Vehicle Driving Character Inference for Individual Stochastic Behavior

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing driving characteristic modeling for vehicles primarily focuses on deterministic characteristics at the cluster level, failing to account for individual stochastic variations in driving behaviors, which hinders accurate prediction of driving characteristics.

Innovation Solution

A driving characteristic inferring apparatus and method that trains a model using learning driving data to infer both deterministic and stochastic driving characteristics of individual vehicles, utilizing a monitoring unit, model training unit, and inference unit, with data collection from sensing modules and RSUs, and employs deep reinforcement learning and recurrent neural networks to minimize distributional distances for accurate inference.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If deterministic characteristic modeling at cluster level is used, then modeling complexity is reduced, but measurement precision of individual driving characteristics deteriorates

Engineering Contradiction:

Improvemodeling complexityVSAvoiddriving characteristic prediction accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent segments the driving characteristic modeling into two distinct components: deterministic characteristics (modeled at cluster level using reinforcement learning) and stochastic characteristics (modeled at individual vehicle level using neural networks). This segmentation allows each component to be optimized independently, maintaining low overall complexity while achieving high precision for individual vehicle prediction.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces stochastic parameters (mean and standard deviation of driving characteristics) alongside deterministic parameters. By changing from purely deterministic parameter modeling to stochastic parameter modeling, the system can capture individual variations while maintaining the structured approach of cluster-based reinforcement learning.

Inventive Principle:

Principle #35Parameter changes

2Quantity of substance

If cluster-level deterministic modeling is used, then data collection requirements are reduced, but reliability of individual vehicle characteristic inference deteriorates

Engineering Contradiction:

Improvedata collection requirementsVSAvoidindividual vehicle characteristic inference accuracy

Core Design Contradiction:

Quantity of substanceVSReliability

Solution Approach 1:

The patent segments the modeling approach into cluster-level deterministic modeling (requiring less data) and individual-level stochastic modeling (requiring more data but providing higher reliability). This segmentation allows the system to leverage available data efficiently while achieving reliable individual vehicle inference.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent uses cluster-level deterministic characteristics as an intermediary that bridges the gap between limited individual vehicle data and accurate individual characteristic inference. The cluster-level model provides a baseline that is then refined using individual vehicle data through the stochastic model.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Measurement precision

If individual stochastic characteristic modeling is implemented, then driving characteristic prediction accuracy is improved, but device complexity increases

Engineering Contradiction:

Improvedriving characteristic prediction accuracyVSAvoidmodeling system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent merges cluster-level reinforcement learning models with individual vehicle neural network models into a unified hybrid system. This merging allows the system to leverage the strengths of both approaches: the structured decision-making of reinforcement learning and the pattern recognition capabilities of neural networks, achieving high accuracy without excessive complexity.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent changes the parameter representation from purely deterministic to stochastic parameters (mean and standard deviation). This parameter change enables the model to capture individual variations efficiently, improving prediction accuracy while maintaining manageable system complexity through parameterized stochastic models.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12552393B2Method and apparatus for inferring driving characteristic of a vehicle in real-time

Publication Date: 2026.02.17 FOUND OF SOONGSIL UNIV IND COOP

AI summary

Provided is a method and an apparatus of inferring a stochastic driving characteristic of a driving vehicle. The driving characteristic inferring apparatus may include a model training unit which trains a plural driving characteristic model and an inference model using learning driving data of a learning driving vehicle, and a driving characteristic inferring unit which infers a driving characteristic coefficient representing a driving characteristic of a target driving vehicle with driving data of the target driving vehicle as an input of the inference model.