examples

Examples

We include a non-exhaustive number of examples, showing common use-cases for Mava. We also have a Quickstart notebook that can be used to quickly create and train your first Multi-Agent System.

Continuous control

We include a number of systems running on continuous control tasks.

Debugging Environment - Simple Spread

• MADDPG: a MADDPG system running on the continuous action space simple_spread MPE environment.

  • Feedforward:

    • Decentralised

      • decentralised
      • decentralised record agents (recording agents acting in the environment)
      • decentralised executor scaling (scaling to 4 executors)
      • decentralised multiple trainers(using multiple trainers)
      • decentralised custom loggers (using custom logging)
      • decentralised lr scheduling(using lr schedule)
      • decentralised evaluator intervals(running the evaluation loop at intervals)

    • centralised , networked (using a fully-connected, networked architecture), networked with custom architecture (using a custom, sparse, networked architecture) and state_based.

  • Recurrent

    • decentralised and state_based.

• MAD4PG: a MAD4PG system running on the continuous action space simple_spread MPE environment.

  • Feedforward
    • decentralised, centralised and state_based.
  • Recurrent
    • decentralised.

PettingZoo - Multiwalker

• MADDPG: a MADDPG system running on the Multiwalker environment.

  • Feedforward
    • decentralised and centralised.
  • Recurrent
    • decentralised.

• MAD4PG: a MAD4PG system running on the Multiwalker environment.

  • Feedforward
    • decentralised
    • decentralised record agents (recording agents acting in the environment).

• MAPPO

  • Feedforward
    • decentralised.

2D RoboCup

• MAD4PG: a MAD4PG system running on the RoboCup environment.
  • Recurrent
    • state_based.

Discrete control

We also include a number of systems running on discrete action space environments.

Debugging Environment - Simple Spread

• MAPPO: a MAPPO system running on the discrete action space simple_spread MPE environment.

  • Feedforward
    • decentralised and centralised.

• MADQN: a MADQN system running on the discrete action space simple_spread MPE environment.

  • Feedforward
    • Decentralised
      • decentralised
      • decentralised lr scheduling (using lr schedule)
      • decentralised custom lr scheduling (using custom lr schedule)
      • decentralised custom epsilon decay scheduling (using configurable epsilon scheduling).
  • Recurrent
    • decentralised.

• VDN: a VDN system running on the discrete action space simple_spread MPE environment.

  • Recurrent
    • centralised.

PettingZoo - Multi-Agent Atari

• MADQN: a MADQN system running on the two-player competitive Atari Pong environment.
  • Recurrent
    • decentralised.

PettingZoo - Multi-Agent Particle Environment

• MADDPG: a MADDPG system running on the Simple Speaker Listener environment.

  • Feedforward
    • decentralised.

• MADDPG: a MADDPG system running on the Simple Spread environment.

  • Feedforward
    • decentralised.

SMAC - StarCraft Multi-Agent Challenge

• MADQN: a MADQN system running on the SMAC environment.

  • Feedforward
    • decentralised.
  • Recurrent
    • decentralised.

• QMIX: a QMIX system running on the SMAC environment.

  • Recurrent
    • centralised.

• VDN: a VDN system running on the SMAC environment.

  • Recurrent
    • centralised.

OpenSpiel - Tic Tac Toe

• MADQN: a MADQN system running on the OpenSpiel environment.
  • Feedforward
    • decentralised.