Laser Learning Environment (LLE)

In LLE, agents start on start tiles, collect gems, and finish by reaching exit tiles. When an agent enters a laser of its own colour, it blocks the beam and lets the others pass; entering a laser of any other colour kills it and ends the episode. This single mechanic makes LLE a benchmark for coordination-critical cooperation.

📖 Documentation: https://yamoling.github.io/lle/

LLE

Highlights

⚡ Fast — game logic implemented in Rust, exposed to Python.
🤝 Coordination-critical — lasers can force agents to actively help each other to reach the exit.
🎚️ Two levels of abstraction — a high-level LLE MARL environment, or a low-level World for full control over maps, states, and steps.
🗺️ Custom maps — write a map as a one-line string or a richer TOML file, or use the 6 built-in levels.
🟰 SAT Solver — retrieve solutions to LLE worlds using a SAT-based solver.
🧪 World analysis — analyse the characteristics of a World: does it require cooperation or not? If so, what kind of cooperation?
🐣 Procedural world generation — generate worlds according to your requirements (cooperative, independent, mutually cooperative, ...)
🔍 Rich observations — layered, flattened, partial views, RGB images, and more, with optional reward shaping (PBRS) and multi-objective rewards.

Installation

Install with uv, pip, poetry, …

pip install laser-learning-environment

Quick start

LLE can be used at two levels of abstraction: as an MARLEnv for cooperative multi-agent reinforcement learning, or as a World for fine-grained control.

As a MARL environment

The LLE class wraps a World and implements the MARLEnv interface from the to add a reward function, observations, states, etc. Build one with lle.level(...), lle.from_str(...), or lle.from_file(...), then chain builder methods before build().

Here is an example on the following map: LLE

import lle

env = lle.from_str("S0 G X").obs_type("layered").build()
obs, state = env.reset()
terminal = False
while not terminal:
    # env.render()                 # uncomment to render
    actions = env.sample_action()
    step = env.step(actions)
    # step.reward, step.obs, step.info, ...
    terminal = step.is_terminal # truncated or done

As a `World` for fine-grained control

The World class exposes the state of the world and the events that happen when the agents move.

from lle import World, Action, EventType

world = World("S0 G X")  # linear world: start S0, gem G, exit X
world.reset()
available = world.available_actions()[0]   # [Action.STAY, Action.EAST]

events = world.step([Action.EAST])
assert events[0].event_type == EventType.GEM_COLLECTED
events = world.step([Action.EAST])
assert events[0].event_type == EventType.AGENT_EXIT

You can save and restore the exact state of the world:

import lle

world = lle.World.level(1)
state = world.get_state()
# ...
events = world.set_state(state)

Query the world through properties such as world.start_pos, world.exit_pos, world.gems, world.lasers, and world.agents.

Procedural generation, solving & analysis

The optional generator module provides procedural generation of proven solvable word capabilities. Call lle.generate(...), chain with other methods to describe the characteristics of your world, and end with build() or take(n=...) to generate one or multiple worlds.

pip install laser-learning-environment[generator]

import lle
from lle import World

# A solvable 5x5 world with 2 agents
world = lle.generate(width=5, height=5, n_agents=2).build(seed=0)

# Find the shortest joint plan (or None if unsolvable within t_max steps)
plan = lle.solve(world, 5)

# A world that *requires* cooperation, SAT-verified
coop = lle.generate(width=6, height=6, n_agents=2).lasers(2).cooperative().build()
assert lle.is_cooperative(coop, t_max=15)

# Prove what every short plan requires (e.g. level 6 is mutually cooperative)
assert lle.is_cooperative(World.level(6), t_max=25)

The builder controls every placement decision:

Layout: random(), lanes(), clustered(), or fine-grained starts(...) / exits(...).
Lasers & walls: lasers(n, placement=..., span=...), walls(n, style=...).
Behaviour: solvable() (default), independent(), cooperative(...), mutual(...).

import lle

world = lle.generate(width=5, height=5, n_agents=3).lanes().walls(4, style="shapes").build()
worlds = list(lle.generate(width=5, height=5, n_agents=2).clustered().lasers(2).mutual().cap(10).take(3))

See the examples/ folder for runnable scripts and the documentation for the full API.

Citing our work

The environment has been presented at EWRL 2023 and at BNAIC 2023 where it received the best paper award.

