BenchMARL

In this notebooks, we train a policy that outputs acceleration and message using MARL algorithms from BenchMARL.

🟧 + 🔵 -> 🔶 + 🟠

• Continuous actions with MLP
  • MASAC
  • MAPPO
• Discrete actions
  • MASAC
  • MAPPO
• Continuous actions using split MLP
  • MASAC
  • Alternate Action and Communication training
• Unidirection communication