A Sarsa(λ)-based control model for real-time traffic light coordination.

Area of Science:

• Intelligent Transportation Systems
• Artificial Intelligence in Traffic Management
• Reinforcement Learning Applications

Background:

• Urban traffic congestion results from high vehicle demand, leading to increased waiting times and reduced traffic flow.
• Traditional traffic signal controllers lack adaptive capabilities, failing to adjust to dynamic traffic flow changes.
• Optimizing traffic signal timing is crucial but complex due to numerous influencing factors.

Purpose of the Study:

• To develop an adaptive traffic control model that optimizes traffic signal timing in real-time.
• To address the limitations of conventional traffic controllers by incorporating learning capabilities.
• To improve traffic flow and minimize average vehicle waiting times through intelligent scheduling.

Main Methods:

• Implementation of a reinforcement learning approach, specifically the Sarsa(λ) algorithm, for traffic signal control.
• The model learns optimal actions by evaluating traffic costs, including vehicle delay, waiting vehicle count, and saturation.
• Real-time data processing and policy updates based on learned experiences.

Main Results:

• The Sarsa(λ)-based model demonstrated significant improvements in traffic control efficiency.
• The proposed model effectively maintains an optimized traffic signal timing policy.
• Inspiring enhancements in real-time dynamic traffic control were observed.

Conclusions:

• Reinforcement learning, particularly the Sarsa(λ) algorithm, offers a viable solution for adaptive traffic signal control.
• The developed model effectively learns and adapts to dynamic traffic conditions, optimizing traffic flow and reducing delays.
• The approach shows strong potential for facilitating real-time dynamic traffic control in urban environments.