@inproceedings{molinghen2023lle,
  title={Laser Learning Environment: A new environment for coordination-critical multi-agent tasks},
  author={Molinghen, Yannick and Avalos, Raphaël and Van Achter, Mark and Nowé, Ann and Lenaerts, Tom},
  year={2023},
  series={BeNeLux Artificial Intelligence Conference},
  booktitle={BNAIC 2023}
}

Development

Clone the repo, install the Python dependencies, then compile with maturin. The example below uses uv, but conda, poetry, or plain pip work too.

git clone https://github.com/yamoling/lle
uv sync                      # creatte a venv and install python dependencies
uv run maturin dev           # build and install lle in the venv
uv run maturin dev --release # build and install lle in the venv (release mode)

Re-generate the Python bindings in python/lle with one of the following:

cargo run --features python-bindings --bin stub-gen
cargo stub-gen # Alias for the above defined in .cargo/config.toml

Tests

Run the Rust and Python test suites with:

cargo test      # Rust unit + integration tests
maturin dev     # (re)build the extension
pytest          # Python tests

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

laser_learning_environment-2.12.0.tar.gz (244.8 kB view details)

Uploaded Sep 1, 2026 Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

laser_learning_environment-2.12.0-cp314-cp314t-win_arm64.whl (833.7 kB view details)

Uploaded Sep 1, 2026 CPython 3.14tWindows ARM64

laser_learning_environment-2.12.0-cp314-cp314t-win_amd64.whl (877.9 kB view details)

Uploaded Sep 1, 2026 CPython 3.14tWindows x86-64

laser_learning_environment-2.12.0-cp314-cp314t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Sep 1, 2026 CPython 3.14tmanylinux: glibc 2.17+ x86-64

laser_learning_environment-2.12.0-cp314-cp314t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Sep 1, 2026 CPython 3.14tmanylinux: glibc 2.17+ ARM64

laser_learning_environment-2.12.0-cp314-cp314t-macosx_11_0_arm64.whl (997.9 kB view details)

Uploaded Sep 1, 2026 CPython 3.14tmacOS 11.0+ ARM64

laser_learning_environment-2.12.0-cp314-cp314t-macosx_10_12_x86_64.whl (1.0 MB view details)

Uploaded Sep 1, 2026 CPython 3.14tmacOS 10.12+ x86-64

laser_learning_environment-2.12.0-cp314-cp314-win_arm64.whl (836.0 kB view details)

Uploaded Sep 1, 2026 CPython 3.14Windows ARM64

laser_learning_environment-2.12.0-cp314-cp314-win_amd64.whl (877.2 kB view details)

Uploaded Sep 1, 2026 CPython 3.14Windows x86-64

laser_learning_environment-2.12.0-cp314-cp314-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Sep 1, 2026 CPython 3.14manylinux: glibc 2.17+ x86-64

laser_learning_environment-2.12.0-cp314-cp314-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Sep 1, 2026 CPython 3.14manylinux: glibc 2.17+ ARM64

laser_learning_environment-2.12.0-cp314-cp314-macosx_11_0_arm64.whl (997.8 kB view details)

Uploaded Sep 1, 2026 CPython 3.14macOS 11.0+ ARM64

laser_learning_environment-2.12.0-cp314-cp314-macosx_10_12_x86_64.whl (1.0 MB view details)

Uploaded Sep 1, 2026 CPython 3.14macOS 10.12+ x86-64

laser_learning_environment-2.12.0-cp313-cp313-win_arm64.whl (834.5 kB view details)

Uploaded Sep 1, 2026 CPython 3.13Windows ARM64

laser_learning_environment-2.12.0-cp313-cp313-win_amd64.whl (877.5 kB view details)

Uploaded Sep 1, 2026 CPython 3.13Windows x86-64

laser_learning_environment-2.12.0-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Sep 1, 2026 CPython 3.13manylinux: glibc 2.17+ x86-64

laser_learning_environment-2.12.0-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Sep 1, 2026 CPython 3.13manylinux: glibc 2.17+ ARM64

laser_learning_environment-2.12.0-cp313-cp313-macosx_11_0_arm64.whl (996.3 kB view details)

Uploaded Sep 1, 2026 CPython 3.13macOS 11.0+ ARM64

laser_learning_environment-2.12.0-cp313-cp313-macosx_10_12_x86_64.whl (1.0 MB view details)

Uploaded Sep 1, 2026 CPython 3.13macOS 10.12+ x86-64

laser_learning_environment-2.12.0-cp312-cp312-win_arm64.whl (834.8 kB view details)

Uploaded Sep 1, 2026 CPython 3.12Windows ARM64

laser_learning_environment-2.12.0-cp312-cp312-win_amd64.whl (878.4 kB view details)

