def add_hurricane_locations():
driver = get_neo4j_driver()
with driver.session() as session:
def create_hurricane(tx, name, lat, lon):
tx.run(
"CREATE (l1:Location {name: $name, latitude: $lat, longitude: $lon})",
name=name,
lat=lat,
lon=lon)
# Add locations where measurements need to be made
session.write_transaction(create_hurricane, "Atlantic1", 18.607736,
-61.288092)
session.write_transaction(create_hurricane, "Atlantic2", 26.308562,
-75.447379)
session.write_transaction(create_hurricane, "Atlantic3", 25.957117,
-78.542334)
session.write_transaction(create_hurricane, "Atlantic4", 27.790243,
-94.164242)
session.write_transaction(create_hurricane, "Atlantic5", 11.426051,
-81.619162)
session.write_transaction(create_hurricane, "Pacific1", 33.079810,
143.703265)
session.write_transaction(create_hurricane, "Pacific2", 22.183588,
136.097657)
session.write_transaction(create_hurricane, "Pacific3", 14.090973,
130.073797)
session.write_transaction(create_hurricane, "Pacific4", 6.680097,
130.780933)
session.write_transaction(create_hurricane, "Pacific5", 0.427483,
138.535974)
def obtain_access_times(mission_id):
# Connect to database, open session
driver = get_neo4j_driver()
with driver.session() as session:
# 1. Obtain a list of satellites that can participate in the mission from the Knowledge Graph
satellites = retrieve_available_satellites(mission_id, session)
# 2. Download the TLEs for the satellites
satellites = add_tle_information(satellites)
# 3. Get mission information
mission = get_mission_information(mission_id, session)
# 4. Save all required information on a file for Orekit:
print(mission)
print_orekit_info(satellites, mission)
driver.close()
# 5. Call Orekit and wait for the results before continuing
jar_path = Path(
os.environ.get("PROPAGATOR_JAR", "./jar_files/propagator.jar"))
orekit_process = subprocess.run(
["java", "-jar", str(jar_path)], cwd=os.getcwd())
# 5. Read Orekit results from file and put them into the right format for this code
java_accesses_path = Path("./int_files/accesses.json")
accesses_folder = Path("./int_files/accesses")
accesses_path = accesses_folder / f'{mission["locations"][0]["name"]}.json'
accesses_folder.mkdir(parents=True, exist_ok=True)
shutil.copyfile(java_accesses_path, accesses_path)
with accesses_path.open("r") as accesses_file:
accesses = json.load(accesses_file)
# Return a map<Satellite, map<Instrument, map<Location, Intervals>>
return accesses
def add_volcano_locations():
driver = get_neo4j_driver()
with driver.session() as session:
def create_volcano(tx, name, lat, lon):
tx.run(
"CREATE (l1:Location {name: $name, latitude: $lat, longitude: $lon})",
name=name,
lat=lat,
lon=lon)
# Add locations where measurements need to be made
session.write_transaction(create_volcano, "Kilauea", 19.4119543,
-155.2747327)
session.write_transaction(create_volcano, "Etna", 37.7510042,
14.9846801)
session.write_transaction(create_volcano, "Piton de la Fournaise",
-21.2494387, 55.7112432)
session.write_transaction(create_volcano, "Stromboli", 38.7918408,
15.1977824)
session.write_transaction(create_volcano, "Merapi", -7.5407171,
110.4282145)
session.write_transaction(create_volcano, "Erta Ale", 13.6069145,
40.6529394)
session.write_transaction(create_volcano, "Ol Doinyo Lengai",
-2.7635781, 35.9056765)
session.write_transaction(create_volcano, "Mount Unzen", 32.7804497,
130.2497246)
session.write_transaction(create_volcano, "Mount Yasur", -19.527192,
169.4307231)
session.write_transaction(create_volcano, "Ambrym", -16.2388854,
168.042517)
def clear_kg():
driver = get_neo4j_driver()
with driver.session() as session:
