Vehicle Abnormality Detection Using Dynamic Threshold Estimation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing abnormality detection systems for road conditions face challenges in accurately determining the presence of abnormalities due to the arbitrary setting of thresholds, leading to erroneous determinations, and lack a clear relationship between threshold settings and detection accuracy.

Innovation Solution

An abnormality detection system that calculates an abnormality degree by comparing driving information from evaluation data sets to reference data sets using two-sample U-statistics, and estimates an optimal threshold based on attribute information to minimize false-positive rates, ensuring predetermined detection accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If a threshold is set arbitrarily for abnormality detection, then the detection process is simple, but the detection accuracy deteriorates due to erroneous determinations

Engineering Contradiction:

Improvethreshold setting simplicityVSAvoidabnormality detection accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent changes the parameter of threshold from a fixed arbitrary value to a dynamically determined value based on statistical analysis of historical data. The threshold is calculated using the standard deviation of past abnormality degrees, transforming it from a static parameter to one that adapts to actual data characteristics, thereby improving detection accuracy while maintaining operational simplicity.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The system performs self-calibration by automatically determining the threshold through statistical analysis of its own historical data without requiring manual intervention. The abnormality detection apparatus uses its accumulated data to self-adjust the threshold parameter, eliminating the need for external expert knowledge or trial-and-error tuning.

Inventive Principle:

Principle #25Self-service

2Productivity

If a fixed threshold is used for abnormality determination, then the detection process is fast and simple, but the false-positive rate increases due to lack of adaptation to actual conditions

Engineering Contradiction:

Improvedetection processing speedVSAvoiddetection accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system performs preliminary statistical analysis on historical abnormality degree data to pre-determine an optimal threshold before actual abnormality detection begins. By calculating the threshold in advance based on past data characteristics (mean and standard deviation), the system prepares the detection criteria beforehand, enabling both fast real-time detection and high reliability through data-driven threshold selection.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system incorporates feedback from historical detection results by continuously analyzing past abnormality degrees to refine the threshold. The threshold is adjusted based on feedback from actual operational data, creating a closed-loop system where detection accuracy improves over time while maintaining fast processing through the established statistical model.

Inventive Principle:

Principle #23Feedback

3Ease of manufacture

If an arbitrary threshold is set without statistical basis, then the system is easy to implement, but the relationship between threshold and detection accuracy remains unclear

Engineering Contradiction:

Improvesystem implementation easeVSAvoidthreshold-accuray relationship clarity

Core Design Contradiction:

Ease of manufactureVSLoss of information

Solution Approach 1:

The patent replaces the mechanical/manual threshold setting process with a statistical calculation system. Instead of relying on human expertise or arbitrary values, the threshold is automatically computed using statistical formulas (involving mean, standard deviation, and confidence levels) applied to historical data, establishing a clear mathematical relationship between threshold and detection accuracy.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces statistical parameters (mean, standard deviation, confidence level) as intermediaries between the raw historical data and the final threshold value. These statistical measures serve as mediators that translate complex historical patterns into a clear, interpretable threshold that has a known relationship with detection accuracy through established statistical theory.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11587372B2Abnormality detection method, abnormality detection apparatus, and abnormality detection system

Publication Date: 2023.02.21 DENSO CORP

AI summary

In an abnormality detection method for detecting an abnormality in vehicle behaviors based on an abnormality degree indicating a degree of difference between first driving information representing a vehicle behavior undergoing abnormality detection and second driving information obtained when no abnormality is found in a travel environment, an abnormality threshold is estimated using an abnormality degree distribution generated from a set of preaccumulated abnormality degrees. The abnormality threshold is an abnormality degree allowing a probability of erroneous determination as to presence or absence of an abnormality to be a predefined probability. Further, in the abnormality detection method, it is detected whether an abnormality is found in the vehicle behavior by comparing the abnormality degree for the first driving information and the estimated abnormality threshold.