Lane Edge Fusion Using Implicit Curves for Noisy Road Geometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face challenges in accurately fusing map lane edge points and perception lane edge points due to noisy data and varying lateral error variance, especially when the true lane edge forms complex curves that fail the vertical line test.
Innovation Solution
A lane edge fusion system that uses an implicit moving least squares approach to fit an implicit function representing the lane edge as an implicit curve, allowing for the estimation of coefficients and covariance, and iteratively determining the nearest point on the curve to build a fused lane edge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If map lane edge points and perception lane edge points are fused directly, then lane edge estimation is achieved, but the noisy data and varying lateral error variance reduce the accuracy of the fused lane edge
Solution Approach 1:
The patent transforms the lane edge fitting problem by changing the parameterization approach from explicit to implicit functions. By representing the lane edge as an implicit curve F(x,y)=0 rather than y=f(x), the system can handle complex curves that fail the vertical line test. The implicit moving least squares method adjusts the parameters dynamically at each evaluation point, weighting the contribution of map and perception data according to their respective error variances, thereby improving accuracy while handling noisy data.
Solution Approach 2:
The system performs preliminary processing by pre-calculating the covariance matrices for map and perception lane edge points before the fusion process. This preliminary action allows the algorithm to know the reliability of each data source in advance, enabling optimal weighting during the implicit curve fitting process. The covariance information is computed beforehand and stored, so when fitting the implicit function, the system can efficiently combine data sources without repeated calculations.
2Ease of operation
If parametric functions are used to fuse map and perception data, then lane edge points can be determined, but it is challenging to find consistent parameterization with noisy data
Solution Approach 1:
The patent inverts the traditional approach by using implicit functions F(x,y)=0 instead of explicit functions y=f(x). This inversion allows the lane edge to be represented without requiring a functional relationship where y is explicitly defined in terms of x. The implicit representation naturally handles vertical segments and complex curves, eliminating the parameterization challenges associated with noisy data while maintaining ease of operation in the fusion process.
Solution Approach 2:
The implicit function F(x,y)=0 serves as an intermediary between the noisy map and perception data points. Rather than directly parameterizing the lane edge, the implicit function acts as a mediator that captures the geometric relationship between points without requiring explicit parameterization. This intermediary representation allows the system to handle noisy data robustly while maintaining computational efficiency.
3Shape
If the true lane edge forms complex curves that fail the vertical line test, then accurate lane representation is achieved, but it becomes difficult to model the lane edge points as a single function
Solution Approach 1:
The patent resolves the modeling complexity issue by inverting the functional relationship from explicit y=f(x) to implicit F(x,y)=0. This inversion allows complex curves including vertical segments to be represented naturally without requiring multiple functions or piecewise definitions. The implicit representation maintains a single unified model for the entire lane edge geometry, simplifying the overall system complexity while accurately representing complex shapes.
Solution Approach 2:
The implicit function representation provides a universal model that can handle all types of lane edge geometries uniformly. Whether the lane edge is a simple curve, a vertical line, or a complex shape failing the vertical line test, the same implicit function framework F(x,y)=0 applies. This multi-functionality eliminates the need for different modeling approaches for different geometric cases, reducing overall system complexity.
Data Source
AI summary
A lane edge fusion system for an autonomous vehicle includes one or more controllers executing instructions to receive perception data and map data of a roadway the autonomous vehicle is traveling along. The one or more controllers derive a plurality of map lane edge points from the map data and a plurality of perception lane edge points from the perception data and select an evaluation point based on the plurality of map lane edge points and the plurality of perception lane edge points. The one or more controllers fit an implicit function for the evaluation point based on an implicit moving least squares approach. The one or more controllers build a fused lane edge by setting a point on the implicit curve as one of a plurality fused lane edge points.