marl-aquarium 0.1.10

Aquarium: A Comprehensive Framework for Exploring Predator-Prey Dynamics through Multi-Agent Reinforcement Learning Algorithms

Aquarium Environment

A Comprehensive Framework for Exploring Predator-Prey Dynamics through Multi-Agent Reinforcement Learning Algorithms using the pettingzoo interface.

Install

pip install marl-aquarium

Example

from marl_aquarium import aquarium_v0

env = aquarium_v0.env()
env.reset(seed=42)

for agent in env.agent_iter():
    observation, reward, termination, truncation, info = env.last()

    if termination or truncation:
        action = None
    else:
        # this is where you would insert your policy
        action = env.action_space(agent).sample()

    env.step(action)
    env.render()
env.close()

Customize the environment

Parameter	Description	Default Value
render_mode	The mode of rendering. Options include "human" for on-screen rendering and "rgb_array".	"human"
observable_walls	Number of observable walls for the agents.	2
width	The width of the environment window.	800
height	The height of the environment window.	800
caption	The caption of the environment window.	"Aquarium"
fps	Frames per second, controlling the speed of simulation.	60
max_time_steps	Maximum number of time steps per episode.	3000
action_count	Number of possible actions an agent can take.	16
predator_count	Number of predators in the environment.	1
prey_count	Number of prey in the environment.	16
predator_observe_count	Number of predators that can be observed by an agent.	1
prey_observe_count	Number of prey that can be observed by an agent.	3
draw_force_vectors	Whether to draw force vectors for debugging.	False
draw_action_vectors	Whether to draw action vectors for debugging.	False
draw_view_cones	Whether to draw view cones for debugging.	False
draw_hit_boxes	Whether to draw hit boxes for debugging.	False
draw_death_circles	Whether to draw death circles for debugging.	False
fov_enabled	Whether field of view is enabled for agents.	True
keep_prey_count_constant	Whether to keep the prey count constant throughout the simulation.	True
prey_radius	Radius of prey entities.	20
prey_max_acceleration	Maximum acceleration of prey entities.	1.0
prey_max_velocity	Maximum velocity of prey entities.	4.0
prey_view_distance	View distance of prey entities.	100
prey_replication_age	Age at which prey entities replicate.	200
prey_max_steer_force	Maximum steering force of prey entities.	0.6
prey_fov	Field of view for prey entities.	120
prey_reward	Reward for prey survival per time step.	1
prey_punishment	Punishment for prey being caught.	1000
max_prey_count	Maximum number of prey entities in the environment.	20
predator_max_acceleration	Maximum acceleration of predator entities.	0.6
predator_radius	Radius of predator entities.	30
predator_max_velocity	Maximum velocity of predator entities.	5.0
predator_view_distance	View distance of predator entities.	200
predator_max_steer_force	Maximum steering force of predator entities.	0.6
predator_max_age	Maximum age of predator entities.	3000
predator_fov	Field of view for predator entities.	150
predator_reward	Reward for predator catching prey.	10
catch_radius	Radius within which predators can catch prey.	100
procreate	Whether entities can procreate within the environment.	False

env = aquarium_v0.env(
    draw_force_vectors=True,
    draw_action_vectors=True,
    draw_view_cones=True,
    draw_hit_boxes=True,
    draw_death_circles=True,
)