Lane Line and Road Edge Mapping From Vehicle Telemetry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating lane line and road edge data for autonomous vehicles are time-consuming and costly, relying on pre-defined maps from aerial imagery.
Innovation Solution
A system and method using machine learning models, specifically a convolutional neural network and expectation maximization model, to process telemetry data from autonomous vehicles to generate lane line and road edge data, reducing reliance on aerial imagery and improving computational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pre-defined maps from aerial imagery are used to generate lane line and road edge data, then map accuracy is improved, but time consumption and cost increase
Solution Approach 1:
The patent uses telemetry data from autonomous vehicles as a copy or alternative source to create map data, replacing the need for expensive aerial imagery. By collecting and processing driving trajectory data from multiple vehicles, the system generates lane line and road edge information that replicates the accuracy previously only achievable through aerial photography, thereby reducing both time and cost while maintaining measurement precision
Solution Approach 2:
The patent replaces the mechanical/aerial surveying system with a data processing system using machine learning models. Instead of physically capturing aerial imagery and manually or algorithmically processing it to extract road features, the system substitutes this with neural network models that process telemetry data to directly generate lane line and road edge data, significantly reducing time consumption while maintaining accuracy
2Measurement precision
If pre-defined maps from aerial imagery are used to generate lane line and road edge data, then map accuracy is improved, but cost increases
Solution Approach 1:
The patent uses telemetry data from autonomous vehicles as a copy or alternative source to create map data, replacing the need for expensive aerial imagery. By collecting and processing driving trajectory data from multiple vehicles, the system generates lane line and road edge information that replicates the accuracy previously only achievable through aerial photography, thereby reducing both time and cost while maintaining measurement precision
Solution Approach 2:
The patent utilizes readily available telemetry data from autonomous vehicles, which is essentially a byproduct of normal vehicle operation. This data is processed through machine learning models to create permanent map data structures. The approach transforms inexpensive, easily obtainable telemetry information into valuable, persistent geographic information, replacing the need for expensive aerial surveying operations
3Productivity
If machine learning models are used to process telemetry data and generate map data, then productivity is improved, but device complexity increases
Solution Approach 1:
The patent divides the complex task of map generation into distinct processing stages handled by specialized machine learning models. The first neural network model processes telemetry data to identify and classify road features, while the second neural network model generates the final lane line and road edge data structures. This segmentation allows each model to be optimized for its specific function, improving overall computational efficiency while making the complex system more manageable and maintainable
Data Source
AI summary
Systems and method are provided for defining map data used in controlling a vehicle. In one embodiment, a method includes: receiving, by a processor, telemetry data; determining, by the processor, distribution data of a path based on the telemetry data; determining, by the processor, a plurality of sample data based on a trained machine learning model and the distribution data; generating, by the processor, at least one of lane line data and road edge data based on the sample data and a second machine learning model; and storing, by the processor, the map data including the lane line data and road edge data for use in controlling the vehicle.