summary = session.run(
'MATCH (m:Mission), (obs:Observation), (l:Location) '
'DETACH DELETE m, obs, l').consume()
print(summary.counters)
def create_logic():
driver = get_neo4j_driver()
evidence = []
predicates = []
name_matching = {}
predicate_types = set()
with driver.session() as session:
# Generate evidence and predicates for sensors and platforms
results = session.run('MATCH (s:Sensor)-[r:HOSTS]-(p:Platform) '
'WHERE p.status = "Currently being flown" '
'RETURN r')
save_evidence(session, results, evidence, predicates, predicate_types)
# Generate evidence and predicates for sensors and observable properties
results = session.run(
'MATCH (p:Platform)-[:HOSTS]-(s:Sensor)-[r:OBSERVES]-(op:ObservableProperty) '
'WHERE p.status = "Currently being flown" '
'RETURN DISTINCT r')
save_evidence(session, results, evidence, predicates, predicate_types)
# Generate evidence and predicates for mission and observations
results = session.run(
'MATCH (m:Mission)-[r:REQUIRES]-(ob:Observation) '
'WHERE m.mid = 1 '
'RETURN r')
save_evidence(session, results, evidence, predicates, predicate_types)
# Generate evidence and predicates for observations and its observable properties
results = session.run(
'MATCH (m:Mission)-[:REQUIRES]-(ob:Observation)-[r:OBSERVEDPROPERTY]-(op:ObservableProperty) '
'WHERE m.mid = 1 '
'RETURN DISTINCT r')
save_evidence(session, results, evidence, predicates, predicate_types)
# Generate evidence and predicates for platforms being in visibility of the target
visibility_threshold = 0.8
predicate_types.add('inVisibilityOfTarget')
predicates.append({
'type': 'inVisibilityOfTarget',
'node_types': ['Platform']
})
results = session.run('MATCH (p:Platform) '
'WHERE p.status = "Currently being flown" '
'RETURN p')
for record in results:
platform_node = record.value()
randnum = random()
if randnum < visibility_threshold:
evidence.append({
'type': 'inVisibilityOfTarget',
'elements': [platform_node]
})
return evidence, predicates
def add_flood_mission(location):
driver = get_neo4j_driver()
with driver.session() as session:
# Count number of missions to get ID
result = session.run('MATCH (m:Mission) RETURN count(m) as count')
mission_count = result.single()[0]
# Create a sample mission
mission_id = mission_count + 1
summary = session.run(
'CREATE (m:Mission {mid: $mission_id, name: $name, description: $description})',
mission_id=mission_id,
name=f"Mission {mission_id} - Active Flood Monitoring",
description='').consume()
print(summary.counters)
# Add the observations that need to be measured
now_time = datetime.now()
month_time = now_time + timedelta(days=14)
summary = session.run(
'MATCH (op1:ObservableProperty), (op2:ObservableProperty), (op3:ObservableProperty), '
'(op4:ObservableProperty), (m:Mission) '
'WHERE op1.name = "Soil moisture at the surface" AND op2.name = "Precipitation Profile (liquid or solid)" '
'AND op3.name = "Land surface imagery" '
'AND m.mid = $mission_id '
'CREATE (o1:Observation {name: $name1, startDate: $start_date, endDate: $end_date, accuracy: $acc1}), '
'(o2:Observation {name: $name2, startDate: $start_date, endDate: $end_date, accuracy: $acc2}), '
'(o3:Observation {name: $name3, startDate: $start_date, endDate: $end_date, accuracy: $acc3}), '
'(m)-[:REQUIRES]->(o1), (m)-[:REQUIRES]->(o2), (m)-[:REQUIRES]->(o3), '
'(o1)-[:OBSERVEDPROPERTY]->(op1), (o2)-[:OBSERVEDPROPERTY]->(op2), '
'(o3)-[:OBSERVEDPROPERTY]->(op3), ',
mission_id=mission_id,
start_date=now_time,
end_date=month_time,
name1='M1 - Soil moisture',
acc1='1 K',
name2='M1 - Precipitation intensity',
acc2='1 K',
name3='M1 - Land cover',
acc3='10 % confidence',
