Smart City Traffic Simulation Using Context-Aware Predictive Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current traffic prediction systems in smart cities rely on bulk historic data and lack contextualization with respect to specific traffic-related actions, making it difficult for authorities to effectively manage and mitigate traffic congestion.
Innovation Solution
A computer-implemented method and apparatus that retrieves training traffic data from a geographic area, determines configurations of smart-city infrastructure corresponding to specific times, and trains a predictive model to predict traffic-related key performance indicators, allowing for informed decision-making on traffic management actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bulk historic traffic data is used for traffic predictions, then the prediction system can operate with simple data collection, but the prediction accuracy lacks contextualization with respect to specific traffic-related actions
Solution Approach 1:
The patent segments traffic data by associating it with specific configurations of traffic-related actions (e.g., traffic signal timing, road closures, lane assignments). Instead of using bulk historic data, the system divides data into context-specific groups where each group corresponds to a particular set of traffic management actions, enabling accurate predictions for specific scenarios
Solution Approach 2:
The patent introduces configuration data as an intermediary that links traffic data to specific traffic-related actions. This configuration information acts as a mediator that provides contextualization, connecting the raw traffic data with the actions taken (such as traffic signal timing or road closures) to enable context-aware predictions
2Ease of operation
If traffic predictions are based on history-based simulation without configuration context, then the system complexity remains low, but authorities cannot effectively determine which actions will yield the best results
Solution Approach 1:
The patent applies preliminary action by pre-training predictive models for different configurations of traffic-related actions before actual traffic management decisions are needed. The system prepares multiple predictive models corresponding to different action configurations (e.g., different traffic signal timing patterns) in advance, so when a decision is needed, authorities can quickly evaluate which pre-trained model predicts the best outcome without performing complex real-time analysis
Solution Approach 2:
The patent utilizes parameter changes by training predictive models with different parameter sets corresponding to various traffic-related action configurations. Each model is trained with specific parameters representing different traffic management scenarios (such as different signal timing parameters, lane assignments, or speed limits), allowing the system to select the appropriate model based on the desired action
Data Source
AI summary
An approach is provided for providing data-driven traffic simulations for ad-hoc reconfigurations of a smart-city infrastructure. The approach involves retrieving training traffic data collected from a geographic area supported by the smart-city infrastructure. The approach also involves determining one or more configurations of the smart-city infrastructure corresponding to one or more times at which the training traffic data was collected, wherein the one or more configurations indicate respective states of one or more traffic-related actions supported by the smart-city infrastructure. The approach further involves training a predictive model to predict a traffic-related key performance indicator based on the training traffic data and the one or more configurations, wherein the predictive model is used to predict the traffic-related key performance indicator for a reconfiguration of at least one of the one or more traffic-related actions.