective sensor fusion. The system performs excessive validation on GNSS data when conditions warrant (e.g., urban canyons, tunnels) while using simpler methods when GNSS is already reliable, optimizing the balance between reliability and processing time.
Data Source
AI summary
A vehicle computing system validates location data received from a Global Navigation Satellite System receiver with other sensor data. In one embodiment, the system calculates velocities with the location data and the other sensor data. The system generates a probabilistic model for velocity with a velocity calculated with location data and variance associated with the location data. The system determines a confidence score by applying the probabilistic model to one or more of the velocities calculated with other sensor data. In another embodiment, the system implements a machine learning model that considers features extracted from the sensor data. The system generates a feature vector for the location data and determines a confidence score for the location data by applying the machine learning model to the feature vector. Based on the confidence score, the system can validate the location data. The validated location data is useful for navigation and map updates.