Road Shape Classification Map Matching for Autonomous Vehicle Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in precise positioning due to under-constrained shape matching errors when using optimization-based map matching techniques for straight and curved lanes, leading to reduced accuracy and reliability in position estimation.
Innovation Solution
A high precision position estimation method is developed using road shape classification-based map matching, which classifies lanes into straight, curved, and clothoid curves, applying geometric feature-based matching for straight and curved lanes and optimization-based matching for clothoid curves, and utilizes an extended Kalman filter to estimate vehicle position, thereby eliminating under-constrained shape matching errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If optimization-based map matching technique (ICP) is applied to all road shapes, then the system is simple to implement, but matching error occurs due to under-constrained shape for straight and curved lanes
Solution Approach 1:
The patent segments the road shape matching process into three distinct categories: straight lanes, curved lanes, and clothoid curve lanes. Each segment type is matched using an appropriate algorithm - geometric feature-based matching for straight and curved lanes, and optimization-based ICP for clothoid curves. This segmentation eliminates the under-constrained shape matching error by applying the right method to each road geometry type.
2Measurement precision
If geometric feature-based map matching is applied to straight and curved lanes, then matching error is eliminated, but computational complexity increases
Solution Approach 1:
The patent applies different matching algorithms with appropriate computational complexity to different road segments. Geometric feature-based matching, which is computationally efficient, is used for straight and curved lanes where it provides sufficient accuracy. Optimization-based ICP, which is computationally intensive, is reserved only for clothoid curve lanes where it is truly needed. This local differentiation reduces overall computational complexity while maintaining high accuracy where required.
3Measurement precision
If road shape classification is performed before map matching, then positioning precision is improved, but processing time increases
Solution Approach 1:
The patent performs road shape classification as a preliminary action before map matching. By classifying the lane shape (straight, curved, or clothoid) in advance, the system can select the most appropriate and efficient matching algorithm for each case. This preliminary classification enables the system to avoid using computationally intensive optimization-based matching for simple straight and curved lanes, thereby reducing overall processing time while maintaining high positioning precision.
Data Source
AI summary
A high precision position estimation method through a road shape classification-based map matching implemented by a high precision position estimation system of an autonomous vehicle includes classifying a road into a straight lane, a curved lane, and a clothoid curve lane by the lane matching between a sensor lane segment and a map lane segment, generating a movement amount value calculated by applying a geometric feature-based map matching or an optimization-based map matching to the classified lane as a covariance matrix, and calculating the latitude, longitude, and traveling direction of a vehicle to be estimated by combining the covariance matrix with sensor measurement values of a vehicle speed and a yaw rate by an extended Kalman filter.