Driver Attention Estimation Using Road Scene Prediction Error
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing driving assistance systems struggle to effectively address cognitive factors that lead to traffic accidents or near-miss events, even when drivers are not fatigued and the traffic environment is easily recognizable.
Innovation Solution
A driving assistance apparatus that calculates a prediction error between predicted and actual images from a vehicle exterior camera, estimates the driver's attention state based on this error, and outputs driving assistance information to prompt appropriate attention and behavior changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If driving assistance systems monitor only attention function degradation and visual recognition, then they can detect fatigue and obvious visual obstacles, but they fail to detect cognitive factors leading to accidents even when the driver is alert and the environment is clearly visible
Solution Approach 1:
The system segments the driver's cognitive state into multiple measurable dimensions: attention function degradation, visual recognition level, and cognitive factor estimation. Each dimension is assessed separately using different input data (gaze information, surrounding environment, prediction error), allowing comprehensive monitoring without requiring a single complex detection mechanism
Solution Approach 2:
The system introduces an intermediary estimation mechanism that uses prediction error from surrounding environment data as an indirect indicator of cognitive state. Rather than directly measuring cognitive factors, the system calculates the difference between predicted and actual road situations, using this prediction error as a mediator to infer when the driver may be overlooking clearly visible hazards
2Reliability
If the system presents comprehensive driving assistance information continuously, then it can provide complete safety guidance, but it may overwhelm the driver with excessive information
Solution Approach 1:
The system applies local quality by tailoring the type, amount, and presentation of assistance information to the specific cognitive state detected in each situation. When cognitive factors are identified as the issue, the system provides targeted prompts about specific overlooked elements rather than generic safety messages, making the information both necessary and easily processable
Solution Approach 2:
The system implements continuous feedback loops where driver responses to assistance information are monitored and used to adjust subsequent information presentation. The prediction error calculation and cognitive state estimation are updated in real-time based on whether the driver corrects their behavior after receiving assistance, allowing the system to refine its information delivery to maintain safety without overwhelming the driver
Data Source
AI summary
A driving assistance apparatus according to the present disclosure includes a memory and a processor. The processor is coupled to the memory, and configured to: calculate a prediction error that is a difference between a predicted image and an actual image, the predicted image being predicted from an image in a traveling direction of a vehicle captured by a vehicle exterior camera that captures a periphery of the vehicle, an actual situation being captured in the actual image; estimate an attention state of a driver, based on the prediction error; and output driving assistance information for prompting an attention state and/or a behavior change and/or a consciousness change in relation to a driving manipulation, which are more appropriate for driving at that time, based on the attention state.