Traffic Light Control by Multiagent Reinforcement Learning Systems

  title={Traffic Light Control by Multiagent Reinforcement Learning Systems},
  author={Bram Bakker and Shimon Whiteson and Leon Kester and Frans C. A. Groen},
  booktitle={Interactive Collaborative Information Systems},
Traffic light control is one of the main means of controlling road traffic. [] Key Method First, the general multi-agent reinforcement learning framework is described, which is used to control traffic lights in this work.

Intelligent traffic light control using collaborative Q-Learning algorithms

The purpose of this research is to optimize the waiting time at traffic light control based on method of collaborative Q-Learning that can be used as a reference model for the solution of traffic congestion in real world.

Multi-agent reinforcement learning for traffic signal control

This paper forms the TSC problem as a discounted cost Markov decision process (MDP) and applies multi-agent reinforcement learning (MARL) algorithms to obtain dynamic TSC policies and shows that these algorithms perform significantly better than both the standard fixed signal timing (FST) algorithm and the saturation balancing (SAT) algorithm over two real road networks.

STMARL: A Spatio-Temporal Multi-Agent Reinforcement Learning Approach for Cooperative Traffic Light Control

A novel Spatio-Temporal Multi-Agent Reinforcement Learning (STMARL) framework for effectively capturing the spatio-temporal dependency of multiple related traffic lights and control these traffic lights in a coordinating way is proposed.

Deep Reinforcement Learning based Traffic Signal optimization for Multiple Intersections in ITS

  • A. PaulS. Mitra
  • Computer Science
    2020 IEEE International Conference on Advanced Networks and Telecommunications Systems (ANTS)
  • 2020
A single DRL agent manages the traffic signal of multiple intersections using policy gradient algorithm that is trained with spatio-temporal data of the environment that allows it to perform action in different deep neural network models.

Multi-Agent Reinforcement Learning for Integrated Network of Adaptive Traffic Signal Controllers (MARLIN-ATSC)

A novel, decentralized and coordinated adaptive real-time traffic signal control system using Multi-Agent Reinforcement Learning for Integrated Network of Adaptive Traffic Signal Controllers (MARLINATSC) that aims to minimize the total vehicle delay in the traffic network.

IntelliLight: A Reinforcement Learning Approach for Intelligent Traffic Light Control

This paper proposes a more effective deep reinforcement learning model for traffic light control and tests the method on a large-scale real traffic dataset obtained from surveillance cameras.

An Experimental Review of Reinforcement Learning Algorithms for Adaptive Traffic Signal Control

This chapter presents a comprehensive review of the applications of RL to the traffic control problem to date, along with a case study that showcases the developing multi-agent traffic control architecture.

Optimised Traffic Light Management Through Reinforcement Learning: Traffic State Agnostic Agent vs. Holistic Agent With Current V2I Traffic State Knowledge

This study designs representative Deep Reinforcement Learning agents that learn the control of multiple traffic lights without and with current traffic state information, and finds that the holistic system substantially increases average vehicle velocities and flow rates, while reducing CO2 emissions, average wait and trip times, as well as a driver stress metric.

Multi-intersections traffic signal intelligent control using collaborative q-learning algorithm

An urban road traffic area-wide coordination control algorithm based on collaborative Q-learning is proposed and the agent model of traffic intersections is demonstrated, showing that the control algorithm can effectively reduce the average delay time and play a very good control effect with multi-intersections.

Recent Advances in Reinforcement Learning for Traffic Signal Control

This survey focuses on investigating the recent advances in using reinforcement learning (RL) techniques to solve the traffic signal control problem, and classify the known approaches based on the RL techniques they use and provides a review of existing models with analysis on their advantages and disadvantages.



Reinforcement Learning of Traffic Light Controllers Adapting to Traffic Congestion

This paper describes the optimization of traffic light controllers using a multi-agent, model-based reinforcement learning or approximate real-time dynamic programming approach, and shows that this approach outperforms existing methods.

Simulation and optimization of traffic in a city

The experimental results show that the adaptive algorithms can strongly reduce average waiting times of cars compared to three hand-designed controllers.

Reinforcement learning for true adaptive traffic signal control

An introduction to Q-learning, a simple yet powerful reinforcement learning algorithm, is presented and a case study involving application to traffic signal control is presented, which involves optimal control of heavily congested traffic across a two-dimensional road network.

Intelligent Traffic Light Control

This paper studies the simulation and optimization of traffic light controllers in a city and presents an adaptive optimization algorithm based on reinforcement learning that outperform other fixed controllers on all studied infrastructures.


The implementation of an intelligent traffic lights control system using fuzzy logic technology which has the capability of mimicking human intelligence for controlling traffic lights is discussed.

Adaptive traffic signal control using fuzzy logic

  • S. Chiu
  • Computer Science
    Proceedings of the Intelligent Vehicles `92 Symposium
  • 1992
Presents a distributed approach to traffic signal control, where the signal timing parameters are a given intersection are adjusted as functions of the local traffic condition and of the signal

Traffic-responsive signal timing for system-wide traffic control

  • J. SpallD. C. Chin
  • Computer Science
    Proceedings of the 1997 American Control Conference (Cat. No.97CH36041)
  • 1997
The approach is based on a neural network serving as the basis for the control law, with the weight estimation occurring in closed-loop mode via the simultaneous perturbation stochastic approximation (SPSA) algorithm.

Learning to Act Using Real-Time Dynamic Programming

Agent-based learning control method for urban traffic signal of single intersection

This technique that is combined with the rules and Q_learning is applied to urban traffic area to control dynamically the traffic signals in an isolated intersection and the result indicates that the effect of the new method is better.


The implementation of a genetic algorithm (GA) to produce optimal or near-optimal intersection traffic signal timing strategies is described, with a focus on examining this application within a simple traffic situation, giving the reader a clear understanding of how the genetic algorithm is used.