#name5='M1 - Volcano Gases',
#acc5='0.1'
).consume()
print(summary.counters)
summary = session.run(
'MATCH (m:Mission), (l:Location) '
'WHERE m.mid = $mission_id AND l.name = $location '
'CREATE (m)-[:HASLOCATION]->(l)',
mission_id=mission_id,
location=location).consume()
print(summary.counters)
return mission_id
def compute_probabilities():
paths = Path('./int_files/simulations/')
simulation_probabilities = {"Full Pipeline": [], "Benchmark Team": []}
for simulation_path in [p for p in paths.iterdir() if p.is_dir()]:
simulation_info_path = simulation_path / 'simulation_information.json'
with simulation_info_path.open() as simulation_info_file:
simulation_info = json.load(simulation_info_file)
# Method 1
# Full process (UniKER - Sensing - Verification)
location = simulation_info["location"]
mission_id = simulation_info["mission_id"]
access_intervals = read_access_times(location)
print_kg_reasoning_files(mission_id, access_intervals, simulation_path)
final_path = train_uniker(simulation_path)
satellite_list = merge_results(final_path)
shutil.rmtree(simulation_path / "record", ignore_errors=True)
driver = get_neo4j_driver()
with driver.session() as session:
team = get_sensors_from_satellite_list(session, satellite_list)
team = run_sensor_planner(team, simulation_info)
team_probs_info_path = simulation_path / 'team_probs.json'
with team_probs_info_path.open('w') as team_probs_info_file:
json.dump(team, team_probs_info_file)
optimal_teams = run_verification(team, simulation_path,
simulation_info, access_intervals)
simulation_probabilities["Full Pipeline"].append(optimal_teams)
simulation_results = paths / 'results.json'
with simulation_results.open('w') as simulation_res_file:
simulation_res_file.write(str(simulation_probabilities))
# Method 2
# ...
print(simulation_probabilities)
simulation_results = paths / 'results.json'
with simulation_results.open('w') as simulation_res_file:
simulation_res_file.write(str(simulation_probabilities))
#json.dump(simulation_probabilities, simulation_res_file)
return simulation_probabilities
def add_flood_locations():
driver = get_neo4j_driver()
with driver.session() as session:
def create_flood(tx, name, lat, lon):
tx.run(
"CREATE (l1:Location {name: $name, latitude: $lat, longitude: $lon})",
name=name,
lat=lat,
lon=lon)
# Add locations where measurements need to be made
session.write_transaction(create_flood, "India", 28.236019, 77.729279)
session.write_transaction(create_flood, "Bangladesh", 23.078129,
90.624984)
session.write_transaction(create_flood, "Texas", 29.695276, -95.351883)
session.write_transaction(create_flood, "Italy", 45.421771, 12.141446)
session.write_transaction(create_flood, "Brazil", -3.248639,
-60.007637)
def add_forest_fire_locations():
driver = get_neo4j_driver()
with driver.session() as session:
def create_forest_fire(tx, name, lat, lon):
tx.run(
"CREATE (l1:Location {name: $name, latitude: $lat, longitude: $lon})",
name=name,
lat=lat,
lon=lon)
# Add locations where measurements need to be made
session.write_transaction(create_forest_fire, "Spain", 41.673299,
1.640078)
session.write_transaction(create_forest_fire, "Greece", 39.774301,
22.058433)
session.write_transaction(create_forest_fire, "California", 37.735543,
-120.793321)
session.write_transaction(create_forest_fire, "Washington", 46.094343,
-121.660813)
session.write_transaction(create_forest_fire, "Kenya", -0.547029,
40.058512)
def generate_volcano_simulation(mission_id, access_intervals, eruption_length, eruption_start, location, speed, size,
max_tir_temperature, max_swir_temperature, max_ash_cloud, max_terrain_displacement,
