Driver Attention Warning Threshold Adaptation Using Neural Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle driver attention warning systems output warnings uniformly based on driving data, failing to account for individual driving styles and situations, leading to inappropriate or missed warnings.
Innovation Solution
A system that compares driver attention levels from a neural network output with those from a traditional driver attention warning system, adjusting the threshold level based on differences and training the neural network with driving data to improve accuracy and sensitivity, thereby tailoring warnings to individual drivers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a uniform threshold level is used for driver attention warnings, then the system is simple to operate, but the accuracy of warnings decreases for individual drivers with different driving styles
Solution Approach 1:
The patent applies dynamics by making the threshold level adjustable and adaptable over time. The system starts with a default threshold and dynamically adjusts it based on comparisons between DAW system outputs and neural network predictions. This allows the system to transition from a static uniform threshold to a dynamic personalized threshold that adapts to individual driver characteristics while maintaining operational simplicity.
Solution Approach 2:
The patent implements feedback by continuously comparing the driver attention level from the DAW system with the prediction from the neural network. When discrepancies are detected beyond a predetermined threshold, the system adjusts the threshold level accordingly. This feedback mechanism enables the system to learn from actual driver behavior and improve warning accuracy while maintaining ease of operation through automated adjustment.
2Measurement precision
If the threshold level is adjusted frequently to improve accuracy, then the measurement precision improves, but the system stability deteriorates
Solution Approach 1:
The patent applies beforehand cushioning by setting predetermined conditions that must be met before threshold adjustment occurs. The system requires that the difference between DAW output and neural network prediction exceeds a predetermined threshold value, and this condition must be satisfied for a predetermined period or number of times before adjustment. This cushioning mechanism prevents premature or excessive adjustments, ensuring system stability while still improving accuracy when truly necessary.
3Measurement precision
If a neural network is trained with all available driving data, then the neural network accuracy improves, but the training time and computational resources increase
Solution Approach 1:
The patent applies partial action by training the neural network iteratively with subsets of driving data rather than requiring all data at once. The system uses available driving data to train the neural network progressively, updating the model with new data as it becomes available. This approach achieves good accuracy with reasonable training time by processing data in manageable portions rather than requiring complete datasets.
Data Source
AI summary
Provided is a vehicle, including: a user interface; a driver attention warning (DAW) system configured to determine a first driver attention level based on driving data, and when the first driver attention level is less than or equal to a threshold level, control the user interface to output a warning; and a controller configured to determine a second driver attention level based on an output of a neural network on driving data, compare the first driver attention level and the second driver attention level at predetermined periods, and when a number of times that a difference between the first driver attention level and the second driver attention level is greater than or equal to a predetermined value is greater than or equal to a predetermined number of times, adjust the threshold level in a decreasing direction. The neural network is configured to be trained with driving data.