Point Cloud Cluster Weighting for HD Map Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately navigating due to sensor limitations and outdated conventional maps, which lack precision and freshness, making safe driving difficult.
Innovation Solution
Generating and maintaining high-definition (HD) maps using data from the vehicles' own sensors, allowing for precise and up-to-date navigation with efficient storage and low latency, and identifying dynamic objects to improve localization and safety.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional maps are used for navigation, then map data is available, but precision and accuracy are insufficient (3-5 meters GPS accuracy, over 100m error conditions)
Solution Approach 1:
The patent segments the point cloud data into multiple clusters representing different objects and features. Each cluster is assigned a weight based on its geometric features, allowing selective emphasis on high-precision features for alignment while filtering out noisy or irrelevant points. This segmentation enables the system to achieve sub-30cm navigation accuracy by focusing computational resources on the most reliable geometric correspondences.
2Measurement precision
If survey teams with high resolution sensors are used to create maps, then map precision improves, but cost and time consumption increase significantly (weeks to months to create comprehensive map)
Solution Approach 1:
The patent implements a self-service mapping approach where autonomous vehicles use their own onboard sensors to collect and contribute point cloud data. The system automatically processes this data through clustering and geometric feature analysis to update maps without requiring external survey teams. This enables continuous, cost-effective map creation and updates at scale, transforming the map creation process from a manual, expensive operation to an automated, vehicle-generated process.
3Duration of action of stationary object
If conventional map updating methods are used, then map data is maintained, but data freshness is insufficient (roads updated 5-10% per year faster than survey capacity)
Solution Approach 1:
The patent establishes continuous map updating by having autonomous vehicles constantly contribute sensor data during normal operation. Rather than periodic survey campaigns, the system processes point cloud data continuously as vehicles traverse roads, automatically detecting and recording changes. This continuous action ensures maps remain fresh and up-to-date with current road conditions, keeping pace with the 5-10% annual road update rate without requiring dedicated survey fleets.
4Area of stationary object
If all point cloud data is used for alignment, then comprehensive coverage is achieved, but computational efficiency decreases and storage requirements increase
Solution Approach 1:
The patent applies local quality by assigning different weights to different clusters of point cloud data based on their geometric features. High-weight clusters with distinctive geometric properties (such as corners, edges, or unique structures) are emphasized for alignment computations, while low-weight clusters with ambiguous or repetitive features are de-emphasized. This selective weighting reduces computational complexity and storage requirements by focusing processing on the most informative local regions rather than uniformly processing all points.
Data Source
AI summary
According to one or more embodiments, operations may comprise obtaining a first point cloud. The operations also comprise performing segmentation of the first point cloud, the segmentation generating one or more clusters of points of the point cloud. The operations also comprise determining, for each respective cluster of the plurality of clusters, a respective geometric feature of a corresponding object that corresponds to the respective cluster. The operations also comprise obtaining a second point cloud. The operations also comprise assigning a plurality of weights that comprises assigning a respective weight to each respective cluster based on the respective geometric feature that corresponds to the respective cluster. The operations also comprise obtaining a second point cloud and aligning the first point cloud with the second point cloud based on the plurality of weights.