max_so2_levels, simulation_information_path, data_streams_path):
# For all data, we assume second-by-second collection for a week (24*7*3600 = 604800 data points)
# We only save the data points for the times when the target is visible for each instrument
# Furthermore, the volcano eruption starts happening at 3AM on the third day (datapoint 180000)
# The peak of the eruption is at 8AM (dp 198000)
# Eruption stops fifth day 12PM (dp 3852000)
# Data back to nominal after 12 hours (dp 428400)
fake_data_generators = {location: {}}
eruption_start_dp = eruption_start*3600
eruption_max_dp = eruption_start_dp + speed*eruption_length*3600
eruption_slope_dp = eruption_start_dp + 0.9*eruption_length*3600
eruption_end_dp = eruption_start_dp + eruption_length*3600
# Generate TIR data (radiance @ 11 um)
state_changes = [eruption_start_dp, eruption_max_dp, eruption_slope_dp, eruption_end_dp]
states = [{"type": "stable", "mean": 10.},
{"type": "slope", "start": 10., "end": max_tir_temperature},
{"type": "stable", "mean": max_tir_temperature},
{"type": "slope", "start": max_tir_temperature, "end": 10.},
{"type": "stable", "mean": 10.}]
fake_data_generators[location]["Land surface temperature"] = {
"changes": state_changes,
"states": states
}
# Generate SWIR data (radiance @ 4 um)
state_changes = [eruption_start_dp, eruption_max_dp, eruption_slope_dp, eruption_end_dp]
states = [{"type": "stable", "mean": 0.},
{"type": "slope", "start": 0., "end": max_swir_temperature},
{"type": "stable", "mean": max_swir_temperature},
{"type": "slope", "start": max_swir_temperature, "end": 0.},
{"type": "stable", "mean": 0.}]
fake_data_generators[location]["Fire temperature"] = {
"changes": state_changes,
"states": states
}
# Generate Plume data (prob of ash plume)
state_changes = [eruption_start_dp, eruption_max_dp, eruption_slope_dp, eruption_end_dp]
states = [{"type": "stable", "mean": 0.},
{"type": "slope", "start": 0., "end": max_ash_cloud},
{"type": "stable", "mean": max_ash_cloud},
{"type": "slope", "start": max_ash_cloud, "end": 0.},
{"type": "stable", "mean": 0}]
fake_data_generators[location]["Cloud type"] = {
"changes": state_changes,
"states": states
}
# Generate SAR data (mean displacement of terrain in mm)
state_changes = [eruption_start_dp, eruption_max_dp, eruption_slope_dp, eruption_end_dp]
states = [{"type": "slope", "start": 0., "end": 20.},
{"type": "slope", "start": 20., "end": max_terrain_displacement},
{"type": "slope", "start": max_terrain_displacement, "end": 30.},
{"type": "slope", "start": 30., "end": 0.},
{"type": "stable", "mean": 0.}]
fake_data_generators[location]["Land surface topography"] = {
"changes": state_changes,
"states": states
}
# Generate SO2 data (Dobson units)
state_changes = [eruption_start_dp, eruption_max_dp, eruption_slope_dp, eruption_end_dp]
states = [{"type": "stable", "mean": 0.5},
{"type": "slope", "start": 0.5, "end": max_so2_levels},
{"type": "stable", "mean": max_so2_levels},
{"type": "slope", "start": max_so2_levels, "end": 0.5},
{"type": "stable", "mean": 0.5}]
fake_data_generators[location]["Atmospheric Chemistry - SO2 (column/profile)"] = {
"changes": state_changes,
"states": states
}
# Connect to database, open session
driver = get_neo4j_driver()
with driver.session() as session:
satellites = retrieve_available_satellites(mission_id, session)
data_locations_json = {}
for satellite in satellites:
if satellite["name"] in access_intervals["output"]:
data_locations_json[satellite["name"]] = {}
for instrument in satellite["sensors"]:
data_locations_json[satellite["name"]][instrument["name"]] = {}
