Adaptive Traffic Model Selection for Unconventional Patterns

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing traffic prediction models face challenges in accurately predicting traffic conditions due to unforeseen factors like weather, accidents, and seasonal variations, leading to discrepancies between predicted and actual traffic, especially on roads with unconventional patterns.

Innovation Solution

A system that selects the most confident model from a plurality of historical models based on real-time traffic data, using confidence metrics to determine the best model for predicting traffic information, and adjusts predictions by decaying real-time data with historical data, ensuring more accurate and nuanced traffic flow determination.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a single historical traffic prediction model is used, then the model structure is simple and easy to implement, but the prediction accuracy deteriorates when facing unconventional traffic patterns caused by weather, accidents, or seasonal variations

Engineering Contradiction:

Improveprediction accuracyVSAvoidmodel selection system complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system dynamically selects different historical models based on real-time traffic conditions rather than using a fixed single model. The model selection is adaptive and changes according to the current traffic situation, allowing the system to maintain high accuracy across varying traffic patterns while keeping individual model structures relatively simple

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the model parameter (which historical model to use) based on confidence metrics calculated from real-time data. By adjusting which model is active based on confidence levels, the system achieves high prediction accuracy for unconventional patterns without requiring each individual model to be overly complex

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If multiple historical models are maintained to handle different traffic patterns, then the prediction accuracy improves for various conditions, but the system complexity and computational overhead increase

Engineering Contradiction:

Improveability to handle unconventional traffic patternsVSAvoidnumber of models and selection mechanism

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system uses confidence metrics calculated from real-time traffic data as feedback to determine which historical model should be selected. This feedback mechanism allows the system to adapt to different traffic patterns automatically, maintaining high versatility while using a systematic approach to manage model selection complexity

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs self-service by automatically selecting the appropriate historical model based on confidence metrics without requiring external intervention. The model selection process is autonomous, using real-time data to determine which model best fits current conditions, thereby handling diverse patterns efficiently

Inventive Principle:

Principle #25Self-service

3Measurement precision

If real-time traffic data is heavily weighted in predictions, then the predictions reflect current conditions better, but the stability and reliability of predictions deteriorate due to data noise and variability

Engineering Contradiction:

Improvereal-time traffic reflection accuracyVSAvoidprediction stability

Core Design Contradiction:

Measurement precisionVSStability of the object's composition

Solution Approach 1:

The system uses historical models as intermediaries between raw real-time data and final predictions. Instead of directly using noisy real-time data, the system processes it through selected historical models that provide stabilization, thereby maintaining precision while improving stability

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system applies beforehand cushioning by using historical traffic patterns to buffer against the noise and variability in real-time data. The historical models provide a stable foundation that cushions the impact of real-time data fluctuations, maintaining prediction stability while still reflecting current conditions

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

Data Source

PatentUS10460599B2Method and apparatus for providing model selection for traffic prediction

Publication Date: 2019.10.29 HERE GLOBAL BV

AI summary

An approach is provided for determining one or more real-time traffic data series associated with one or more road segments over one or more time intervals. The approach involves causing, at least in part, a determination of one or more differences of the real-time data series with one or more historical models. The approach also involves determining whether the one or more real-time difference exceeds a threshold difference. The approach involves causing, at least in part, a selection of one or more other historical models for the one or more road segments over the one or more time intervals based, at least in part, on a determination that the real-time difference exceeds the threshold difference.