Driver Fatigue Prediction Using Time-Segmented Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for predicting driver fatigue and inattentiveness require large storage capacities to manage extensive data over long driving periods, limiting their ability to include all relevant situations and data in fatigue forecasts.

Innovation Solution

A method that determines representative values of driving behavior and situational data across different time periods, using a vigilance index to calculate break recommendations, reducing memory requirements by focusing on immediate, short-term, medium-term, and long-term changes, allowing for a high-quality fatigue prognosis with minimal storage.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If all measurement data over the entire travel time are stored for fatigue prediction, then prediction quality is improved, but memory requirements increase significantly

Engineering Contradiction:
Improveprediction qualityVSAvoidmemory requirements
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent divides the entire travel time into multiple time ranges (e.g., recent, medium-term, long-term periods) and processes measurement data separately for each time range. This segmentation allows the system to capture fatigue-related variations across different time scales without storing all raw data points, thereby reducing memory requirements while maintaining prediction quality.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts representative values (such as mean values, minima, maxima, or standard deviations) from the measurement data in each time range. By storing only these extracted representative values rather than the complete raw measurement data set, the system achieves significant memory reduction while preserving the essential information needed for accurate fatigue prediction.

Inventive Principle:
Principle #2Taking out (Extraction)

2Quantity of substance

If representative values are determined in multiple time ranges, then memory requirements are reduced, but processing complexity increases

Engineering Contradiction:
Improvememory requirementsVSAvoidprocessing complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

By segmenting the time data into distinct ranges and assigning specific processing tasks to each range, the patent organizes the complexity in a structured manner. This makes the processing more manageable and systematic, even though multiple time ranges are involved, as each range can be handled independently with standardized operations.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the parameter representation by transforming raw measurement data into representative values (mean, min, max, standard deviation) for each time range. This parameter transformation simplifies the data structure and reduces the computational burden compared to processing complete raw data sets, thereby managing processing complexity effectively.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP2607140B1Method for generating and triggering a pause recommendation
Publication Date: 2020.02.12 VOLKSWAGEN AG
  • EP2607140B1 patent drawingFigure 1
  • EP2607140B1 patent drawingFigure 2
  • EP2607140B1 patent drawingFigure 3~4

AI summary

The method involves determining a driving characteristic of a driver corresponding to measurement data. Driving conditions of a vehicle are determined corresponding to the measurement data. The measurement data are observed in a time domain (14), and a preset representative value of the measurement data is determined in the time domain. A pause recommendation is calculated with a help of a preset vigilance index based on the representative value and output to the driver of the vehicle. Mean-, minimum- and maximum values and a standard deviation are determined as the representative value.