POPGym Arcade - GPU-Accelerated POMDPs

POPGym Arcade contains 7 pixel-based POMDPs in the style of the Arcade Learning Environment. Each environment provides:

3 Difficulty settings
Common observation and action space shared across all envs
Fully observable and partially observable configurations
Fast and easy GPU vectorization using jax.vmap and jax.jit

Gradient Visualization

We also provide tools to visualize how policies use memory.

See below for further instructions.

Throughput

You can expect millions of frames per second on a consumer-grade GPU. With obs_size=128, most policies converge within 30-60 minutes of training.

Getting Started

Installation

To install the environments, run

pip install popgym-arcade

If you plan to use our training scripts, install the baselines as well

pip install 'popgym-arcade[baselines]'

Human Play

To best understand the environments, you should try and play them yourself. The play script lets you play the games yourself using the arrow keys and spacebar.

popgym-arcade-play NoisyCartPoleEasy        # play MDP 256 pixel version
popgym-arcade-play BattleShipEasy -p -o 128 # play POMDP 128 pixel version

Creating and Stepping Environments

Our envs are gymnax envs, so you can use your wrappers and code designed to work with gymnax. The following example demonstrates how to integrate POPGym Arcade into your code.

import popgym_arcade
import jax

# Create both POMDP and MDP env variants
pomdp, pomdp_params = popgym_arcade.make("BattleShipEasy", partial_obs=True)
mdp, mdp_params = popgym_arcade.make("BattleShipEasy", partial_obs=False)

# Let's vectorize and compile the envs
# Note when you are training a policy, it is better to compile your policy_update rather than the env_step
pomdp_reset = jax.jit(jax.vmap(pomdp.reset, in_axes=(0, None)))
pomdp_step = jax.jit(jax.vmap(pomdp.step, in_axes=(0, 0, 0, None)))
mdp_reset = jax.jit(jax.vmap(mdp.reset, in_axes=(0, None)))
mdp_step = jax.jit(jax.vmap(mdp.step, in_axes=(0, 0, 0, None)))
    
# Initialize four vectorized environments
n_envs = 4
# Initialize PRNG keys
key = jax.random.key(0)
reset_keys = jax.random.split(key, n_envs)
    
# Reset environments
observation, env_state = pomdp_reset(reset_keys, pomdp_params)

# Step the POMDPs
for t in range(10):
    # Propagate some randomness
    action_key, step_key = jax.random.split(jax.random.key(t))
    action_keys = jax.random.split(action_key, n_envs)
    step_keys = jax.random.split(step_key, n_envs)
    # Pick actions at random
    actions = jax.vmap(pomdp.action_space(pomdp_params).sample)(action_keys)
    # Step the env to the next state
    # No need to reset, gymnax automatically resets when done
    observation, env_state, reward, done, info = pomdp_step(step_keys, env_state, actions, pomdp_params)

# POMDP and MDP variants share states
# We can plug the POMDP states into the MDP and continue playing 
action_keys = jax.random.split(jax.random.key(t + 1), n_envs)
step_keys = jax.random.split(jax.random.key(t + 2), n_envs)
markov_state, env_state, reward, done, info = mdp_step(step_keys, env_state, actions, mdp_params)

Memory Introspection Tools

We implement visualization tools to probe which pixels persist in agent memory, and their impact on Q value predictions. Try code below or vis example to visualize the memory your agent uses

from popgym_arcade.baselines.model.builder import QNetworkRNN
from popgym_arcade.baselines.utils import get_saliency_maps, vis_fn
import equinox as eqx
import jax

config = {
    # Env string
    "ENV_NAME": "NavigatorEasy",
    # Whether to use full or partial observability
    "PARTIAL": True,
    # Memory model type (see models directory)
    "MEMORY_TYPE": "lru",
    # Evaluation episode seed
    "SEED": 0,
    # Observation size in pixels (128 or 256)
    "OBS_SIZE": 128,
}

# Initialize the random key
rng = jax.random.PRNGKey(config["SEED"])

# Initialize the model
network = QNetworkRNN(rng, rnn_type=config["MEMORY_TYPE"], obs_size=config["OBS_SIZE"])
# Load the model
model = eqx.tree_deserialise_leaves("PATH_TO_YOUR_MODEL_WEIGHTS.pkl", network)
# Compute the saliency maps
grads, obs_seq, grad_accumulator = get_saliency_maps(rng, model, config)
# Visualize the saliency maps
# If you have latex installed, set use_latex=True
vis_fn(grads, obs_seq, config, use_latex=False)

Other Useful Libraries

gymnax - The (deprecated) jax-capable gymnasium API
stable-gymnax - A maintained and patched version of gymnax
popgym - The original collection of POMDPs, implemented in numpy
popjaxrl - A jax version of popgym
popjym - A more readable version of popjaxrl environments that served as a basis for our work

Citation

@article{wang2025popgym,
  title={POPGym Arcade: Parallel Pixelated POMDPs},
  author={Wang, Zekang and He, Zhe and Zhang, Borong and Toledo, Edan and Morad, Steven},
  journal={arXiv preprint arXiv:2503.01450},
  year={2025}
}