Traffic Jam Prediction Model Using Facility Departure Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional traffic jam prediction systems are limited to predicting traffic conditions up to several hours into the future and cannot accurately forecast traffic jams over longer periods such as several days ahead, making it difficult to plan travel effectively.

Innovation Solution

A traffic jam predicting device that utilizes a learned model to predict traffic conditions over a long period by integrating vehicle information, weather data, and attendance records, specifically using a random forest algorithm to analyze the relationship between the number of people departing from facilities and traffic congestion.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional traffic jam prediction systems use real-time vehicle information to predict traffic conditions, then prediction accuracy for near-future traffic (several hours) is improved, but the ability to predict long-term traffic conditions (several days ahead) deteriorates

Engineering Contradiction:
Improvetraffic jam prediction accuracyVSAvoidprediction time horizon
Core Design Contradiction:
Measurement precisionVSDuration of action of moving object

Solution Approach 1:

The patent segments the prediction problem into two distinct models: a short-term prediction model using real-time vehicle information for near-future traffic (several hours), and a long-term prediction model using machine learning with multiple factors (weather, events, historical data) for extended periods (several days). This segmentation allows each model to optimize for its specific time horizon without compromise.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs preliminary data collection and processing by gathering weather information, event information, and historical traffic data in advance before the actual prediction is needed. This preliminary preparation enables the long-term prediction model to function effectively several days ahead, overcoming the limitation of real-time data-only approaches.

Inventive Principle:
Principle #10Preliminary action

2Duration of action of moving object

If a machine learning model uses multiple factors including weather and event information to predict long-term traffic conditions, then prediction capability for extended periods is improved, but system complexity increases

Engineering Contradiction:
Improveprediction time horizonVSAvoidsystem complexity
Core Design Contradiction:
Duration of action of moving objectVSDevice complexity

Solution Approach 1:

The patent implements a universal machine learning model that handles multiple prediction functions simultaneously: it processes weather information, event information, historical traffic data, and real-time vehicle information through a single integrated system. This multi-functional approach manages complexity by consolidating diverse data sources and prediction tasks into one cohesive model rather than requiring separate systems for each function.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system introduces an information processing unit that acts as an intermediary between diverse data sources (weather services, event databases, historical records, real-time vehicle sensors) and the prediction model. This intermediary layer standardizes and integrates multiple data types into a unified format, simplifying the overall system architecture while enabling comprehensive long-term prediction capabilities.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4210019B1Method for generating learned model, non-transitory storage medium, and traffic jam predicting device
Publication Date: 2025.12.10 TOYOTA JIDOSHA KK
  • EP4210019B1 patent drawingFigure 1
  • EP4210019B1 patent drawingFigure 2
  • EP4210019B1 patent drawingFigure 3

AI summary

A method includes, by a processor (30A), acquiring number of persons information that indicates a number of users, including users who ride in vehicles (12), who depart from a facility at each of a predetermined time period, weather information for each predetermined time period, and vehicle information relating to vehicles (12) traveling in a periphery of the facility, determining traffic jam status that indicates absence/presence of a traffic jam on a road located in a vicinity of the facility in the predetermined time period, by using the vehicle information; and generating a learned model (150) for predicting a traffic jam of a road by machine learning using, as teaching data, the number of persons information, the weather information, and the traffic jam status that is associated with the number of persons information and the weather information.