import numpy as np
from multiagent.core import World, Agent, Landmark
from multiagent.scenario import BaseScenario
from scipy.optimize import linear_sum_assignment
def get_thetas(poses):
# compute angle (0,2pi) from horizontal
thetas = [None]*len(poses)
for i in range(len(poses)):
# (y,x)
thetas[i] = find_angle(poses[i])
return thetas
def find_angle(pose):
# compute angle from horizontal
angle = np.arctan2(pose[1], pose[0])
if angle<0:
angle += 2*np.pi
return angle
class Scenario(BaseScenario):
def __init__(self, num_agents=4, dist_threshold=0.1, arena_size=1, identity_size=0):
self.num_agents = num_agents
self.target_radius = 0.5 # fixing the target radius for now
self.ideal_theta_separation = (2*np.pi)/self.num_agents # ideal theta difference between two agents
self.arena_size = arena_size
self.dist_thres = 0.05
self.theta_thres = 0.1
self.identity_size = identity_size
def make_world(self):
world = World()
# set any world properties first
world.dim_c = 2
num_agents = self.num_agents
num_landmarks = 1
world.collaborative = False
# add agents
world.agents = [Agent(iden=i) for i in range(num_agents)]
for i, agent in enumerate(world.agents):
agent.name = 'agent %d' % i
agent.collide = True
agent.silent = True
agent.size = 0.05
agent.adversary = False
# add landmarks
world.landmarks = [Landmark() for i in range(num_landmarks)]
for i, landmark in enumerate(world.landmarks):
landmark.name = 'landmark %d' % i
landmark.collide = False
landmark.movable = False
landmark.size = 0.03
# make initial conditions
self.reset_world(world)
world.dists = []
return world
def reset_world(self, world):
# random properties for agents
# colors = [np.array([0,0,0.1]), np.array([0,1,0]), np.array([0,0,1]), np.array([1,1,0]), np.array([1,0,0])]
for i, agent in enumerate(world.agents):
agent.color = np.array([0.35, 0.35, 0.85])
# agent.color = colors[i]
# random properties for landmarks
for i, landmark in enumerate(world.landmarks):
landmark.color = np.array([0.25, 0.25, 0.25])
# set random initial states
for agent in world.agents:
agent.state.p_pos = np.random.uniform(-self.arena_size, self.arena_size, world.dim_p)
agent.state.p_vel = np.zeros(world.dim_p)
agent.state.c = np.zeros(world.dim_c)
for i, landmark in enumerate(world.landmarks):
# bound on the landmark position less than that of the environment for visualization purposes
landmark.state.p_pos = np.random.uniform(-.5*self.arena_size, .5*self.arena_size, world.dim_p)
landmark.state.p_vel = np.zeros(world.dim_p)
world.steps = 0
world.dists = []
def reward(self, agent, world):
if agent.iden == 0:
landmark_pose = world.landmarks[0].state.p_pos
relative_poses = [agent.state.p_pos - landmark_pose for agent in world.agents]
thetas = get_thetas(relative_poses)
# anchor at the agent with min theta (closest to the horizontal line)
theta_min = min(thetas)
expected_poses = [landmark_pose + self.target_radius * np.array(
[np.cos(theta_min + i*self.ideal_theta_separation),
np.sin(theta_min + i*self.ideal_theta_separation)])
for i in range(self.num_agents)]
dists = np.array([[np.linalg.norm(a.state.p_pos - pos) for pos in expected_poses] for a in world.agents])
# optimal 1:1 agent-landmark pairing (bipartite matching algorithm)
self.delta_dists = self._bipartite_min_dists(dists)
world.dists = self.delta_dists
total_penalty = np.mean(np.clip(self.delta_dists, 0, 2))
self.joint_reward = -total_penalty
return self.joint_reward
def _bipartite_min_dists(self, dists):
ri, ci = linear_sum_assignment(dists)
min_dists = dists[ri, ci]
return min_dists
def observation(self, agent, world):
# positions of all entities in this agent's reference frame
entity_pos = [entity.state.p_pos - agent.state.p_pos for entity in world.landmarks]
default_obs = np.concatenate([agent.state.p_vel] + [agent.state.p_pos] + entity_pos)
if self.identity_size != 0:
identified_obs = np.append(np.eye(self.identity_size)[agent.iden],default_obs)
return identified_obs
return default_obs
def done(self, agent, world):
condition1 = world.steps >= world.max_steps_episode
self.is_success = np.all(self.delta_dists < self.dist_thres)
return condition1 or self.is_success
def info(self, agent, world):
return {'is_success': self.is_success, 'world_steps': world.steps,
'reward':self.joint_reward, 'dists':self.delta_dists.mean()}
