A variable speed limit control approach for freeway tunnels based on the model-based reinforcement learning framework...

Area of Science:

• Traffic Engineering
• Artificial Intelligence
• Control Systems

Background:

• Freeway tunnels face challenges in maintaining traffic safety and efficiency.
• Traditional variable speed limit (VSL) strategies have limitations in adapting to real-time conditions.

Purpose of the Study:

• To propose a novel VSL control strategy for freeway tunnels using model-based reinforcement learning (MBRL) with safety perception.
• To enhance traffic safety and efficiency in freeway tunnels through an advanced VSL approach.

Main Methods:

• Developed a multi-lane cell transmission model for freeway tunnels as an environment model for MBRL.
• Integrated a real-time crash risk prediction model with random deep and cross networks for safety perception.
• Trained the VSL control agent using a deep dyna-Q method with a safety trigger mechanism.

Main Results:

• The proposed VSL strategies increased traffic safety by 16.00%-20.00% and efficiency by 3.00%-6.50% compared to fixed speed limits.
• Outperformed traditional VSL strategies based on traffic flow prediction and model-free reinforcement learning.
• VSL strategies with safety triggers demonstrated superior safety compared to those without.

Conclusions:

• MBRL-based VSL strategies with safety perception offer significant improvements in freeway tunnel traffic safety and efficiency.
• The developed strategy is effective and adaptable to current tunnel application conditions.
• This approach shows strong potential for real-world implementation in intelligent transportation systems.