Vehicle Alert Display Control Using Prediction Error Cues
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing display control technologies for vehicles often alert drivers unnecessarily, causing annoyance and overlooking critical objects due to lack of differentiation between perceived and non-perceived scenes, leading to inefficient alert systems.
Innovation Solution
A display control system that uses a deep predictive coding network to predict scenes and identify objects with high prediction errors, selectively displaying alerts for objects likely to be overlooked, reducing unnecessary alerts and improving driver attention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If alerts are displayed for all detected objects, then the driver is informed of all potential risks, but the driver feels annoyed due to unnecessary alerts
Solution Approach 1:
The patent applies local quality by differentiating alert presentation based on prediction error characteristics. Objects with high prediction errors (unexpected objects) are highlighted with special alert icons, while objects with low prediction errors (expected objects) are displayed without special alerts. This selective alert strategy reduces driver annoyance while maintaining reliability by focusing attention on critical unexpected objects.
2Loss of information
If alerts are displayed for all detected objects, then comprehensive information is provided, but critical objects may be overlooked due to information overload
Solution Approach 1:
The patent applies local quality by differentiating alert presentation based on prediction error characteristics. Objects with high prediction errors (unexpected objects) are highlighted with special alert icons, while objects with low prediction errors (expected objects) are displayed without special alerts. This selective alert strategy reduces driver annoyance while maintaining reliability by focusing attention on critical unexpected objects.
Solution Approach 2:
The patent extracts and separates critical information from general information by using different alert icons. Expected objects (low prediction error) are displayed with standard icons, while unexpected objects (high prediction error) are displayed with special alert icons. This extraction of critical unexpected objects from the general set of detected objects prevents information overload and ensures critical objects are not overlooked.
3Measurement precision
If the system uses deep learning prediction models, then accurate scene prediction is achieved, but the device complexity increases
Solution Approach 1:
The patent uses prediction error as an intermediary metric to bridge the gap between complex deep learning predictions and simple alert decisions. The system calculates prediction errors between consecutive frames and uses this intermediate value to determine which objects require special alerts. This intermediary approach allows the system to leverage accurate deep learning predictions while maintaining simple and interpretable alert logic.
Data Source
AI summary
An input unit receives a first image that captures a scene in front of a vehicle or a scene in front of a user on board a vehicle at a first point of time. A generator generates a third image showing a result of prediction of the scene in front at the first point of time, based on a second image that captures the scene in front at a second point of time preceding the first point of time. A display controller controls display of information presentation based on a result of comparison between the first image received by the input unit and the third image generated by the generator using a predetermined criterion.