Abstract:This paper introduces MoveLight, a novel traffic signal control system that enhances urban traffic management through movement-centric deep reinforcement learning. By leveraging detailed real-time data and advanced machine learning techniques, MoveLight overcomes the limitations of traditional traffic signal control methods. It employs a lane-level control approach using the FRAP algorithm to achieve dynamic and adaptive traffic signal control, optimizing traffic flow, reducing congestion, and improving overall efficiency. Our research demonstrates the scalability and effectiveness of MoveLight across single intersections, arterial roads, and network levels. Experimental results using real-world datasets from Cologne and Hangzhou show significant improvements in metrics such as queue length, delay, and throughput compared to existing methods. This study highlights the transformative potential of deep reinforcement learning in intelligent traffic signal control, setting a new standard for sustainable and efficient urban transportation systems.

What problem does this paper attempt to address?

### Problems Addressed by the Paper This paper introduces a novel traffic signal control system called **MoveLight**, which aims to enhance urban traffic management through deep reinforcement learning techniques centered on movement. Specifically, MoveLight overcomes the limitations of traditional traffic signal control methods by leveraging detailed real-time data and advanced machine learning technologies. #### Core Issues - **Alleviating Urban Traffic Congestion**: With the growth of urban populations and the surge in the number of vehicles, the development of urban infrastructure has lagged, leading to severe traffic congestion problems. Traffic signal control can help maximize the utilization of existing resources, improving traffic efficiency without the need for new road construction. - **Dynamic Adaptive Signal Control**: Traditional fixed-time control and actuated control methods perform poorly in dynamically changing traffic environments. MoveLight employs lane-level control methods and uses the FRAP algorithm to achieve dynamic adaptive traffic signal control, optimizing traffic flow, reducing congestion, and improving overall efficiency. #### Main Contributions - **Fine-Grained Data-Driven Approach**: By utilizing detailed real-time data from each lane, higher precision control is achieved, better combining lanes from different directions. - **Multi-Intersection and Network-Level Scalability**: This method is not only applicable to single intersections but also extends to arterial roads and network levels, demonstrating the model's scalability. - **Experimental Validation**: Experiments using real-world datasets from Cologne and Hangzhou show significant improvements in metrics such as queue length, delay, and throughput. In summary, this paper addresses the shortcomings of traditional traffic signal control methods in terms of dynamic adaptability and real-time data processing by proposing the MoveLight system, setting a new standard for intelligent traffic signal control.

MoveLight: Enhancing Traffic Signal Control through Movement-Centric Deep Reinforcement Learning

DNLight: Learning Efficient Evaluation for Traffic Signal Control

ADLight: A Universal Approach of Traffic Signal Control with Augmented Data Using Reinforcement Learning

Intelligent Traffic Light via Policy-based Deep Reinforcement Learning

MetaLight: Value-Based Meta-Reinforcement Learning for Traffic Signal Control

ModelLight: Model-Based Meta-Reinforcement Learning for Traffic Signal Control

PhaseLight: an Universal and Practical Traffic Signal Control Algorithms Based on Reinforcement Learning

MTLight: Efficient Multi-Task Reinforcement Learning for Traffic Signal Control

Deep Reinforcement Learning for Traffic Light Control in Vehicular Networks

IntelliLight: A Reinforcement Learning Approach for Intelligent Traffic Light Control.

Traffic light control with reinforcement learning

SafeLight: A Reinforcement Learning Method toward Collision-free Traffic Signal Control

TransferLight: Zero-Shot Traffic Signal Control on any Road-Network

PDLight: A Deep Reinforcement Learning Traffic Light Control Algorithm with Pressure and Dynamic Light Duration

DenseLight: Efficient Control for Large-scale Traffic Signals with Dense Feedback.

Hierarchically and Cooperatively Learning Traffic Signal Control

A Deep Reinforcement Learning Approach for Traffic Signal Control Optimization

PressLight

DynamicLight: Dynamically Tuning Traffic Signal Duration with DRL

Reinforcement Learning-Based Adaptive Optimal Control Model for Signal Light in Intelligent Transportation Systems

CycLight: learning traffic signal cooperation with a cycle-level strategy