def make_world(self):
world = HazardousWorld()
world.dim_c = 2
# add landmarks
# TODO: decide on desired landmark properties instead of using a 'default' collection
world.landmarks = []
world.landmarks.append(RiskRewardLandmark(risk_fn=None,
reward_fn=PolynomialRewardFunction2D(coefs=[10.0, 0.0, 0.0, -1.0, 0.0, -1.0],
bounds={'xmin':-2.0, 'xmax':2.0, 'ymin':-2.0, 'ymax':2.0})))
for i, landmark in enumerate(world.landmarks):
landmark.name = 'landmark %d' % i
landmark.collide = True
landmark.movable = False
landmark.size = DP.landmark_size
# properties for landmarks
if isinstance(landmark, RiskRewardLandmark) and landmark.is_hazard:
landmark.color = np.array([landmark.risk(landmark.size) + .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=0.0)
# observation-based communication
world.dim_c = 0
world.max_communication_distance = _MAX_COMMUNICATION_DISTANCE
# collaborative rewards
world.collaborative = True
world.systemic_rewards = False
world.identical_rewards = False
# add landmarks
world.landmarks = []
for lm in _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
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
def make_world(self):
world = HazardousWorld()
# observation-based communication
world.dim_c = 0
world.max_communication_distance = DP.max_communication_distance
# add landmarks
world.landmarks = [Landmark() for i in range(self.num_landmarks)]
for i, landmark in enumerate(world.landmarks):
landmark.name = 'landmark %d' % i
landmark.collide = True
landmark.movable = False
landmark.size = DP.landmark_size
# add obstacles
world.obstacles = [Obstacle() for i in range(self.num_obstacles)]
for i, obstacle in enumerate(world.obstacles):
obstacle.name = 'obstacle %d' % i
obstacle.collide = True
obstacle.size = 0.05
# make initial conditions
self.reset_world(world)
return world
def make_world(self):
world = HazardousWorld()
# set scenario-specific world parameters
world.collaborative = True
world.systemic_rewards = True
world.identical_rewards = False
world.dim_c = 0 # observation-based communication
world.max_communication_distance = _MAX_COMMUNICATION_DISTANCE
world.distance_resistance_gain = _DISTANCE_RESISTANCE_GAIN
# create and add terminal landmarks
# no intermediate landmarks in this scenario
world.origin_terminal_landmark = RiskRewardLandmark(
risk_fn=None, reward_fn=RadialReward(1.0, 10.0))
world.origin_terminal_landmark.name = 'origin'
world.origin_terminal_landmark.state.p_pos = np.array([-0.75, -0.75])
world.destination_terminal_landmark = RiskRewardLandmark(
risk_fn=None, reward_fn=RadialReward(1.0, 10.0))
world.destination_terminal_landmark.name = 'destination'
world.destination_terminal_landmark.state.p_pos = np.array(
[0.75, 0.75])
# create landmark list and set properties
world.landmarks = [
world.origin_terminal_landmark, world.destination_terminal_landmark
]
for i, landmark in enumerate(world.landmarks):
landmark.p_vel = np.zeros(world.dim_p)
landmark.collide = False
landmark.movable = False
landmark.size = _LANDMARK_SIZE
landmark.color = np.array(
[landmark.risk_fn.get_failure_probability(0, 0) + .1, 0, 0])
# make initial conditions
self.reset_world(world)
return world
def make_world(self):
world = HazardousWorld()
# set scenario-specific world parameters
world.collaborative = True
world.systemic_rewards = True
world.identical_rewards = False
world.dim_c = 0 # observation-based communication
world.connection_reward = _CONNECTION_REWARD
world.termination_reward = _TERMINATION_REWARD
world.render_connections = True
# add landmarks
world.origin_terminal_landmark = RiskRewardLandmark(
risk_fn=None, reward_fn=RadialReward(1.0, 10.0))
world.destination_terminal_landmark = RiskRewardLandmark(
risk_fn=None, reward_fn=RadialReward(1.0, 10.0))
world.hazard_landmark = _NON_TERMINAL_LANDMARKS[0]
world.hazard_landmark.ignore_connection_rendering = True
world.landmarks = [
world.origin_terminal_landmark,
world.destination_terminal_landmark, world.hazard_landmark
]
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()
# set scenario-specific world parameters
world.collaborative = True
world.systemic_rewards = True
world.identical_rewards = False
world.dim_c = 0 # observation-based communication
world.max_communication_distance = _MAX_COMMUNICATION_DISTANCE
world.distance_resistance_gain = _DISTANCE_RESISTANCE_GAIN
world.distance_resistance_exponent = _DISTANCE_RESISTANCE_EXPONENT
# add landmarks
world.origin_terminal_landmark = RiskRewardLandmark( risk_fn=None, reward_fn=RadialReward(1.0, 10.0))
world.destination_terminal_landmark = RiskRewardLandmark( risk_fn=None, reward_fn=RadialReward(1.0, 10.0))
world.landmarks = [world.origin_terminal_landmark, world.destination_terminal_landmark]
for lm in _NON_TERMINAL_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