2D Map Localization Using Vertical Structure Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in fast and accurate localization due to the inefficiencies of 3D LiDAR data storage and alignment, particularly with traditional 3D alignment algorithms like Iterative Closest Point (ICP) having poor convergence properties and being susceptible to occlusion and non-uniformity, while 2D LiDAR sensors are prone to errors from transient obstacles and off-plane hazards.
Innovation Solution
A method is developed to generate a 2D representation of structures from 3D point data by partitioning the scene into cells, projecting points, recording heights, determining properties through bit-wise operations, and matching these representations to candidate maps using particle filters or factor graphs for localization, thereby leveraging the robustness of 3D LiDAR data efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If 3D LiDAR data is used for localization, then robustness against occlusion and measurement accuracy are improved, but storage requirements and computational complexity increase
Solution Approach 1:
The patent extracts only the essential vertical structure information from complete 3D point clouds, creating a simplified representation that retains robustness against occlusion while reducing storage requirements. The system extracts height data and vertical structures rather than storing all 3D points.
Solution Approach 2:
Instead of downsampling 3D data to reduce storage, the patent inverts the approach by selectively extracting only the critical vertical structure information needed for localization, discarding redundant data while preserving essential features for robust matching.
2Measurement precision
If traditional 3D alignment algorithms like ICP are used, then localization accuracy can be achieved, but convergence properties are poor and computational efficiency decreases
Solution Approach 1:
The patent performs preliminary processing to create a simplified 2D representation with vertical structure information before alignment. This preprocessing step prepares the data in advance, enabling faster and more reliable convergence during the alignment process while maintaining accuracy.
Solution Approach 2:
The patent transforms the alignment problem by changing the data representation from full 3D point clouds to 2D maps with height information. This parameter transformation simplifies the alignment computation while preserving the vertical structure necessary for accurate localization.
3Productivity
If 2D LiDAR sensors are used for localization, then storage efficiency and computational speed are improved, but susceptibility to occlusion and off-plane hazards increases
Solution Approach 1:
The patent enhances 2D LiDAR data by incorporating the vertical dimension through height information extracted from 3D data. This creates a hybrid representation that maintains the computational efficiency of 2D maps while adding vertical structure information to improve robustness against occlusion and off-plane hazards.
Data Source
AI summary
Autonomous vehicles require precise localization to support safe and reliable operation. Current systems aim to localize in 6DOF based on observations from a combination of cameras and 3D LiDAR, matching against dense, 3D prior maps. These maps are quite large and complex, presenting both computational and physical challenges in terms of matching, storage, and retrieval. Most of the environments where vehicles operate in contain frequent and distinct vertical structure sufficient for 2D localization, while state-of-the-art IMUs can be used to recover roll and pitch. This disclosure introduces a fast method for constructing 2D maps summarizing the vertical structure in the environment and demonstrate that it can be used to localize accurately in vehicular and other applications.