Lane Detection via Road Model and Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane detection systems for automated vehicles face challenges in poor visibility, missing or poorly visible road markings, and inaccurate GNSS-based position determination, leading to erroneous lane detection and potentially dangerous driving situations, especially in areas like roadworks, entrances, exits, widenings, and narrowings.
Innovation Solution
A method and system that utilize navigation data, such as map data and road sign detection, to determine a robust road model by weighting detected features, prioritizing reliable road markings and boundaries, and adjusting the route guidance curve to prevent missteering and improve lane detection accuracy in challenging conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If optical sensors are used to detect lanes based on road markings, then lane detection can be performed in normal conditions, but detection accuracy deteriorates in poor visibility, missing road markings, or complex road situations
Solution Approach 1:
The patent combines multiple data sources including optical sensor data, GNSS position information, and pre-stored road model data to create a comprehensive lane detection system. This merging of information sources compensates for the weaknesses of individual sensors in challenging conditions, maintaining both accuracy and reliability when road markings are missing or visibility is poor
Solution Approach 2:
The patent introduces a road model as an intermediary layer that mediates between raw sensor data and lane detection results. This road model, pre-stored with road geometry information, acts as a reference framework that helps interpret optical sensor data and GNSS position information, improving detection reliability in complex situations by providing contextual guidance
2Loss of information
If GNSS-based position determination is used to determine vehicle location, then position information can be obtained, but accuracy is insufficient for precise lane detection
Solution Approach 1:
The patent changes the parameter of position determination by combining GNSS absolute position data with relative positioning based on detected road markings and pre-stored road models. This multi-parameter approach transforms the single-source GNSS positioning into a hybrid positioning system that achieves higher precision while maintaining information availability
3Loss of information
If additional lane markings are detected at road features like entrances and exits, then more road information is available, but erroneous lane detection occurs due to misidentification of the actual lane
Solution Approach 1:
The patent implements a feedback mechanism where the road model continuously validates detected lane markings against pre-stored road geometry information. When additional markings are detected at road features, the system uses feedback from the road model to verify whether these markings correspond to the actual vehicle lane, preventing erroneous identification while maintaining complete information utilization
Data Source
AI summary
A method detects a lane for a transverse guidance of a vehicle. The transverse guidance of the vehicle is based on a roadway model. The method has the steps of ascertaining one or more features which are suitable for influencing the detection of the lane; detecting a lane on the basis of a sensor system of the vehicle; and ascertaining the roadway model on the basis of the detected lane and the ascertained one or more features. The method optionally has the steps of additionally receiving navigation data and transversely guiding the vehicle on the basis of the ascertained roadway model.