def make_world(self):
world = HazardousWorld()
# set scenario-specific world parameters
world.collaborative = True
world.systemic_rewards = True
world.identical_rewards = self.identical_rewards
world.dim_c = 0 # observation-based communication
world.connection_reward = _CONNECTION_REWARD
world.termination_reward = _TERMINATION_REWARD
world.render_connections = True
# add landmarks. terminals first then hazards (if any)
# world.origin_terminal_landmark = RiskRewardLandmark( risk_fn=None, reward_fn=RadialReward(1.0, 0.0))
# world.destination_terminal_landmark = RiskRewardLandmark( risk_fn=None, reward_fn=RadialReward(1.0, 0.0))
world.origin_terminal_landmark = Landmark()
world.destination_terminal_landmark = Landmark()
world.landmarks = [
world.origin_terminal_landmark, world.destination_terminal_landmark
]
world.hazard_landmarks = []
for i in range(self.num_hazards):
lm = RiskRewardLandmark(risk_fn=RadialRisk(_HAZARD_SIZE,
self.hazard_risk),
reward_fn=RadialReward(_HAZARD_SIZE, 0.0))
lm.silent = True
lm.deaf = True
lm.ignore_connection_rendering = True
world.hazard_landmarks.append(lm)
world.landmarks.append(lm)
for i, landmark in enumerate(world.landmarks):
landmark.name = 'landmark_%d' % i
landmark.collide = False
landmark.movable = False
landmark.size = _LANDMARK_SIZE
# properties for landmarks
if isinstance(landmark, RiskRewardLandmark) and landmark.is_hazard:
#TODO: make colors heatmap of risk probability over all bounds
landmark.color = np.array([
landmark.risk_fn.get_failure_probability(0, 0) + .1, 0, 0
])
else:
landmark.color = np.array([0.25, 0.25, 0.25])
# make initial conditions
self.reset_world(world)
return world
def make_world(self):
world = HazardousWorld(
collision_termination_probability=self.collision_risk,
flyoff_termination_radius=10.0,
flyoff_termination_speed=50.0,
spontaneous_termination_probability=1.0 /
(8.0 * 50.0 * self.num_agents))
# set scenario-specific world parameters
world.collaborative = True
world.systemic_rewards = True
world.identical_rewards = self.identical_rewards
world.dim_c = 0 # observation-based communication
world.termination_reward = _TERMINATION_REWARD
world.render_connections = True
# add landmarks to world
world.landmarks = []
for lm in self.scenario_landmarks:
world.landmarks.append(lm)
for i, landmark in enumerate(world.landmarks):
landmark.name = 'landmark_%d' % i
landmark.collide = False
landmark.movable = False
landmark.size = _LANDMARK_SIZE
# properties for landmarks
if isinstance(landmark, RiskRewardLandmark) and landmark.is_hazard:
#TODO: make colors heatmap of risk probability over all bounds
landmark.color = np.array([
landmark.risk_fn.get_failure_probability(0, 0) + .1, 0, 0
])
else:
landmark.color = np.array([0.25, 0.25, 0.25])
# make initial conditions
self.reset_world(world)
return world