Lane Description Using Sensor Fusion and ML Assignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital maps used in autonomous vehicles lack timeliness and accuracy, making them insufficient for real-time navigation and highly automated driving.
Innovation Solution
A method using a machine learning-based assignment function to determine the description of a lane by assigning feature vectors from environmental data, including camera, radar, and LIDAR data, to provide precise and current information about the road or lane, enabling vehicles to autonomously determine their surroundings and navigate accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If digital maps are updated by a provider at specific intervals, then the map creation process is simple and centralized, but the maps are not sufficiently up-to-date and accurate for autonomous vehicles
Solution Approach 1:
The autonomous vehicle performs self-updating of digital map information by detecting lane markings and determining lane descriptions using its own sensors and machine learning models, eliminating dependence on periodic provider updates
Solution Approach 2:
The vehicle continuously captures environmental data, processes it through the assignment function, and updates its lane description in real-time, creating a feedback loop that maintains current and accurate map information
2Measurement precision
If traditional image processing methods are used for lane detection, then the system is simpler, but the precision and reliability of lane description are insufficient for highly automated driving
Solution Approach 1:
Traditional image processing methods are replaced with a machine learning-based assignment function that processes feature vectors to determine lane descriptions, achieving higher precision through learned patterns rather than rule-based processing
Solution Approach 2:
The system transforms raw environmental data into feature vectors with specific dimensions and properties, then uses the assignment function to map these features to lane descriptions, changing the parameter representation to achieve better detection accuracy
Data Source
Figure 1~3
Figure 4
AI summary
A method (400) for ascertaining a description of a lane (112) situated in the surroundings of a vehicle (102) is described. The method (400) comprises ascertainment (401) of an assignment function that is set up to assign different descriptions of a lane (112) to different values of a feature vector. Furthermore, the method (400) comprises ascertainment (402) of surroundings data for the vehicle (112), wherein the surroundings data indicate information about a carriageway marking and/or about one or more objects (101, 103) in the surroundings of the vehicle (102). Moreover, the method (400) comprises ascertainment (403) of a current value of the feature vector on the basis of the surroundings data. The method (400) further comprises determination (404) of a description of the lane (112) on the basis of the assignment function and on the basis of the current value of the feature vector.