Driver State Assessment Using Residual Error

The patent applies preliminary action by using a driver model to predict the driver's intended steering operations before the driver actually executes them. The system identifies a driver model from historical data, then uses this model to generate predicted steering angles in advance. By comparing these predicted angles with actual steering inputs, the system can detect deviations that indicate drowsiness before they manifest in vehicle behavior changes, thus eliminating the time delay while maintaining assessment precision.

The patent applies parameter changes by transforming the complex Bode plot comparison into a simpler residual error metric. Instead of comparing entire frequency response curves and interpreting ambiguous shape characteristics (peak vs flat), the system calculates a single residual error value representing the difference between predicted and actual steering angles. This quantitative parameter provides clear, unambiguous assessment criteria while maintaining the ability to detect driver state changes with high precision.

The patent applies self-service by implementing a recursive identification process where the driver model automatically updates itself using ongoing driver behavior data. The system continuously refines the driver model parameters without requiring external recalibration or complex manual adjustments. This self-updating mechanism maintains high driver model accuracy and adaptability while minimizing the operational complexity for users, as the system performs the complex identification tasks autonomously.