Perimeter Control Method of Road Traffic Regions Based on MFD-DDPG

Area of Science:

• Traffic Engineering
• Urban Planning
• Artificial Intelligence

Background:

• Urban expansion leads to significant traffic congestion, hindering economic growth and social activities.
• Effective regional traffic management and control are crucial for urban governance.
• Optimizing traffic control based on regional congestion characteristics is a complex research challenge.

Purpose of the Study:

• To develop advanced traffic control models for urban areas.
• To address regional traffic optimization without relying on real-time traffic determination.
• To apply reinforcement learning algorithms for perimeter control strategies.

Main Methods:

• Focus on the macroscopic fundamental diagram (MFD) for traffic analysis.
• Introduction of Q-learning (QL) algorithm from reinforcement learning.
• Implementation of Deep Deterministic Policy Gradient (DDPG) algorithm from deep reinforcement learning.
• Development of MFD-QL and MFD-DDPG perimeter control models.

Main Results:

• Numerical analysis and simulation experiments were conducted.
• The MFD-QL and MFD-DDPG algorithms demonstrated rapid convergence to stable states.
• Superior control effects were achieved in optimizing regional perimeter control.

Conclusions:

• The proposed MFD-QL and MFD-DDPG models are effective for regional perimeter control.
• Reinforcement learning approaches offer promising solutions for complex traffic management challenges.
• These algorithms contribute to improved urban traffic flow and reduced congestion.