def _default_agents(self):
agents = Agents(agent_type=self.agent_type)
obstacles = geom_to_linear_obstacles(self.field.obstacles)
# Add new spawns to the field for the leaders
self.field.spawns.extend([
rectangle(25, 45, 10, 10),
rectangle(80, 65, 10, 10),
rectangle(75, 35, 10, 10),
rectangle(35, 5, 10, 10),
])
for i in range(self.size_leaders):
group_leader = AgentGroup(
agent_type=self.agent_type,
size=1,
attributes=self.attributes(target=i, is_follower=False))
agents.add_non_overlapping_group(
group_leader,
position_gen=self.field.sample_spawn(i + 1),
obstacles=obstacles)
group_herding = AgentGroup(
agent_type=self.agent_type,
size=self.size_herding,
attributes=self.attributes(target=NO_TARGET, is_follower=True))
agents.add_non_overlapping_group(
group_herding,
position_gen=self.field.sample_spawn(0),
obstacles=obstacles)
return agents
def _default_agents(self):
agents = Agents(agent_type=self.agent_type)
group1 = AgentGroup(size=self.size // 2,
agent_type=self.agent_type,
attributes=self.attributes1)
group2 = AgentGroup(size=self.size // 2,
agent_type=self.agent_type,
attributes=self.attributes2)
agents.add_non_overlapping_group(
group=group1, position_gen=self.field.sample_spawn(0))
agents.add_non_overlapping_group(
group=group2, position_gen=self.field.sample_spawn(1))
return agents
def agents_three_circle(size=10):
agents = Agents(agent_type=ThreeCircle)
group = AgentGroup(agent_type=ThreeCircle,
size=size,
attributes=attributes)
agents.add_non_overlapping_group(
group, position_gen=lambda: np.random.uniform(-10.0, 10.0, 2))
return agents
def test_agent_group(agent_type, attributes):
group = AgentGroup(size=SIZE, agent_type=agent_type, attributes=attributes)
assert isinstance(group.size, int)
assert issubclass(group.agent_type, AgentType)
assert isinstance(group.attributes, (Collection, Generator, Callable))
assert isinstance(group.members, list)
assert len(group.members) == SIZE
def _default_agents(self):
agents = Agents(agent_type=self.agent_type)
group_leader = AgentGroup(agent_type=self.agent_type,
size=self.size_leaders,
attributes=self.attributes(is_leader=True))
agents.add_non_overlapping_group(
group_leader, position_gen=self.field.sample_spawn(0))
group_herding = AgentGroup(agent_type=self.agent_type,
size=self.size_herding,
attributes=self.attributes(is_leader=False))
agents.add_non_overlapping_group(
group_herding, position_gen=self.field.sample_spawn(0))
return agents
def test_agent_agent_block_list(benchmark, size, agent_type, algorithm):
# Grow the area with size. Keeps agent density constant.
area_size = np.sqrt(2 * size)
agents = Agents(agent_type=agent_type)
group = AgentGroup(agent_type=agent_type, size=size, attributes=attributes)
agents.add_non_overlapping_group(
group,
position_gen=lambda: np.random.uniform(-area_size, area_size, 2))
benchmark(agent_agent_block_list, agents.array)
assert True
def _default_agents(self):
agents = Agents(agent_type=self.agent_type)
group_active = AgentGroup(
agent_type=self.agent_type,
size=self.size_leaders,
attributes=self.attributes(has_target=True, is_follower=False))
group_herding = AgentGroup(
agent_type=self.agent_type,
size=self.size_herding,
attributes=self.attributes(has_target=False, is_follower=True))
for group in (group_active, group_herding):
agents.add_non_overlapping_group(
group,
position_gen=self.field.sample_spawn(0),
obstacles=geom_to_linear_obstacles(self.field.obstacles))
return agents
def _default_agents(self):
agents = Agents(agent_type=self.agent_type)
group_leader = AgentGroup(agent_type=self.agent_type,
size=self.size_leaders,
attributes=self.attributes(is_leader=True))
agents.add_non_overlapping_group(
group_leader,
position_gen=lambda: np.array((self.horizontal_ratio * self.width,
self.vertical_ratio * self.height)))
group_herding = AgentGroup(agent_type=self.agent_type,
size=self.size_herding,
attributes=self.attributes(is_leader=False))
agents.add_non_overlapping_group(
group_herding, position_gen=self.field.sample_spawn(0))
return agents
def _default_agents(self):
agents = Agents(agent_type=self.agent_type)
group = AgentGroup(agent_type=self.agent_type,
size=self.size,
attributes=self.attributes)
agents.add_non_overlapping_group(
