Traffic State Prediction Using Dynamic Assignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current traffic forecasting methods lack accuracy and reliability, particularly in predicting future traffic states, which hinders operators' ability to take proactive measures for smooth traffic flow, as they often rely on statistical approaches that do not incorporate real-world data or spatial correlations between road sections.
Innovation Solution
A computer-implemented method that utilizes time-stamped location data from GPS and other sensors to create detailed speed profiles, turn probabilities, and attraction shares, combined with dynamic traffic assignment methodologies, to provide near-real-time traffic state predictions, enabling operators to control traffic systems effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If pure statistical methods (Machine-Learning and AI techniques) are used for traffic state forecast, then the method can be implemented without explicit transportation theory modeling, but the forecast accuracy and reliability deteriorate due to lack of real-world traffic data incorporation and spatial correlation analysis
Solution Approach 1:
The patent merges statistical time-series mining techniques with explicit transportation theory modeling to create a hybrid approach. This combination integrates the ease of implementation of statistical methods with the forecast accuracy of transportation theory, resolving the contradiction between ease of manufacture and reliability.
Solution Approach 2:
The patent creates a composite forecasting methodology that combines statistical learning models with physics-based transportation models. This composite approach leverages the strengths of both methodologies - the data-driven flexibility of statistical methods and the theoretical rigor of transportation models - to achieve both ease of implementation and high forecast accuracy.
2Reliability
If explicit modeling approach based on transportation theory is used, then the physical interpretation of network and traffic conditions is achieved, but real-world traffic data is not utilized leading to reduced forecast accuracy
Solution Approach 1:
The patent merges statistical time-series mining techniques with explicit transportation theory modeling to create a hybrid approach. This combination integrates the ease of implementation of statistical methods with the forecast accuracy of transportation theory, resolving the contradiction between ease of manufacture and reliability.
Solution Approach 2:
The patent incorporates real-world traffic data as feedback to calibrate and validate the transportation theory models. This feedback mechanism ensures that the physically-based models are grounded in actual traffic conditions, simultaneously achieving physical interpretation accuracy and forecast precision.
3Device complexity
If different road sections are treated as independent without spatial correlation, then the prediction method is simpler to implement, but the overall traffic system forecast accuracy deteriorates
Solution Approach 1:
The patent segments the traffic network into discrete road sections while maintaining spatial correlation through the graph structure. This segmentation allows for manageable computation at each node while preserving the interconnectedness of the overall system, resolving the contradiction between system complexity and prediction reliability.
Solution Approach 2:
The patent introduces the spatial dimension through graph-based representation, where nodes and edges capture the topological relationships between road sections. This dimensional approach allows the system to model spatial correlations efficiently without exponentially increasing computational complexity, achieving both simplicity and reliability.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
Computer system (100), method and computer program product are provided for supporting an operator to control a traffic system (200) including a traffic infrastructure (210, 211 to 213, 221 to 223) configured to allow the movement of real world traffic participants (251 to 253). A state prediction module (130) of the computer system determines, based on time-stamped location data of trajectories, time dependent speed profiles (131), time dependent turn probabilities (132), and time dependent attraction shares (133) corresponding to time dependent turn probabilities. It further determines a state forecast (FC1) for a given future time point based on the time dependent traffic parameters including the speed profiles (131), turn probabilities (132), and attraction shares (133), in conjunction with at least one existing time-dependent origin-destination-matrix (134) and a suitable Sequential Dynamic Traffic assignment methodology.