Map Representation Updating Through LIDAR Localization Convergence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing navigation and self-driving systems face challenges in accurately localizing vehicles due to outdated map representations, leading to poor performance of localization techniques when the map data is not up to date.
Innovation Solution
A method and server system that utilizes a localization algorithm to generate candidate locations using point clouds and evaluate their convergence, identifying outdated map portions and updating them using LIDAR data, employing machine learning algorithms to enhance localization accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a map representation is used for localization, then the localization process can be performed using pre-built map data, but the localization accuracy deteriorates when the map data is outdated
Solution Approach 1:
The system performs preliminary actions by continuously collecting LIDAR point cloud data during vehicle operation and pre-processing this data for future map updates. The server receives and stores point cloud data from multiple vehicles in advance, preparing updated map representations before they are needed for localization, thus preventing accuracy deterioration.
Solution Approach 2:
The system implements feedback by using the localization algorithm to generate candidate locations and evaluating their convergence. When candidate locations diverge (indicating outdated map data), the system triggers a map update process. The updated map is then fed back into the localization system, creating a closed-loop feedback mechanism that maintains localization accuracy.
2Measurement precision
If map updates are performed frequently to maintain accuracy, then localization accuracy is improved, but the system complexity and computational resources increase
Solution Approach 1:
Instead of updating the entire map uniformly, the system performs localized updates only in specific geographic portions where map data has become outdated. The server identifies candidate portions based on vehicle trajectories and updates only those specific areas, reducing computational complexity while maintaining localization accuracy where it matters.
Solution Approach 2:
The system performs partial updates by processing only the necessary portions of the map rather than complete re-updates. By using convergence evaluation to identify only the outdated portions, the system applies updates selectively, reducing the overall computational burden while sufficient for maintaining accuracy.
3Manufacturing precision
If LIDAR point cloud data is collected and processed continuously, then map update accuracy is improved, but the data processing time and computational energy increase
Solution Approach 1:
The system performs preliminary data collection by continuously gathering LIDAR point cloud data during normal vehicle operation and storing it for later processing. This preliminary action ensures that when map updates are needed, the data is already collected and ready, reducing the actual processing time while maintaining high update accuracy.
Solution Approach 2:
The system implements self-service by automatically evaluating map data freshness using convergence metrics and triggering updates only when necessary. The server autonomously determines when map portions are outdated and initiates updates without external intervention, optimizing the balance between update accuracy and processing time.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enhances localization accuracy by identifying and updating outdated map portions, ensuring reliable navigation and operation of self-driving cars by reducing divergence in localization results.
Implementation Method 1
LIDAR-based object detection generally comprises transmitting beams of light towards a region of interest, and detecting reflected light beams, such as from objects in the region of interest, to generate a representation of the region of interest including any objects
Implementation Method 2
determining, by the server using a convergence metric, a parameter for evaluating quality of localization of the localization algorithm when the SDC is located in the candidate portion of the map representation. The parameter is indicative of convergence of candidate locations in the set of candidate locations
Data Source
AI summary
Methods and server for updating a candidate portion of a map representation of a geographical region is disclosed. The server may generate, using a localization algorithm, a set of candidate locations based on a set of point clouds and the map representation. The set of point clouds is being associated with the candidate portion of the map representation. A server may determine, using a convergence metric, a parameter for evaluating quality of localization of the localization algorithm when the SDC is located in the candidate portion of the map representation. The parameter is being indicative of convergence of candidate locations in the set of candidate locations. A server may identify, using the parameter, the candidate portion of the map representation as an outdated portion of the map representation. A server may update the outdated portion of the map representation using the set of point clouds.