Single Action Environments¶
MaMuJoCo also supports single action Gymansium/MuJoCo/ environments such as Gymnasium/Mujoco/InvertedPendulum and Gymnasium/Mujoco/InvertedDoublePendulum.
And can be instantiated (without a partition):
env = mamujoco_v0.parallel_env("InvertedDoublePendulum", None)
env = mamujoco_v0.parallel_env("InvertedPendulum", None)
In which case, they simply are the same environments with a single agent using the PettingZoo
APIs.
The Purpose of these is to allow researchers to debug multi-agent learning algorithms.
Action Space¶
The shape of the action space depends on the partitioning. The partitioning has the following form:
Observation Space¶
The agent receives the same observations as the single agent Gymnasium environment.
Rewards¶
The agent receive the same reward as the single agent Gymnasium environment.
Starting state¶
The starting state of the environment is the same as single agent Gymnasium environment.
Episode End¶
The agent terminates and truncates at the same time, given the same conditions as the single agent Gymnasium environment.
Version History¶
v1:
Now based on
Gymnasium/MuJoCo-v5
instead ofGymnasium/MuJoCo-v4
(https://github.com/Farama-Foundation/Gymnasium/pull/572).
v0: Initial version release, uses Gymnasium.MuJoCo-v4, and is a fork of the original multiagent_mujuco. No Changes from the original
MaMuJoCo
(schroederdewitt/multiagent_mujoco).