for observation in instrument["characteristics"]:
data_locations_json[satellite["name"]][instrument["name"]][observation] = {}
for location in access_intervals["output"][satellite["name"]][instrument["name"]]:
if access_intervals["output"][satellite["name"]][instrument["name"]][location]["timeArray"]:
observation_name = slugify(satellite["name"] + "__" + instrument["name"] + "__" + location + "__" + observation)
data_locations_json[satellite["name"]][instrument["name"]][observation][location] = observation_name + ".npy"
array_path = data_streams_path / f"{observation_name}.npy"
#time_array, data_array = generate_fake_timeline(instrument["characteristics"][observation]["Q"],
# access_intervals["output"][satellite["name"]][instrument["name"]][location],
# fake_data_generators[location][observation])
# with open(array_path, 'wb') as f:
# np.save(f, time_array)
# np.save(f, data_array)
print(observation_name)
observable_properties = ["Land surface temperature", "Fire temperature", "Cloud type" , "Land surface topography"]
with simulation_information_path.open('w', encoding='utf8') as simulation_information_file:
simulation_information_json = {
"mission_id": mission_id,
"length": eruption_length,
"start": eruption_start,
"location": location,
"speed": speed,
"size": size,
"max_tir_temperature": max_tir_temperature,
"max_swir_temperature": max_swir_temperature,
"max_ash_cloud": max_ash_cloud,
"max_terrain_displacement": max_terrain_displacement,
"max_so2_levels": max_so2_levels,
"data_locations": data_locations_json,
"observable_properties": observable_properties
}
json.dump(simulation_information_json, simulation_information_file)
def run_verification(original_team, simulation_path: Path, simulation_info, access_intervals):
# data from knowledge graph
driver = get_neo4j_driver()
# Save kg with names first, at the end substitute for indices
with driver.session() as session:
mission_info = get_mission_information(simulation_info["mission_id"], session)
path_to_dict = Path('./int_files/output.dict')
prism_path = Path(os.environ.get("PRISM_PATH", 'D:/Dropbox/darpa_grant/prism/prism/bin'))
print(prism_path)
prism_wsl = (os.environ.get("PRISM_WSL", "yes") == "yes")
# name of files for PRISM (saved to current directory)
mission_file = simulation_path / "prop1.txt" # specification
mdp_filename = "KG_MDP1.txt" # MDP
output_filename = "output1.txt" # output log
# Make paths absolute
mission_file = mission_file.resolve()
simulation_path = simulation_path.resolve()
# Iterate teams until we have a manageable number of states (~1000)
entity_dict, inv_entity_dict = retrieve_entity_dict(driver)
num_states = inf
base_team = copy.deepcopy(original_team)
num_agents = 15
while num_states > 1000:
mission_length = find_mission_length(mission_info)
base_team = random_team_choice(base_team, num_agents)
team = construct_team_from_list(base_team)
target = simulation_info["location"]
team_time = find_time_bounds(team, target, access_intervals)
prefix_list = ['a', 's', 'm']
a_prefix, s_prefix, m_prefix = prefix_list
team_time_id = generate_team_time_id(entity_dict, team_time, a_prefix, s_prefix)
a_list, s_list = generate_as_lists(team, entity_dict, a_prefix, s_prefix)
m_list = generate_m_list(team, simulation_path / "simulation_information.json", entity_dict, prefix_list[2])
num_asm = [len(a_list), len(s_list), len(m_list)]
num_a, num_s, num_m = num_asm
print('# of agents, sensors, meas: ', num_asm)
if num_s > 16:
num_agents -= 1
continue
check_time(team, team_time_id, m_list, entity_dict, s_prefix, m_prefix)
# relationship matrices
relation_as = construct_as_matrix(team, entity_dict, num_a, num_s, a_prefix, s_prefix, a_list, s_list)
