Geometry-Aware Vector Map Generation With Convex Polygon Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current map generation models for autonomous vehicles often produce inaccurate representations of environmental features, such as irregular-shaped pedestrian crossings and randomly-shaped center lines and road boundaries, which hinder safe navigation.
Innovation Solution
A map generation model is trained using a loss function that includes polygon and polyline losses to promote convexity and smoothness, ensuring that features are represented as convex polygons and smoothly varying polylines, thereby improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional deep-learning methods are used to generate maps from sensor data, then the map generation process can be completed, but the accuracy and precision of the generated maps deteriorate
Solution Approach 1:
The patent introduces a loss function with specific mathematical parameters that enforce geometric constraints on generated maps. The loss function includes terms that penalize non-convex polygons and irregular polyline shapes, transforming the optimization objective to produce maps with guaranteed geometric properties such as convex pedestrian crossings and smooth road boundaries.
Solution Approach 2:
The patent implements a feedback mechanism through the loss function that continuously evaluates and corrects generated map features. By calculating polygon convexity and polyline smoothness during training, the system receives feedback signals that guide the neural network to produce geometrically accurate representations, iteratively improving map quality until convergence.
2Manufacturing precision
If traditional deep-learning methods are used for map generation, then processing can be performed, but the geometric accuracy of features like pedestrian crossings and road boundaries deteriorates
Solution Approach 1:
The patent modifies the training objective by introducing geometric constraint parameters in the loss function. Specifically, the loss includes terms that measure polygon convexity (ensuring all internal angles are ≤ 180 degrees) and polyline curvature smoothness, which transform the neural network's output from arbitrary shapes to geometrically regular forms.
Solution Approach 2:
The patent applies curvature-based constraints to ensure that generated features have smooth, regular shapes. The loss function penalizes sharp turns and irregular curvatures in polylines representing road boundaries and center lines, enforcing smooth geometric transitions that match real-world infrastructure characteristics.
3Reliability
If conventional map generation models are used, then the system remains simple, but the reliability of autonomous navigation deteriorates
Solution Approach 1:
The patent enhances navigation reliability by modifying training parameters through a specialized loss function. This loss function incorporates geometric validation parameters that ensure generated maps meet minimum quality thresholds for safe navigation, such as convex polygon representations and smooth polyline paths, thereby improving safety without requiring complete system redesign.
Data Source
AI summary
Certain aspects of the present disclosure provide a method for generating a map. A method generally includes obtaining sensor data corresponding to one or more sensors. The method includes inputting the sensor data to a map generation model. The method obtains as output from the map generation model a vectorized map of one or more features of an environment. The vectorized map represents the first feature as a convex polygon, such that all internal angles of the convex polygon are less than or equal to 180 degrees.