Intersection Passage Time Prediction Using Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately predict intersection passage times due to low traffic information collection rates, especially in interrupted flow scenarios, leading to unreliable driving time predictions.
Innovation Solution
A method involving the collection of traffic information, including vehicle speeds through waiting line sections behind intersections, calculation of average passage times, and use of a prediction model trained with deep learning to forecast future passage times, thereby providing accurate route guidance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traffic information is collected from interrupted flow intersections, then passage time prediction can be provided, but the collection rate for traffic information is low leading to inaccurate predictions
Solution Approach 1:
The patent introduces an artificial neural network as an intermediary between the collected traffic information and the passage time prediction. The neural network processes and analyzes the limited traffic data from interrupted flow intersections, extracting meaningful patterns and relationships that would be difficult to obtain through traditional methods. This intermediary component enables accurate passage time prediction despite the low collection rate of raw traffic information.
Solution Approach 2:
The patent transforms the input traffic information parameters through the artificial neural network, changing their representation and relationships. By processing speeds, waiting times, and other traffic parameters through neural network layers, the system extracts new features and patterns that improve prediction accuracy. The neural network changes the parameter space from raw measurements to meaningful predictive features.
2Reliability
If traditional prediction methods are used for interrupted flow, then simple calculations can be performed, but reliable passage time guidance cannot be provided
Solution Approach 1:
The patent replaces traditional mechanical prediction methods (simple calculations based on average speeds and fixed waiting times) with an artificial neural network system. This substitution enables the model to learn complex, non-linear relationships from historical traffic data, significantly improving the reliability of passage time predictions. The neural network captures irregular patterns in interrupted flow that traditional mechanical methods cannot detect.
Solution Approach 2:
The patent performs preliminary training of the artificial neural network using historical traffic information from interrupted flow intersections. This preliminary action prepares the model with learned patterns and relationships before actual passage time prediction is needed. The pre-trained neural network can then provide reliable predictions even with limited real-time data, as it has already learned from extensive historical patterns.
Data Source
AI summary
Disclosed are an apparatus and a method for predicting a passage time and a driving time of vehicles in a waiting line section, which pass through the waiting line section disposed behind an intersection. The apparatus inputs information on speeds of the vehicles to a trained prediction model and more accurately predicts the time for users to arrive at his or her destinations.