Corpus ID: 202583612

Emergent Tool Use From Multi-Agent Autocurricula

  title={Emergent Tool Use From Multi-Agent Autocurricula},
  author={Bowen Baker and I. Kanitscheider and T. Markov and Yi Wu and Glenn Powell and Bob McGrew and Igor Mordatch},
  • Bowen Baker, I. Kanitscheider, +4 authors Igor Mordatch
  • Published 2020
  • Computer Science, Mathematics
  • ArXiv
  • Through multi-agent competition, the simple objective of hide-and-seek, and standard reinforcement learning algorithms at scale, we find that agents create a self-supervised autocurriculum inducing multiple distinct rounds of emergent strategy, many of which require sophisticated tool use and coordination. We find clear evidence of six emergent phases in agent strategy in our environment, each of which creates a new pressure for the opposing team to adapt; for instance, agents learn to build… CONTINUE READING
    100 Citations
    Natural Emergence of Heterogeneous Strategies in Artificially Intelligent Competitive Teams
    • 1
    • Highly Influenced
    • PDF
    ROMA: Multi-Agent Reinforcement Learning with Emergent Roles
    • 8
    • PDF
    Multi-Agent Collaboration via Reward Attribution Decomposition
    Evolutionary Population Curriculum for Scaling Multi-Agent Reinforcement Learning
    • 5
    • PDF
    Emergent Reciprocity and Team Formation from Randomized Uncertain Social Preferences
    Multi-Agent Reinforcement Learning for Problems with Combined Individual and Team Reward
    • 1
    • PDF
    Continual Learning of Control Primitives: Skill Discovery via Reset-Games


    Emergent Complexity via Multi-Agent Competition
    • 174
    • PDF
    Emergent Coordination Through Competition
    • 51
    • PDF
    Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments
    • 815
    • PDF
    Social Influence as Intrinsic Motivation for Multi-Agent Deep Reinforcement Learning
    • 66
    • PDF
    Learning to Play with Intrinsically-Motivated Self-Aware Agents
    • 49
    • PDF
    Learning to Communicate with Deep Multi-Agent Reinforcement Learning
    • 553
    • PDF
    Emergence of Locomotion Behaviours in Rich Environments
    • 431
    • PDF
    Counterfactual Multi-Agent Policy Gradients
    • 455
    • PDF
    Coevolution of a Backgammon Player
    • 129
    • PDF
    Structured agents for physical construction
    • 17
    • PDF