group, position_gen=self.field.sample_spawn(0))
return agents
def _observe_agents(self, change):
if self.agent_type is not None:
agents = Agents(agent_type=self.agent_type)
group = AgentGroup(agent_type=self.agent_type,
size=1,
attributes=partial(self.attributes,
orientation=0.0))
agents.add_non_overlapping_group(
group, position_gen=lambda: self.start_pos1)
group = AgentGroup(agent_type=self.agent_type,
size=1,
attributes=partial(self.attributes,
orientation=np.pi))
agents.add_non_overlapping_group(
group, position_gen=lambda: self.start_pos2)
self.agents = agents
def _observe_agents(self, change):
if self.agent_type is not None:
agents = Agents(agent_type=self.agent_type)
group = AgentGroup(agent_type=self.agent_type,
size=self.size,
attributes=self.attributes)
agents.add_non_overlapping_group(group,
position_gen=lambda: np.array(
(0, 0)))
self.agents = agents
def _default_agents(self):
agents = Agents(agent_type=self.agent_type)
for spawn, name in enumerate(
['finlandiahall', 'foyer', 'helsinkihall']):
group = AgentGroup(agent_type=self.agent_type,
size=getattr(self, f'size_{name}'),
attributes=self.attributes(has_target=True,
is_follower=False))
agents.add_non_overlapping_group(
group,
position_gen=self.field.sample_spawn(spawn),
obstacles=geom_to_linear_obstacles(self.field.obstacles))
return agents
def test_agents(agent_type, attributes):
agents = Agents(agent_type=agent_type)
group = AgentGroup(size=SIZE, agent_type=agent_type, attributes=attributes)
agents.add_non_overlapping_group(
group=group, position_gen=lambda: np.random.uniform(XMIN, XMAX, 2))
assert True
def _default_agents(self):
agents = Agents(agent_type=self.agent_type)
# Generate iterators for group of leaders.
#target_exits = [0,1,0]
#cells = [913,695,652]
target_exits = []
cells = []
n_guides = len(target_exits)
speed_left = "fast"
speed_lower = "fast"
speed_right = "fast"
speed_upper = "fast"
# Exiting agents in left spawn
group_follower_spawn_left = AgentGroup(
agent_type=self.agent_type,
size=getattr(self, 'size_spawn_left'),
attributes=self.attributes(familiar=2, has_target=True, is_follower=True))
agents.add_non_overlapping_group(
speed_left,
"spawn_left",
group_follower_spawn_left,
position_gen=False,
position_iter=iter([]),
spawn=0,
obstacles=geom_to_linear_obstacles(self.field.obstacles))
# Exiting agents in lower spawn
group_follower_spawn_lower = AgentGroup(
agent_type=self.agent_type,
size=getattr(self, 'size_spawn_lower'),
attributes=self.attributes(familiar=3, has_target=True, is_follower=True))
agents.add_non_overlapping_group(
speed_lower,
"spawn_lower",
group_follower_spawn_lower,
position_gen=False,
position_iter=iter([]),
spawn=0,
obstacles=geom_to_linear_obstacles(self.field.obstacles))
# Exiting agents in right spawn
group_follower_spawn_right = AgentGroup(
agent_type=self.agent_type,
size=getattr(self, 'size_spawn_right'),
attributes=self.attributes(familiar=0, has_target=True, is_follower=True))
agents.add_non_overlapping_group(
speed_right,
"spawn_right",
group_follower_spawn_right,
position_gen=False,
position_iter=iter([]),
spawn=2,
obstacles=geom_to_linear_obstacles(self.field.obstacles))
# Exiting agents in upper spawn
group_follower_spawn_upper = AgentGroup(
agent_type=self.agent_type,
size=getattr(self, 'size_spawn_upper'),
attributes=self.attributes(familiar=1, has_target=True, is_follower=True))
agents.add_non_overlapping_group(
speed_upper,
"spawn_upper",
group_follower_spawn_upper,
position_gen=False,
position_iter=iter([]),
spawn=3,
obstacles=geom_to_linear_obstacles(self.field.obstacles))
if n_guides != 0:
init_pos = self.generate_leader_pos(cells, n_guides)
print(init_pos)
# init_pos = [[8, 6]]
target_exits = iter(target_exits)
init_pos = iter(init_pos)
# Guides in Variance Dilemma
group_leader = AgentGroup(
agent_type=self.agent_type,
size=n_guides,
attributes=self.attributes_leader(target_iter=target_exits, has_target=True, is_follower=False))
agents.add_non_overlapping_group(
"group_leader",
group_leader,
position_gen=True,
position_iter=init_pos,
spawn=0,
obstacles=geom_to_linear_obstacles(self.field.obstacles))
return agents
def run(individual, n_leaders, scenario):
# Import simulation
simulation = ComplexFloor()
