pogema

Partially-Observable Grid Environment for Multiple Agents

Partially Observable Multi-Agent Pathfinding (PO-MAPF) is a challenging problem that fundamentally differs from regular MAPF. In regular MAPF, a central controller constructs a joint plan for all agents before they start execution. However, PO-MAPF is intrinsically decentralized, and decision-making, such as planning, is interleaved with execution. At each time step, an agent receives a local observation of the environment and decides which action to take. The ultimate goal for the agents is to reach their goals while avoiding collisions with each other and the static obstacles.

POGEMA stands for Partially-Observable Grid Environment for Multiple Agents. It is a grid-based environment that was specifically designed to be flexible, tunable, and scalable. It can be tailored to a variety of PO-MAPF settings. Currently, the somewhat standard setting is supported, in which agents can move between the cardinal-adjacent cells of the grid, and each action (move or wait) takes one time step. No information sharing occurs between the agents. POGEMA can generate random maps and start/goal locations for the agents. It also accepts custom maps as input.

Installation

For now, install directly from github:

pip install git+https://github.com/Cognitive-AI-Systems/pogema.git

This is a temporary installation option while we are working on Pogema 2.0. The current PyPI package still contains the older version. Once Pogema 2.0 is fully ready, it will also be published on PyPI with a simpler installation command.

Using Example

from pogema import pogema_v0, GridConfig

env = pogema_v0(grid_config=GridConfig())

obs, info = env.reset()

while True:
    # Using random policy to make actions
    obs, reward, terminated, truncated, info = env.step(env.sample_actions())
    env.render()
    if all(terminated) or all(truncated):
        break

Baselines and Evaluation Protocol

The baseline implementations and evaluation pipeline are presented in POGEMA Benchmark repository.

Interfaces

Pogema provides integrations with a range of MARL frameworks: PettingZoo, PyMARL and SampleFactory.

PettingZoo

from pogema import pogema_v0, GridConfig

# Create Pogema environment with PettingZoo interface
env = pogema_v0(GridConfig(integration="PettingZoo"))

PyMARL

from pogema import pogema_v0, GridConfig

env = pogema_v0(GridConfig(integration="PyMARL"))

SampleFactory

from pogema import pogema_v0, GridConfig

env = pogema_v0(GridConfig(integration="SampleFactory"))

Gymnasium

Pogema is fully capable for single-agent pathfinding tasks.

from pogema import pogema_v0, GridConfig

env = pogema_v0(GridConfig(integration="gymnasium"))

Example of training stable-baselines3 DQN to solve single-agent pathfinding tasks:

Customization

Random maps

from pogema import pogema_v0, GridConfig

# Define random configuration
grid_config = GridConfig(num_agents=4,  # number of agents
                         size=8, # size of the grid
                         density=0.4,  # obstacle density
                         seed=1,  # set to None for random 
                                  # obstacles, agents and targets 
                                  # positions at each reset
                         max_episode_steps=128,  # horizon
                         obs_radius=3,  # defines field of view
                         )

env = pogema_v0(grid_config=grid_config)
env.reset()
env.render()

Custom maps

from pogema import pogema_v0, GridConfig

grid = """
.....#.....
.....#.....
...........
.....#.....
.....#.....
#.####.....
.....###.##
.....#.....
.....#.....
...........
.....#.....
"""

# Define new configuration with 8 randomly placed agents
grid_config = GridConfig(map=grid, num_agents=8)

# Create custom Pogema environment
env = pogema_v0(grid_config=grid_config)

Citation

If you use this repository in your research or wish to cite it, please make a reference to our paper:

@inproceedings{skrynnik2025pogema,
  title={POGEMA: A Benchmark Platform for Cooperative Multi-Agent Pathfinding},
  author={Skrynnik, Alexey and Andreychuk, Anton and Borzilov, Anatolii and Chernyavskiy, Alexander and Yakovlev, Konstantin and Panov, Aleksandr},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2025}
}