posggym 0.3.2

A standard API and diverse set of reference environments for reinforcement learning and planning in Partially Observable Stochastic Games (POSGs).

POSGGym

POSGGym is an open source Python library providing implementations of Partially Observable Stochastic Game (POSG) environments coupled with dynamic models of each environment, all under a unified API. While there are a number of amazing open-source implementations for POSG environments, very few have support for dynamic models that can be used for planning. The aim of this library is to fill this gap. Another aim it to provide open-source implementations for many of the environments commonly used in the Partially-Observable multi-agent planning literature. While some open-source implementations exist for some of the common environments, we hope to provide a central repository, with easy to understand and use implementations in order to make reproducibility easier and to aid in faster research.

POSGGym is directly inspired by and adapted from the Gymnasium (formerly Open AI Gym) and PettingZoo libraries for reinforcement learning. The key addition in POSGGym is the support for environment models. POSGGym's API aims to mimic the Gymnasium API as much as possible while incorporating multiple-agents.

Documentation

The documentation for the project is available at posggym.readthedocs.io/.

Installation

The latest version of POSGGym can be installed by running:

pip install posggym

This will install the base dependencies for running the main environments, but may not include all dependencies for all environments or for rendering some environments. You can install all dependencies for a family of environments like pip install posggym[grid_world] or dependencies for all environments using pip install posggym[all].

We support and test for Python>=3.8.

Environments

POSGGym includes the following families of environments. The code for implemented environments are located in the posggym/envs/ subdirectory.

• Classic - These are classic POSG problems from the literature.
• Grid-World - These environments are all based in a 2D Gridworld.

You can see a list of all environments by running:

import posggym
posggym.pprint_registry()

Environment API

POSGGym models each environment as a python env class. Creating environment instances and interacting with them is very simple, and flows almost identically to the Gymnasium user flow. Here's an example using the PredatorPrey-v0 environment:

import posggym
env = posggym.make("PredatorPrey-v0")

observations, info = env.reset(seed=42)

for t in range(50):
    actions = {i: env.action_spaces[i].sample() for i in env.agents}
    observations, rewards, terminated, truncated, done, info = env.step(actions)

    if done:
        observation, info = env.reset()

env.close()

Model API

Every environment provides access to a model of the environment in the form of a python model class. Each model implements a generative model, which can be used for planning, along with functions for sampling initial states. Some environments also implement a full POSG model including the transition, joint observation and joint reward functions.

The following is an example of accessing and using the environment model:

import posggym
env = posggym.make("PredatorPrey-v0")
model = env.model

model.seed(seed=42)

state = model.sample_initial_state()
observations = model.sample_initial_obs(state)

for t in range(50):
    actions = {i: env.action_spaces[i].sample() for i in model.get_agents(state)}
    state, observations, rewards, terminated, truncated, all_done, info = model.step(state, actions)

    if all_done:
        state = model.sample_initial_state()
        observations = model.sample_initial_obs(state)

The base model API is very similar to the environment API. The key difference that all methods are stateless so can be used repeatedly for planning. Indeed the env class for the built-in environments are mainly just a wrappers over the underlying model class that manage the state and add support for rendering.

Note that unlike for the env class, for convenience the output of the model.step() method is a dataclass instance and so it's components can be accessed as attributes. For example:

result = model.step(state, actions)
observations = result.observations
info = result.info

Both the env and model classes support a number of other methods, please see the documentation TODO for details.

Authors

Jonathon Schwartz - [e-mail]

License

MIT © 2022, Jonathon Schwartz

Versioning

The POSGGym library uses semantic versioning.