# Import Hallway floor field
field = simulation.field
# Generate iterators for group of guides.
target_exits = [] # Target exit for the guide
cells = []
if n_leaders > 0:
for i in range(n_leaders):
target_exits.append(individual[i][0])
cells.append(individual[i][1])
# Save the target_exits, in another list for later use
exits = target_exits
# Number of followers
size_spawn_left = 50
size_spawn_lower = 50
size_spawn_right = 50
size_spawn_upper = 50
# Set the familiar exit for the followers based on the scenario.
if scenario == 0:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
# Absolute worst
elif scenario == 1:
familiar_left = 2
familiar_lower = 3
familiar_right = 0
familiar_upper = 1
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
# Quite bad
elif scenario == 2:
familiar_left = 2
familiar_lower = 3
familiar_right = 0
familiar_upper = 1
speed_left = "fast"
speed_lower = "fast"
speed_right = "fast"
speed_upper = "fast"
# Not so bad
elif scenario == 3:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "fast"
speed_lower = "fast"
speed_right = "fast"
speed_upper = "fast"
elif scenario == 4:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
elif scenario == 5:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
elif scenario == 6:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
elif scenario == 7:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
elif scenario == 8:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
elif scenario == 9:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
elif scenario == 10:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
elif scenario == 11:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
elif scenario == 12:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
elif scenario == 13:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
elif scenario == 14:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
else:
familiar_left = 0
familiar_lower = 1
familiar_right = 2
familiar_upper = 3
speed_left = "slow"
speed_lower = "slow"
speed_right = "slow"
speed_upper = "slow"
# Exiting agents in left spawn
group_follower_spawn_left = AgentGroup(agent_type=Circular,
size=size_spawn_left,
attributes=attributes(
simulation,
familiar=familiar_left,
has_target=True,
is_follower=True))
simulation.agents.add_non_overlapping_group(
speed_left,
"spawn_left",
group_follower_spawn_left,
position_gen=False,
position_iter=iter([]),
spawn=0,
obstacles=geom_to_linear_obstacles(field.obstacles))
# Exiting agents in lower spawn
group_follower_spawn_lower = AgentGroup(agent_type=Circular,
size=size_spawn_lower,
attributes=attributes(
simulation,
familiar=familiar_lower,
has_target=True,
is_follower=True))
simulation.agents.add_non_overlapping_group(
speed_lower,
"spawn_lower",
group_follower_spawn_lower,
position_gen=False,
position_iter=iter([]),
spawn=1,
obstacles=geom_to_linear_obstacles(field.obstacles))
# Exiting agents in right spawn
group_follower_spawn_right = AgentGroup(agent_type=Circular,
size=size_spawn_right,
attributes=attributes(
simulation,
familiar=familiar_right,
has_target=True,
is_follower=True))
simulation.agents.add_non_overlapping_group(
speed_right,
"spawn_right",
group_follower_spawn_right,
position_gen=False,
position_iter=iter([]),
spawn=2,
obstacles=geom_to_linear_obstacles(field.obstacles))
# Exiting agents in upper spawn
group_follower_spawn_upper = AgentGroup(agent_type=Circular,
size=size_spawn_upper,
attributes=attributes(
simulation,
familiar=familiar_upper,
has_target=True,
is_follower=True))
simulation.agents.add_non_overlapping_group(
speed_upper,
"spawn_upper",
group_follower_spawn_upper,
position_gen=False,
position_iter=iter([]),
spawn=3,
obstacles=geom_to_linear_obstacles(field.obstacles))
if n_leaders == 0:
# Check if the solution has already been evaluated, if it has print the penalty and total evacuation time and return
bank_evactime = solutionbank(cells, target_exits, n_leaders, scenario)
if bank_evactime != 0:
print(bank_evactime)
return
if n_leaders > 0:
# generate_leader_pos() should check that guides are not spawned in unfeasible positions
init_pos = generate_leader_pos(simulation, cells, n_leaders)
print(init_pos)
target_exits = iter(target_exits)
init_pos = iter(init_pos)
# Check if the solution has already been evaluated, if it has print the penalty and total evacuation time and return
bank_evactime = solutionbank(cells, target_exits, n_leaders, scenario)
if bank_evactime != 0:
print(bank_evactime)
