Driving Data Analyzer Latent Feature Extraction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing driving assist systems face challenges in extracting common driving scenes that are independent of driver habits and vehicle types, leading to inconsistent scene extraction for similar driving situations.

Innovation Solution

A driving data analyzer that collects and correlates driving data sequences with identification data, using a data compression network model to extract latent features independent of external factors such as driver habits and vehicle types, enabling robust data analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If driving scenes are extracted based on driving data groups that include driver habits and vehicle types, then the driving scenes reflect individual driving characteristics, but the extracted driving scenes become inconsistent for similar driving situations across different drivers and vehicle types

Engineering Contradiction:

Improvedriving scene extraction accuracyVSAvoiddriving scene consistency

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent extracts and removes the influence of external factors (driver habits, vehicle types) from the driving data before analyzing driving scenes. By separating these variable factors from the core driving behavior patterns, the system achieves consistent driving scene extraction across different drivers and vehicle types while maintaining accuracy in identifying actual driving situations.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms the driving data by changing the parameters used for analysis - specifically, it normalizes driving data to remove variations caused by external factors. This parameter transformation allows the system to focus on essential driving behavior patterns rather than being influenced by driver-specific or vehicle-specific characteristics.

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If driving data is collected from multiple drivers and vehicle types, then the driving data group includes diverse driving characteristics, but it becomes difficult to extract common driving scenes independent of external factors

Engineering Contradiction:

Improvedriving data diversityVSAvoiddata analysis complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary processing step that acts as a mediator between diverse driving data and driving scene extraction. This intermediary layer normalizes and standardizes the data from different drivers and vehicle types, transforming heterogeneous data into a unified format that enables common driving scene extraction without requiring complex individualized analysis for each data source.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11061911B2Driving data analyzer

Publication Date: 2021.07.13 DENSO CORP

AI summary

In a driving data analyzer, a data collector collects, from at least one vehicle, driving data sequences while each of the driving data sequences is correlated with identification data. Each driving data sequence includes sequential driving data items, and each driving data item represents at least one of a driver's operation of at least one vehicle and a behavior of the at least one vehicle based on the at least one of a driver's operation. The identification data represents a type of at least one external factor that contributes to variations in the driving data items. A feature extractor applies a data compression network model to the driving data sequences to thereby extract, from the driving data sequences, at least one latent feature independently from the type of the at least one external factor.