Multi-Agent Reinforcement Learning

Summary

• Extends RL to environments with multiple interacting agents.
• Challenges:

  • Non-stationarity: Agents’ policies change over time.
  • Scalability: Large state-action spaces.

• Approaches:

  • Centralized Training, Decentralized Execution: Train a joint policy but allow decentralized decision-making.
  • Multi-Agent Policy Gradients: Extend policy gradients to handle multiple agents.