A Deep Reinforcement Learning Approach Using Asymmetric Self-Play for Robust Multirobot Flocking

Yunjie JIA, Yong SONG, Jiyu CHENG, Jiong JIN, Wei ZHANG, Simon X. YANG, Sam KWONG

Research output: Journal PublicationsJournal Article (refereed)peer-review

Abstract

Flocking control, as an essential approach for survivable navigation of multirobot systems, has been widely applied in fields, such as logistics, service delivery, and search and rescue. However, realistic environments are typically complex, dynamic, and even aggressive, posing considerable threats to the safety of flocking robots. In this article, based on deep reinforcement learning, an Asymmetric Self-play-empowered Flocking Control framework is proposed to address this concern. Specifically, the flocking robots are trained concurrently with learnable adversarial interferers to stimulate the intelligence of the flocking strategy. A two-stage self-play training paradigm is developed to improve the robustness and generalization of the model. Furthermore, an auxiliary training module regarding the learning of transition dynamics is designed, dramatically enhancing the adaptability to environmental uncertainties. Feature-level and agent-level attention are implemented for action and value generation, respectively. Both extensive comparative experiments and real-world deployment demonstrate the superiority and practicality of the proposed framework.
Original languageEnglish
Number of pages10
JournalIEEE Transactions on Industrial Informatics
Early online date23 Jan 2025
DOIs
Publication statusE-pub ahead of print - 23 Jan 2025

Bibliographical note

Publisher Copyright:
© 2005-2012 IEEE.

Keywords

  • Adversarial training
  • autonomous vehicles
  • flocking
  • multiagent deep reinforcement learning (MADRL)

Fingerprint

Dive into the research topics of 'A Deep Reinforcement Learning Approach Using Asymmetric Self-Play for Robust Multirobot Flocking'. Together they form a unique fingerprint.

Cite this