Driving Scenario Classification With Crowdsourced Feedback Rules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for classifying driving scenario data in autonomous driving simulation systems are not precise enough, especially when dealing with data from multiple dimensions, and trained models often show poor generalization performance due to incomplete training samples.
Innovation Solution
A method involving crowdsourcing for annotating scenario data, selecting representative samples for training, and using machine learning to generate and update a classification model based on user-provided rules to improve annotation and training quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional classification methods are used for driving scenario data, then the classification process is simple, but the classification accuracy and generalization performance are poor
Solution Approach 1:
The patent applies preliminary action by using crowdsourcing to pre-annotate driving scenario data with standard labels before training the classification model. This pre-processing step ensures that the training data is already classified and labeled, improving the model's learning efficiency and final classification accuracy without adding complexity to the model architecture itself.
Solution Approach 2:
The patent implements feedback mechanisms where user corrections and rule-based annotations are continuously fed back into the training dataset. The classification model is iteratively retrained with this refined data, progressively improving its accuracy. The system also uses prediction results to generate rules that further refine the training data, creating a closed-loop feedback system that enhances performance over time.
2Reliability
If comprehensive driving scenario data is collected to improve model generalization, then the model performance improves, but the data processing time and computational resources increase
Solution Approach 1:
The patent extracts only the most representative and informative driving scenario samples from the comprehensive dataset for training the classification model. By selecting key samples that capture the essential characteristics of different driving scenarios, the system achieves good generalization performance while significantly reducing the amount of data that needs to be processed, thereby decreasing training time and computational resource requirements.
Solution Approach 2:
The patent changes the parameter of data representation by transforming raw driving scenario data into structured formats with standardized labels through crowdsourcing annotation. This parameter transformation organizes the data more efficiently, enabling faster processing and training while maintaining or improving model generalization performance through better data quality and structure.
3Manufacturing precision
If manual annotation of driving scenario data is performed to improve data quality, then the training data accuracy improves, but the annotation cost and time consumption increase
Solution Approach 1:
The patent creates a multi-functional system where the classification model serves multiple purposes: it automatically annotates new driving scenario data, provides suggestions for manual annotation to improve consistency, and generates rules for data processing. This universal system reduces reliance on pure manual annotation while maintaining high data quality, thereby improving annotation efficiency without sacrificing precision.
Solution Approach 2:
The system implements self-service by enabling the classification model to automatically annotate and pre-process driving scenario data before it enters the manual review stage. The model also generates prediction results that can be directly used or refined, reducing the need for extensive manual annotation. This self-annotation capability significantly improves productivity while maintaining acceptable data quality through automated preprocessing.
Data Source
AI summary
Embodiments of a method, apparatus, device and computer readable storage medium for classifying driving scenario data includes: acquiring a first driving scenario data set from a crowdsourcing platform, driving scenario data in the first driving scenario data set having been classified; generating a driving scenario classification model at least based on the first driving scenario data set, for classifying driving scenario data collected by a collection entity; acquiring a rule for classifying the driving scenario data, the rule is generated based on a result of classifying the driving scenario data collected by the collection entity using the driving scenario classification model; updating the driving scenario classification model at least based on the rule.