Autonomous Vehicle Localization Using Sector-Based Map Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face inaccuracies in location due to discrepancies between sensed data and map data caused by road slopes or 3-dimensional objects, leading to potential traffic accidents.
Innovation Solution
A method involving a computing device that determines the validity of sensed data by using object-based and vehicle-based sectors, adjusting weights based on data errors and previous calibration values, to accurately locate the vehicle.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If map data is used for locating the autonomous vehicle, then the vehicle can be positioned on the top view map, but the location accuracy deteriorates when there is a slope or 3-dimensional object causing discrepancy between sensed data and map data
Solution Approach 1:
The patent divides the surrounding environment into multiple sectors (e.g., 360 degrees divided into 10-degree intervals creating 36 sectors) based on the autonomous vehicle's heading direction. Each sector independently processes object detection and matching with map data, allowing localized corrections without affecting the entire localization system. This segmentation enables the system to handle discrepancies in specific areas while maintaining overall localization accuracy.
Solution Approach 2:
The patent dynamically adjusts weighting parameters when fusing sensed object data with map object data. When discrepancies are detected between sensed data and map data (indicating slopes or 3D objects), the system changes the weight given to each data source, relying more on map data in problematic sectors and more on sensed data in reliable sectors. This parameter adjustment resolves the contradiction by adaptively balancing data consistency and measurement precision.
2Ease of operation
If the autonomous vehicle uses standard mapping methods, then the process is simple, but the driving path may lead the vehicle through walls due to inaccurate location mapping
Solution Approach 1:
The patent transitions from 2D top-view map representation to 3D spatial understanding by incorporating elevation information and object height data. The system detects 3-dimensional objects and adjusts the mapping to account for vertical discrepancies, allowing the vehicle to navigate around obstacles that would otherwise appear passable in a flat 2D map. This dimensional enhancement maintains operational simplicity while dramatically improving driving path reliability.
3Measurement precision
If the system uses multiple data sources for location determination, then accuracy can be improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent implements selective processing where not all sectors require full data fusion and validation. When map data and sensed data match within acceptable thresholds, the system accepts the location with minimal processing. Only in sectors where discrepancies exceed thresholds does the system perform extensive multi-source data fusion and validation. This partial action approach maintains high location accuracy while significantly reducing average computational complexity.
Data Source
AI summary
A method for locating a specific autonomous vehicle is provided. The method includes steps of: (a) receiving, by a computing device of the specific autonomous vehicle, map data related to a position of the specific autonomous vehicle and its adjacent regions, from a server; (b) obtaining, by the computing device, a specific object-based sector and a specific vehicle-based sector; and (c) determining, by the computing device, a validity of the specific sensed data by referring to the specific object-based sector, the specific vehicle-based sector and the map object data, and in case the specific sensed data is determined as valid, locating the specific autonomous vehicle by referring to a specific matched result data generated from a matching process between the map object data and the specific sensed data.