Traffic Signal Optimization Through Intersection-Group Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing traffic signal control methods based on fixed signals fail to incorporate real-time traffic situations, leading to increased congestion in large-scale road networks and are hindered by the complexity of training multiple intersections using reinforcement learning.
Innovation Solution
A training method and system that segments a large-scale road network into intersection groups, trains reinforcement learning models for each group, and iteratively refines the models until they meet optimization goals, using a training controller and execution units to manage the process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single reinforcement learning agent performs traffic signal optimization on multiple intersections, then the system can control multiple intersections with one model, but the action space increases significantly making training difficult and ineffective
Solution Approach 1:
The patent divides the large-scale road network into multiple intersection groups, where each group is trained by a dedicated reinforcement learning agent. This segmentation reduces the action space for each agent compared to a single agent controlling all intersections, making training more effective and manageable while still achieving network-wide optimization.
2Productivity
If traffic signal control is performed at only some congested intersections, then congestion is reduced in those areas, but congestion increases in other areas due to the balloon effect
Solution Approach 1:
The patent merges multiple intersection groups into a coordinated system where reinforcement learning agents communicate and cooperate. This allows the system to optimize traffic flow across the entire road network rather than isolated intersections, preventing congestion from simply transferring to other areas while maintaining overall network efficiency.
3Ease of operation
If fixed signals based on past traffic volume data are used, then the signal order and duration are predetermined and easy to implement, but real-time traffic situations cannot be incorporated leading to increased congestion
Solution Approach 1:
The patent transitions from static fixed signals to dynamic reinforcement learning-based signal control. The system continuously learns from real-time traffic data and adapts signal timing and duration dynamically, allowing it to respond to changing traffic conditions while maintaining ease of operation through automated decision-making.
Solution Approach 2:
The reinforcement learning agents receive feedback from real-time traffic data and adjust their control strategies accordingly. This feedback mechanism enables the system to incorporate current traffic situations into signal decisions, improving traffic flow efficiency while maintaining operational simplicity through automated adaptation.
Data Source
AI summary
Disclosed is a training method for the optimization of a traffic signal. The method includes segmenting a large-scale road network into intersection groups each including a plurality of intersections, training a reinforcement learning model by allocating a training agent to each intersection group, performing traffic signal control through the inference of an optimal signal based on the trained reinforcement learning model, evaluating whether the results of the execution of the traffic signal control satisfy a preset goal of learning for traffic signal optimization, and repeatedly training the reinforcement learning model based on the results of the evaluation.