Ultra-scalable JaxMARL based warfare simulation engine.
- Obstacles and terrain integration
- Rasterized maps
- Blast radii simulation
- Friendly fire mechanics
- Pygame visualization
- JAX-based parallelization
pip install parabellum
import parabellum as pb
from jax import random
terrain = pb.terrain_fn("Thun, Switzerland", 1000)
scenario = pb.make_scenario("Thun", terrain, 10, 10)
env = pb.Parabellum(scenario)
rng, key = random.split(random.PRNGKey(0))
obs, state = env.reset(key)
# Simulation loop
for _ in range(100):
rng, rng_act, key_step = random.split(key)
key_act = random.split(rng_act, len(env.agents))
act = {a: env.action_space(a).sample(k) for a, k in zip(env.agents, key_act)}
obs, state, reward, done, info = env.step(key_step, act, state)
# Visualize
vis = pb.Visualizer(env, state_sequence)
vis.animate()
Full documentation: parabellum.readthedocs.io
- Noah Syrkis
- Timothée Anne
- Supervisor: Sebastian Risi
MIT