Vehicle Trajectory Correction via Curvature-Weighted Loop Closure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Simultaneous Localization and Mapping (SLAM) techniques for vehicle trajectory estimation suffer from cumulative errors in position changes, which can be reduced by loop closure detection but still result in significant errors within the looped trajectory sections.
Innovation Solution
A method that detects loop closure through image matching and corrects the trajectory by computing curvature to control the distribution of corrections over time, concentrating corrections at high-curvature points, and using a likelihood function with weighted sums to optimize position changes, where weights are reduced at curved sections relative to straight parts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If loop closure detection is used to correct trajectory errors, then cumulative position errors are reduced, but significant errors remain in the looped trajectory sections
Solution Approach 1:
The patent applies local quality by differentiating the treatment of trajectory points based on their local geometric properties. High-curvature points receive different correction handling compared to low-curvature points. Specifically, the method identifies high-curvature regions and applies targeted corrections that preserve the natural turning behavior while correcting drift, whereas low-curvature regions receive corrections that maintain straightness. This localized differentiation resolves the contradiction by improving overall accuracy without compromising loop consistency.
Solution Approach 2:
The patent changes the parameter of curvature along the trajectory to control where corrections are applied. By computing curvature values at each trajectory point and using these as a basis for selective correction, the method dynamically adjusts the correction strategy. High-curvature points are identified and treated differently from low-curvature points, allowing the system to maintain accuracy in turning sections while ensuring consistency in straight sections, thus resolving the contradiction between overall accuracy and loop consistency.
2Measurement precision
If corrections are distributed uniformly over the trajectory, then cumulative errors are reduced, but curvature artifacts appear in straight sections
Solution Approach 1:
The patent applies local quality by making the correction distribution non-uniform based on local curvature characteristics. Instead of applying corrections uniformly across all trajectory points, the method identifies high-curvature regions and concentrates corrections there, while applying minimal or no corrections in low-curvature (straight) sections. This preserves the straightness of linear trajectory segments while still correcting cumulative errors in the overall trajectory, thereby resolving the contradiction between accuracy improvement and shape preservation.
Solution Approach 2:
The patent applies partial action by selectively applying corrections only where needed rather than uniformly across the entire trajectory. By using curvature thresholds to identify high-curvature points and limiting corrections to these specific regions, the method avoids over-correcting straight sections. This partial application of correction action resolves the contradiction by achieving sufficient accuracy improvement without introducing artifacts in straight trajectory segments.
3Shape
If curvature-based selective correction is applied, then curvature artifacts are reduced, but computational complexity increases
Solution Approach 1:
The patent applies partial action by computing curvature only at selected trajectory points rather than continuously across the entire trajectory. By using a threshold-based approach that identifies and processes only high-curvature regions, the method significantly reduces the number of curvature computations required. This selective computation maintains trajectory smoothness in critical regions while minimizing overall computational overhead, thereby resolving the contradiction between shape quality and computational complexity.
Data Source
Figure 1~2
Figure 3
AI summary
Measurements of a trajectory of a vehicle are corrected when it is detected that the vehicle has travelled along a loop, returning to an earlier position. A first estimate of the trajectory of a vehicle is obtained from measurements that provide temporally localized indications of position changes of the vehicle. During travel along the trajectory images are captured from the vehicle. A search is made for images with matching content to detect closure of the loop. A corrected estimate of the trajectory is determined that accounts for the completion of the loop. The distribution density of the correction is at least partly controlled by a measure of curvature, so that the corrections are concentrated at high curvature. Thus, correction artifacts in straight trajectory parts, such as during travel along a street, are avoided. The measure of curvature of the trajectory may be provided using image based measurements or from additional heading measurements.