Uploaded Sep 1, 2026 CPython 3.12Windows x86-64

laser_learning_environment-2.12.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Sep 1, 2026 CPython 3.12manylinux: glibc 2.17+ x86-64

laser_learning_environment-2.12.0-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Sep 1, 2026 CPython 3.12manylinux: glibc 2.17+ ARM64

laser_learning_environment-2.12.0-cp312-cp312-macosx_11_0_arm64.whl (996.5 kB view details)

Uploaded Sep 1, 2026 CPython 3.12macOS 11.0+ ARM64

laser_learning_environment-2.12.0-cp312-cp312-macosx_10_12_x86_64.whl (1.0 MB view details)

Uploaded Sep 1, 2026 CPython 3.12macOS 10.12+ x86-64

laser_learning_environment-2.12.0-cp311-cp311-win_arm64.whl (838.6 kB view details)

Uploaded Sep 1, 2026 CPython 3.11Windows ARM64

laser_learning_environment-2.12.0-cp311-cp311-win_amd64.whl (881.3 kB view details)

Uploaded Sep 1, 2026 CPython 3.11Windows x86-64

laser_learning_environment-2.12.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Sep 1, 2026 CPython 3.11manylinux: glibc 2.17+ x86-64

laser_learning_environment-2.12.0-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Sep 1, 2026 CPython 3.11manylinux: glibc 2.17+ ARM64

laser_learning_environment-2.12.0-cp311-cp311-macosx_11_0_arm64.whl (998.0 kB view details)

Uploaded Sep 1, 2026 CPython 3.11macOS 11.0+ ARM64

laser_learning_environment-2.12.0-cp311-cp311-macosx_10_12_x86_64.whl (1.0 MB view details)

Uploaded Sep 1, 2026 CPython 3.11macOS 10.12+ x86-64

File details

Details for the file laser_learning_environment-2.12.0.tar.gz.

File metadata

Download URL: laser_learning_environment-2.12.0.tar.gz
Upload date: Sep 1, 2026
Size: 244.8 kB
Tags: Source
Uploaded using Trusted Publishing? No
Uploaded via: maturin/1.15.0

File hashes

Hashes for laser_learning_environment-2.12.0.tar.gz
Algorithm	Hash digest
SHA256	`2842ef92bac7116e1983a2c0e896031f97eda5356f26bac596be6264475a49e3`
MD5	`62233bceb684b749916834d6ea94b919`
BLAKE2b-256	`c75ca15337e705f764d4fced8777e95c00ea975b9134167c0899c9f5197f57f7`

Algorithm	Hash digest
SHA256	`87b4513a3fc7f41deee09b5ae9eaf37b812a341dca0b17044b9ab424c1a1560c`
MD5	`dec50d6ef28e89f2c8638553dcdc690e`
BLAKE2b-256	`b6278d9e9450cf37e45b039c67f445034fa36b63896e9c6443d150eb5bd9ea8d`

Algorithm	Hash digest
SHA256	`6a97dd27273ce6b49665c8887cbcb298802b5520784a3f23df02fb7b43548784`
MD5	`eebd27dc8bff51e1457f86336a0b7f5c`
BLAKE2b-256	`343aa7fc6cd8178a38976ff08fe3dc0bf7c2df94ae00f491dfc365f724cec392`

Algorithm	Hash digest
SHA256	`61bf8956bee70ab8655fefee7b212f87e0ddceee5319183663377ea575978532`
MD5	`c14724159ccc20bb39ff1cc4eafa2e65`
BLAKE2b-256	`f95310dfb5c5a970a89ea46a97894ec7c4ecdc58fef406a51ff40a9afa2f6a1a`

Algorithm	Hash digest
SHA256	`8f0b190ed2862ae1e3fc59a82de6f093d086038ba398411f5eb8699a6b8d3958`
MD5	`2e1025ca2c488966bef240c80db93c0e`
BLAKE2b-256	`0f38881d83f7bb1ba846bbdb04b13af1805696f3e45c1994ae8fe0e9dfdfa96e`

Algorithm	Hash digest
SHA256	`a467345a0482976e9d70b899d3473589ed8e78f1282ff1d4d57884e998419be4`
MD5	`6d632cdbd628e9628193da4557eb662d`
BLAKE2b-256	`fdb93d0b992be6abeb9be3b0f25e640f808a30448c8ea97ae8927cf16945127c`

Algorithm	Hash digest
SHA256	`c333dbe0b508a8dc9ea6cf8c93cd0c2b43a501e5f2b0244c1c5b10798d41dac7`
MD5	`0c06e705ca8263cb6976ae1f517e650f`
BLAKE2b-256	`2191b90755712768a9a54a61e89c91293079218d52fb79e90e39d1cb201758f0`