return
# Add guides.
# NB! If there are multiple guides, the function that is set to create the guides should check that the guides do
# not overlap each other.
group_leader = AgentGroup(agent_type=Circular,
size=n_leaders,
attributes=attributes_leader(
simulation,
target_iter=target_exits,
has_target=True,
is_follower=False))
# If it is not taken care before hand that leaders can't overlap, the function will terminate here.
simulation.agents.add_non_overlapping_group(
"dummy_speed",
"group_leader",
group_leader,
position_gen=True,
position_iter=init_pos,
spawn=0,
obstacles=geom_to_linear_obstacles(field.obstacles))
# We do not need to set the seed number, since we use the deterministic social force model
#np.random.seed(seed)
simulation.update()
simulation.run()
print(simulation.data['time_tot'])
# Write the solution to the solution bank
if n_leaders == 0:
banksolution = "{}{}{}{}".format(simulation.data['time_tot'], ' ',
scenario, '\n')
if n_leaders == 1:
banksolution = "{}{}{}{}{}{}{}{}".format(simulation.data['time_tot'],
' ', scenario, ' ', cells[0],
' ', exits[0], '\n')
if n_leaders == 2:
banksolution = "{}{}{}{}{}{}{}{}{}{}{}{}".format(
simulation.data['time_tot'], ' ', scenario, ' ', cells[0], ' ',
exits[0], ' ', cells[1], ' ', exits[1], '\n')
if n_leaders == 3:
banksolution = "{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}".format(
simulation.data['time_tot'], ' ', scenario, ' ', cells[0], ' ',
exits[0], ' ', cells[1], ' ', exits[1], ' ', cells[2], ' ',
exits[2], '\n')
if n_leaders == 4:
banksolution = "{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}".format(
simulation.data['time_tot'], ' ', scenario, ' ', cells[0], ' ',
exits[0], ' ', cells[1], ' ', exits[1], ' ', cells[2], ' ',
exits[2], ' ', cells[3], ' ', exits[3], '\n')
if n_leaders == 5:
banksolution = "{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}".format(
simulation.data['time_tot'], ' ', scenario, ' ', cells[0], ' ',
exits[0], ' ', cells[1], ' ', exits[1], ' ', cells[2], ' ',
exits[2], ' ', cells[3], ' ', exits[3], ' ', cells[4], ' ',
exits[4], '\n')
if n_leaders == 6:
banksolution = "{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}".format(
simulation.data['time_tot'], ' ', scenario, ' ', cells[0], ' ',
exits[0], ' ', cells[1], ' ', exits[1], ' ', cells[2], ' ',
exits[2], ' ', cells[3], ' ', exits[3], ' ', cells[4], ' ',
exits[4], ' ', cells[5], ' ', exits[5], '\n')
if n_leaders == 7:
banksolution = "{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}".format(
simulation.data['time_tot'], ' ', scenario, ' ', cells[0], ' ',
exits[0], ' ', cells[1], ' ', exits[1], ' ', cells[2], ' ',
exits[2], ' ', cells[3], ' ', exits[3], ' ', cells[4], ' ',
exits[4], ' ', cells[5], ' ', exits[5], ' ', cells[6], ' ',
exits[6], '\n')
if n_leaders == 8:
banksolution = "{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}".format(
simulation.data['time_tot'], ' ', scenario, ' ', cells[0], ' ',
exits[0], ' ', cells[1], ' ', exits[1], ' ', cells[2], ' ',
exits[2], ' ', cells[3], ' ', exits[3], ' ', cells[4], ' ',
exits[4], ' ', cells[5], ' ', exits[5], ' ', cells[6], ' ',
exits[6], ' ', cells[7], ' ', exits[7], '\n')
if n_leaders == 9:
banksolution = "{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}".format(
simulation.data['time_tot'], ' ', scenario, ' ', cells[0], ' ',
exits[0], ' ', cells[1], ' ', exits[1], ' ', cells[2], ' ',
exits[2], ' ', cells[3], ' ', exits[3], ' ', cells[4], ' ',
exits[4], ' ', cells[5], ' ', exits[5], ' ', cells[6], ' ',
exits[6], ' ', cells[7], ' ', exits[7], ' ', cells[8], ' ',
exits[8], '\n')
if n_leaders == 10:
banksolution = "{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}{}".format(
simulation.data['time_tot'], ' ', scenario, ' ', cells[0], ' ',
exits[0], ' ', cells[1], ' ', exits[1], ' ', cells[2], ' ',
exits[2], ' ', cells[3], ' ', exits[3], ' ', cells[4], ' ',
exits[4], ' ', cells[5], ' ', exits[5], ' ', cells[6], ' ',
exits[6], ' ', cells[7], ' ', exits[7], ' ', cells[8], ' ',
exits[8], ' ', cells[9], ' ', exits[9], '\n')
if os.path.isfile("complex/bank_complex.out"):
bankfile = open("complex/bank_complex.out", "a")
bankfile.write(banksolution)
bankfile.close()
else:
bankfile = open("complex/bank_complex.out", "w")
bankfile.write(banksolution)
bankfile.close()
return
