mlsanityData to 7 Oct 2026

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

arXiv 2404.00886 ↗First posted 1 Apr 2024cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Traffic signal control has a great impact on alleviating traffic congestion in modern cities. Deep reinforcement learning (RL) has been widely used for this task in recent years, demonstrating promising performance but also facing many challenges such as limited performances and sample inefficiency. To handle these challenges, MTLight is proposed to enhance the agent observation with a latent state, which is learned from numerous traffic indicators. Meanwhile, multiple auxiliary and supervisory tasks are constructed to learn the latent state, and two types of embedding latent features, the task-specific feature and task-shared feature, are used to make the latent state more abundant. Extensive experiments conducted on CityFlow demonstrate that MTLight has leading convergence speed and asymptotic performance. We further simulate under peak-hour pattern in all scenarios with increasing control difficulty and the results indicate that MTLight is highly adaptable.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 18 of its 63 references that are in the map. First posted on arXiv 1 Apr 2024.

Builds on

18 of 63 references.

References · 18

  1. 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,861
  2. 2017A Survey on Multi-Task Learningcited by 838
  3. 2017An Overview of Multi-Task Learning in Deep Neural Networkscited by 791
  4. 2016Reinforcement Learning with Unsupervised Auxiliary Taskscited by 400
  5. 2019Deep Learning for Spatio-Temporal Data Mining: A Surveycited by 171
  6. 2019Multi-Agent Deep Reinforcement Learning for Large-scale Traffic Signal Controlcited by 159
  7. 2019CoLight: Learning Network-level Cooperation for Traffic Signal Controlcited by 133
  8. 2019CityFlow: A Multi-Agent Reinforcement Learning Environment for Large Scale City Traffic Scenariocited by 114
  9. 2018Multi-Agent Generative Adversarial Imitation Learningcited by 99
  10. 2017Zero-Shot Task Generalization with Multi-Task Deep Reinforcement Learningcited by 97
  11. 2019Learning Phase Competition for Traffic Signal Controlcited by 84
  12. 2020AttendLight: Universal Attention-Based Reinforcement Learning Model for Traffic Signal Controlcited by 60
6 more
  1. 2021On The Effect of Auxiliary Tasks on Representation Dynamicscited by 56
  2. 2019A Geometric Perspective on Optimal Representations for Reinforcement Learningcited by 54
  3. 2020GeneraLight: Improving Environment Generalization of Traffic Signal Control via Meta Reinforcement Learningcited by 17
  4. 2021ModelLight: Model-Based Meta-Reinforcement Learning for Traffic Signal Controlcited by 8
  5. 2019Generalization in multitask deep neural classifiers: a statistical physics approachcited by 6
  6. 2020Exploiting Interpretable Patterns for Flow Prediction in Dockless Bike Sharing Systemscited by 4

Built on by

Not cited yet.