Autonomous Vehicle Road Boundary Filtering for Lateral Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face inaccuracies in lateral position estimation due to lidar perception limitations, particularly when road structures deviate from rectangular shapes, leading to incorrect matching of road boundaries and reduced precision in autonomous driving.
Innovation Solution
An autonomous vehicle system utilizing a global positioning system (GPS), high-definition (HD) maps, and sensors to determine road boundaries based on a line of sight (LOS) condition, filtering targets through property, area, and direction checks, and maintaining a filtering target across frames to enhance precision positioning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If lidar perception is used to detect road boundaries, then autonomous driving capability is enabled, but lateral position estimation accuracy deteriorates when road structures are non-rectangular
Solution Approach 1:
The patent introduces an intermediary filtering process that uses LOS conditions as a mediator between raw lidar detection and final road boundary determination. This filtering mechanism selectively processes detected boundaries based on geometric validity, preventing incorrect matches while preserving autonomous driving functionality.
Solution Approach 2:
The patent implements feedback through the LOS condition verification process, where detected road boundaries are continuously validated against geometric constraints. This feedback loop corrects positioning errors by rejecting invalid detections and reinforcing accurate boundary identifications, thereby improving lateral position estimation accuracy.
2Ease of operation
If iterative closest point (ICP) algorithm is used to match lidar output lines with road boundaries, then road boundary detection is achieved, but positioning accuracy deteriorates due to incorrect matching
Solution Approach 1:
The patent applies preliminary action by establishing LOS conditions before performing ICP matching. These pre-established geometric constraints serve as validation criteria that must be satisfied before a road boundary match is accepted, preventing incorrect matches from degrading positioning accuracy.
Solution Approach 2:
The LOS condition acts as an intermediary validation layer between the ICP algorithm and final boundary determination. This intermediary process filters out incorrect matches that would otherwise be accepted by the ICP algorithm, thereby maintaining positioning accuracy while preserving detection capability.
3Area of stationary object
If all detected road boundaries are used for positioning, then comprehensive coverage is achieved, but positioning accuracy deteriorates due to inclusion of irrelevant boundaries
Solution Approach 1:
The patent extracts only the relevant subset of road boundaries that satisfy LOS conditions from the complete set of detected boundaries. This extraction process removes irrelevant boundaries (such as those obscured by median strips or non-visible structures) while retaining comprehensive coverage of actually visible boundaries, thereby improving positioning accuracy.
Solution Approach 2:
The patent applies local quality by treating different detected boundaries differently based on their LOS validity. Boundaries satisfying LOS conditions are processed with high priority for positioning, while those not satisfying conditions are filtered out or given lower priority, creating a quality-based differentiation that improves overall positioning accuracy.
Data Source
AI summary
An apparatus for controlling autonomous driving of a vehicle is introduced. The apparatus may comprise a GPS receiver, a memory storing a high-definition (HD) map, at least one sensor for sensing surroundings of the vehicle, and a processor. The processor is configured to estimate the vehicle's position based on GPS information, HD map information, and sensor information. The apparatus further determines one or more road boundaries based on a line of sight (LOS) condition, which is verified using the estimated position. A filtering target is selected from the road boundaries based on a reference boundary. A driving distance traveled by the vehicle is set periodically at predetermined intervals based on the filtering target. A signal associated with the periodically set driving distance is output, and autonomous driving of the vehicle is controlled based on the signal.