Algorithm	Hash digest
SHA256	`84074aa753cacc48adcc82805f2a688ad90ce5e275aa1f556d043a1dd2a3fabf`
MD5	`c1130c042a93550cbb8f73000306a501`
BLAKE2b-256	`95c39272c832a03c7f9f65489db18f55b136d277d525ba187723b3de8ade27fb`

Algorithm	Hash digest
SHA256	`c180125803d83e87ae548295964a87dc128e340f180d6067fc8b1ff24c392346`
MD5	`124b73f67803bdbd35f7f7df48d4ef42`
BLAKE2b-256	`9fb82a5dc5b4416a653b5fa3d03e76ae793ef26f23c2e7cb90a8cbbacc792c6b`

Algorithm	Hash digest
SHA256	`eb00ba0cdb72cf70be2a7d049e8b727b229d5aa303a7a7867154802a86a50eea`
MD5	`b5fa0bd274274b6ff5f4ab0d4d418fb7`
BLAKE2b-256	`b916b1c6375944d056af30969dba14ebdef5fb885235afbbb42c3ddc4db09432`

Algorithm	Hash digest
SHA256	`6904b519c4d28ea5d058f1e4641987919aebf02418413ba6cfa9033440b3de37`
MD5	`e8471ed7f48e2f6e24c65724293747f1`
BLAKE2b-256	`17732b2af3f2474ca1e321b0404c0b4b26e31e0b4200b1deb03ba3f31ff3df14`

Algorithm	Hash digest
SHA256	`c9cbce5d128df16cb73bb081b0d2b657b7a8c05b20f8b390cefa50ab113a0d3c`
MD5	`77e00337ed55d1a8ea59aceea8991df0`
BLAKE2b-256	`3b38ca9a4ed443f199a48dc4b74e6aa9026ec59bd89356255ed47f3a8819d4cb`

Algorithm	Hash digest
SHA256	`2d3bc9cb32a2bdcb96ac80a374a49316c80adbbd9559163ce8a0c6684ca727f2`
MD5	`8f2cfd156633cb970f3c828f64c403a1`
BLAKE2b-256	`d247c3f0e31a0e383128fa8b379e6380d81b55fabf82eb99eed79e6002248f0c`

Algorithm	Hash digest
SHA256	`feaa850c922a80f44ff53ee0a984cea499c1268ffaed9989a79cb422984dc2f0`
MD5	`131e6b5b043e8d448ef09903d82dd3ac`
BLAKE2b-256	`0fb0139f1687418103408312e8258fc2b9135cab4c5dd0ffb2d2783bf4f6685e`

Algorithm	Hash digest
SHA256	`13762b0d63098ba6f96e11f8bc6aa5ac72ccbe1e14ec2dc622374dbaafc7fde9`
MD5	`d8d2c3bd312d18cf549eff5ddfd28974`
BLAKE2b-256	`01e5671a5d0057088f72d0eb3987b6356a2f1cffa2a19eeed60908e8d3e7a6df`

Algorithm	Hash digest
SHA256	`bb3a9e45c8b1dcdca3c948695acc1f3ecf9394eb25f23497b468234e90122555`
MD5	`bc529e868dbbd57735d1219127075951`
BLAKE2b-256	`89c96aee53759b4db259fa545b046480d13a7e592bba57f97f2352bd837fadd6`

Algorithm	Hash digest
SHA256	`08e949f35969b8b579ee169acf6181406da0c2bb65d7af26310cea1566ccf7ee`
MD5	`d15e00f052a3529c3163ad1be7fba53a`
BLAKE2b-256	`7b2d2a0a6396d857007d66c15ee4fe389913db1f0337996811072df0df8f99ff`

Algorithm	Hash digest
SHA256	`341e778c45980f0991b71e9a92a5f192d73e8bff599c060decbf64128c831354`
MD5	`620f503ab5efd88203731fd82c9dcebc`
BLAKE2b-256	`4e8db6ebfaa53d01f11ddc54075208ff23e337f3507028c7e4d295d03083d3bd`

Algorithm	Hash digest
SHA256	`4bfbe7a5510317ac2e4c48f8a732a0e192d11b536b28c615175c42dc4a242e8a`
MD5	`f797d221be4d2b50bbef57814469082f`
BLAKE2b-256	`18bbadcf9b25fb2e0588dbe2641221e028eeae258a413085f5d08bcf807774e5`

