Vehicle Absolute Position Determination Using Odometry and Map Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the absolute initial position of a vehicle, especially in environments without global positioning systems like parking garages, are inaccurate and require precise knowledge of landmark orientations and dimensions.
Innovation Solution
A method involving relative navigation and odometry to record a vehicle's trajectory, detection of at least three objects with non-intersecting detection beams, and comparison with a digital map using a generalized PnP method to determine the absolute initial position.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If classic PnP method with AR markers is used for localization, then initial pose can be determined, but measurement precision deteriorates for distant markers and requires precise knowledge of marker orientation and dimensions
Solution Approach 1:
The patent extracts only the essential geometric information (2D image coordinates of marker corners) needed for localization, discarding the need for precise 3D marker parameters such as orientation and dimensions. This extraction approach simplifies the input requirements while maintaining localization capability.
Solution Approach 2:
The patent changes the parameter set from requiring precise 3D marker parameters (orientation, dimensions) to using only 2D image coordinates. This parameter transformation enables the system to work with simpler, more easily obtainable data while achieving accurate localization through the proposed optimization method.
2Adaptability or versatility
If P3P method is used for pose estimation, then initial pose can be determined with fewer points, but the method cannot be generalized to multi-camera systems because lines of sight must converge at a single point
Solution Approach 1:
The patent develops a generalized PnP method that works universally for both single-camera and multi-camera systems. The optimization framework accommodates multiple cameras by treating each camera's detection beams independently, removing the geometric constraint that lines of sight must converge at a single point and enabling broad applicability across different camera configurations.
3Measurement precision
If 3D reconstruction method is used for localization, then vehicle position can be determined, but tracking must be performed over extended time periods which increases loss of time
Solution Approach 1:
The patent uses pre-stored 3D map data of landmarks as prior knowledge, eliminating the need for real-time 3D reconstruction. By having the 3D information prepared in advance, the system can perform rapid localization by comparing current 2D detections with the pre-existing 3D map, significantly reducing the time required for position determination.
Data Source
Figure 1
Figure 2~4
Figure 5
AI summary
The absolute position of a vehicle is to be determined. For this purpose, the vehicle is moved along a trajectory (1) starting from its initial absolute position. The trajectory (1) is recorded by odometry, starting from a defined initial position (2) that corresponds to the absolute initial position. Three objects (O1, O2, O3) are detected from the vehicle while it is on the trajectory (1). A detection beam (3) is captured for each object. The trajectory (1) and the detection beams (3) are compared with a map. Finally, the absolute initial position is determined from this comparison.