Multi-agent (MARL)¶

Multi-agent support is new and has only had superficial testing compared to the single-agent API.

Use make_parallel_env() to create a PettingZoo ParallelEnv.

from mudgym import make_parallel_env

env = make_parallel_env(agents=4)
observations, infos = env.reset()

observations, rewards, terminations, truncations, infos = env.step(
    {agent: "look" for agent in env.agents}
)
env.close()

Agents are named player_0, player_1, and so on. Finished agents are removed from env.agents, so only submit actions for agents still listed there.