Algorithm	Hash digest
SHA256	`2577c25644428f090b177addb2b453ec96464a2740efee816fbbf10ebe7a1e31`
MD5	`829e638a5f6715d639696b36abd85cee`
BLAKE2b-256	`f7a3324820eadb8e01f748e55f0981f108be849b48fca054eb32f1402782d53f`

Algorithm	Hash digest
SHA256	`10c0575326dc8b9f77b5ee7b7e01a954445d6ae4a56ea0d76dd8a3de1b8b2092`
MD5	`6490380228a31feacd791f604ea2324a`
BLAKE2b-256	`fffe5c3fab8b73fef4773d3f312fa2b1d8b8206ef199aff4fec8bc116bf1f216`

laser-learning-environment 2.12.0

Laser Learning Environment (LLE)

Highlights

Installation

Quick start

As a MARL environment

As a World for fine-grained control

Procedural generation, solving & analysis

Citing our work

Development

Tests

Project links

Key dates

1 maintainer

Credits

Requires

Classifiers

Download files

Source Distribution

Built Distributions

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

As a `World` for fine-grained control

Algorithm	Hash digest
SHA256	`27a36db0ca94568ac651b8265ee64d8060006ab1ef5d20ffdbe46618a358a02c`
MD5	`29ed1f5d742d54933e44822358d31683`
BLAKE2b-256	`9f47409b7623dfa389c2435225b81aceb7490df77c1beb20f29d659a6a870b2f`

Algorithm	Hash digest
SHA256	`337ddd1243b2d50b4533eb1b4ca4be0d729b29826714875512bc7e0274244da3`
MD5	`c94610cdab780dbb59d4cce27f331db8`
BLAKE2b-256	`28928958700ca4954ab7a7c489805406f8de230177e1f10a300a8210a408790a`

Algorithm	Hash digest
SHA256	`43aa9d16831e8668968cbeb149adbd652f9d54ac5053a8cc1b9b8fa3d4d73222`
MD5	`ca936fdc1fa104dccd0cd0698d5bdf4a`
BLAKE2b-256	`605afe3119372a224df99d5d4d766704af7b15cb59b990e68635a1fa7aaf3154`

Algorithm	Hash digest
SHA256	`238a3f899b3433e97ad629764e2a83d287e2cf19fe46fa7545f95c2aa3f71e85`
MD5	`c334d7076ea7caae6dccbcec7991ace8`
BLAKE2b-256	`266d2ddebedb9fabc56bd83003fde657d8218f3f6d9abe0526a72ddda56ed8d2`

Algorithm	Hash digest
SHA256	`6b0f739735a8fcafff8da86ede282f32fd43da1fde9a24fc6a389da8b4c0d049`
MD5	`616af26f231d9084dae0b59a4383576c`
BLAKE2b-256	`b44a5fe2c3e33ce191aa7175230e30d790aae4bbdd35a38dd218ff313c682aa9`

Algorithm	Hash digest
SHA256	`2d624f3aadaab2dfdc9b9d65b1546562d3cbfb025c5e9226f36ca9b3ce0bb3ec`
MD5	`4e25aebc3238c70e050e4d666f4c7acb`
BLAKE2b-256	`05b2d53ea3c785deeb0f6b68629dd89940a78c1a6d0b73114f1ba249d9fb4aef`

Algorithm	Hash digest
SHA256	`715c83f36ac5fdf67e816d36243444c43c27f6130ecb8900e946a42a2228ea45`
MD5	`25c0295ab1c35b0374fb080fa1c298e9`
BLAKE2b-256	`9599dba03dd17dd19c456a61f1e94ada9c3aa06b3882bbd617c269ef08459793`

Algorithm	Hash digest
SHA256	`c78778e3ffcdf37ce9025fa06c94e46fd19c3fcb7e2b0a6ed6df40821e53e67a`
MD5	`86f7e737303b19862797adf3c56aed36`
BLAKE2b-256	`76fee500149f6c37e6e94ff6583534b5b69d79022471a4bf22c31222301fda70`

Algorithm	Hash digest
SHA256	`7ec15a068606169b52bee55d1de5ffbe27c7ee5915223916a05a77c270ad5160`
MD5	`7f7c90cf130c9b64c1bc2004d1778b38`
BLAKE2b-256	`087da83d31a9d9adb20ad3eeb5ebeb3c0bb1fe0acec380ef02347118a93d93b8`

Algorithm	Hash digest
SHA256	`56e22ab2800c4acb16a4721747006e2098be0760079c064a29a4d56c092460ff`
MD5	`d28150040fcbadf5d5719caa12c4dee7`
BLAKE2b-256	`682be179b13e7c9e9913ba494c82820092403bde768d6b028a7bd3f1e3a9cc33`