# modules for PRISM MDP
all_states = all_states_as(num_a, num_s, relation_as, a_list, s_list, team_time_id)
num_states = len(all_states) # total number of states
print(f"Num agents: {num_agents}; Num states: {num_states}")
num_agents -= 1
#print(amy_team(team, "probabilities"))
#print(amy_team(team, "times"))
prefix_list = ['a', 's', 'm']
m_list = generate_m_list(team, simulation_path / "simulation_information.json", entity_dict, prefix_list[2])
qout = mp.Queue()
processes = [mp.Process(target=parallelize, args=(team, team_time, entity_dict, inv_entity_dict, mission_file, mdp_filename, output_filename, simulation_path, prism_path, m_list, prefix_list, i, qout, prism_wsl)) for i in range(mission_length)]
for p in processes:
p.start()
for p in processes:
p.join()
result = []
teaming = []
times = []
for p in processes:
result_p, teaming_p, time_dict = qout.get()
result.append(result_p)
teaming.append(teaming_p)
times.append(time_dict)
# merge all teaming dictionaries into one
teams = {k: v for d in teaming for k, v in d.items()}
timestep_dict = {k: v for d in times for k, v in d.items()}
optimal_teaming = pareto_plot_all(result, teams, timestep_dict)
print('\n ===================== OPTIMAL TEAM ===================== ')
#print(result, teams)
print(optimal_teaming)
return optimal_teaming
def print_kg_reasoning_files(mission_id, access_intervals,
simulation_path: Path):
# Generate the Knowledge Base relationship by relationship, saving the entities in a set to later generate the
# dictionary
# Connect to database, open session
driver = get_neo4j_driver()
entities = set()
relations = set()
kg = []
# Save kg with names first, at the end substitute for indices
with driver.session() as session:
# HOSTS
# isInstanceOf
satellites_info = get_all_active_satellites_with_instruments(session)
relations.add("HOSTS")
relations.add("isInstanceOf")
for satellite in satellites_info:
sat_name = satellite["name"]
entities.add(sat_name)
for sensor in satellite["sensors"]:
sensor_name = sensor["name"]
sensor_instance_name = f"{sat_name}|{sensor_name}"
entities.add(sensor_instance_name)
kg.append({
"head": sat_name,
"relationship": "HOSTS",
"tail": sensor_instance_name
})
entities.add(sensor_name)
kg.append({
"head": sensor_instance_name,
"relationship": "isInstanceOf",
"tail": sensor_name
})
# MEASURES
relations.add("OBSERVES")
measures_relations = get_observes_relationships(session)
for relation in measures_relations:
entities.add(relation["head"])
entities.add(relation["tail"])
kg.extend(measures_relations)
# OBSERVEDPROPERTY
relations.add("OBSERVEDPROPERTY")
observedproperty_relations = get_observedproperty_relations(session)
for relation in observedproperty_relations:
entities.add(relation["head"])
entities.add(relation["tail"])
kg.extend(observedproperty_relations)
# REQUIRES
relations.add("REQUIRES")
requires_relations = get_requires_relations(mission_id, session)
for relation in requires_relations:
entities.add(relation["head"])
entities.add(relation["tail"])
kg.extend(requires_relations)
# HASLOCATION
relations.add("HASLOCATION")
haslocation_relations = get_haslocation_relations(mission_id, session)
for relation in haslocation_relations:
entities.add(relation["head"])
entities.add(relation["tail"])
kg.extend(haslocation_relations)
# inVisibilityOfTarget
relations.add("inVisibilityOfTarget")
for sat_name, sat_info in access_intervals["output"].items():
for instr_name, instr_info in sat_info.items():
for target_name, accesses in instr_info.items():
if len(accesses["timeArray"]) > 0:
