Vehicle Pose Estimation Using Visual, Inertial, and Wheel Odometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional vehicle localization systems, particularly in GPS-deprived environments like tunnels or urban areas with skyscrapers, suffer from inaccuracies and lack robustness due to limited satellite visibility.
Innovation Solution
A drive device and vehicle system that integrates inertial measurement units, wheel odometry, and electromagnetic sensors to estimate vehicle pose by associating feature points across images, utilizing inertial and positional data for enhanced accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPS-based vehicle localization is used, then the system is simple and cost-effective, but the localization accuracy deteriorates in GPS-deprived environments such as tunnels and urban areas with skyscrapers
Solution Approach 1:
The patent combines multiple sensor systems (inertial measurement unit, wheel odometry unit, and electromagnetic sensor unit with visual odometry) into an integrated localization system. This merging allows the system to maintain reliable pose estimation in GPS-deprived environments by fusing data from multiple sources, each compensating for the limitations of the others.
Solution Approach 2:
The localization system uses a composite approach by integrating heterogeneous sensor types (inertial sensors, odometry sensors, and visual electromagnetic sensors) into a unified system. This composite sensor fusion architecture enables the system to achieve reliable localization in challenging environments where single-sensor systems fail.
2Adaptability or versatility
If visual odometry is used alone for pose estimation, then the system can work without GPS, but the measurement precision deteriorates due to feature point association errors
Solution Approach 1:
The system implements feedback by using inertial measurement unit data and wheel odometry data to validate and correct visual odometry results. The inertial data provides feedback on vehicle acceleration and orientation changes, while wheel odometry provides feedback on distance traveled, both of which are used to verify and refine the pose estimates derived from visual feature point associations.
Solution Approach 2:
The patent merges visual odometry with inertial navigation and wheel odometry to create a robust multi-sensor fusion system. This combination allows the system to maintain high measurement precision by cross-validating results across different sensor modalities, compensating for the weaknesses of visual odometry alone.
3Measurement precision
If multiple sensor data sources are integrated for pose estimation, then the localization accuracy improves, but the device complexity increases
Solution Approach 1:
The patent segments the localization system into distinct functional modules: an inertial measurement unit for acceleration and orientation data, a wheel odometry unit for distance and speed data, and an electromagnetic sensor unit for visual feature extraction. This segmentation allows each component to be optimized independently while maintaining overall system precision.
Solution Approach 2:
The system achieves multi-functionality by designing a processor that can handle and fuse data from multiple different sensor types (inertial, odometry, and visual sensors). This universal processing capability allows the system to maintain high precision pose estimation while managing the complexity of multiple sensor inputs through a unified computational framework.
Data Source
AI summary
This invention refers to vehicle (10) comprising an inertial measurement unit (26) configured to output inertial related data, a wheel odometry unit (28) configured to output position related data, an electromagnetic sensor unit (20) configured to output first image data and second image data that is previous image data of the first image data, and a visual odometry part (31) configured to receive the inertial related data, the position related data, the first image data, and the second image data and to output an estimated pose of the vehicle (10) computed by associating a first feature point extracted from the first image data with a second feature point extracted from the second image data and evaluating the association of first feature point and second feature point based on the inertial related data from the inertial measurement unit (26) and the position related data from wheel odometry.