sensor_instance_name = sat_name + "|" + instr_name
entities.add(sensor_instance_name)
entities.add(target_name)
kg.append({
"head": sensor_instance_name,
"relationship": "inVisibilityOfTarget",
"tail": target_name
})
# SENSORTYPE
relations.add("SENSORTYPE")
sensortype_relations = get_sensortype_relations(session)
for relation in sensortype_relations:
entities.add(relation["head"])
entities.add(relation["tail"])
kg.extend(sensortype_relations)
# SENSORBAND
relations.add("SENSORBAND")
sensorband_relations = get_sensorband_relations(session)
for relation in sensorband_relations:
entities.add(relation["head"])
entities.add(relation["tail"])
kg.extend(sensorband_relations)
# SENSORRULE
relations.add("SENSORRULE")
sensorrule_relations = get_sensorrule_relations(session)
for relation in sensorrule_relations:
entities.add(relation["head"])
entities.add(relation["tail"])
kg.extend(sensorrule_relations)
# TYPEOBSERVES
relations.add("TYPEOBSERVES")
typeobserves_relations = get_typeobserves_relations(session)
for relation in typeobserves_relations:
entities.add(relation["head"])
entities.add(relation["tail"])
kg.extend(typeobserves_relations)
relations.add("canParticipate")
ground_truth = retrieve_available_satellites(mission_id, session)
mission_info = get_mission_information(mission_id, session)
# Print a file with a relation between entities and indices
entities_dict_path = simulation_path / "entities.dict"
inv_entity_dict = {}
with entities_dict_path.open('w', encoding='utf8') as entities_dict_file:
for idx, entity in enumerate(entities):
entities_dict_file.write(f"{idx}\t{entity}\n")
inv_entity_dict[entity] = idx
# Print a file with a relation between predicates and indices
relations_dict_path = simulation_path / "relations.dict"
inv_relation_dict = {}
with relations_dict_path.open('w', encoding='utf8') as relations_dict_file:
for idx, relation in enumerate(relations):
relations_dict_file.write(f"{idx}\t{relation}\n")
inv_relation_dict[relation] = idx
# Print the knowledge base into a file
kg_path = simulation_path / "train.txt"
kg_val_path = simulation_path / "valid.txt"
train_val_split = 0.1
with kg_path.open('w', encoding='utf8') as kg_file, kg_val_path.open(
'w', encoding='utf8') as kg_val_file:
for fact in kg:
if random.random() < train_val_split:
kg_file.write(
f'{fact["head"]}\t{fact["relationship"]}\t{fact["tail"]}\n'
)
kg_val_file.write(
f'{fact["head"]}\t{fact["relationship"]}\t{fact["tail"]}\n'
)
else:
kg_file.write(
f'{fact["head"]}\t{fact["relationship"]}\t{fact["tail"]}\n'
)
# Print a file with the logic rules
src_rules_path = Path("./knowledge_reasoning/MLN_rule.txt")
dst_rules_path = simulation_path / "MLN_rule.txt"
shutil.copy(src_rules_path, dst_rules_path)
(simulation_path / "final_rules").mkdir(exist_ok=True)
shutil.copy(Path("./knowledge_reasoning/fc_observation.txt"),
simulation_path / "final_rules" / "fc_observation.txt")
shutil.copy(Path("./knowledge_reasoning/fc_visibility.txt"),
simulation_path / "final_rules" / "fc_visibility.txt")
# Print a ground truth with the set of satellites we know have a chance of participating at all
ground_truth_path = simulation_path / "test.txt"
with ground_truth_path.open('w', encoding='utf8') as ground_truth_file:
for satellite in ground_truth:
ground_truth_file.write(
f'{satellite["name"]}\tcanParticipate\t{mission_info["name"]}\n'
)
def add_forest_fire_mission(location):
driver = get_neo4j_driver()
with driver.session() as session:
# Count number of missions to get ID
result = session.run('MATCH (m:Mission) RETURN count(m) as count')
mission_count = result.single()[0]
# Create a sample mission
mission_id = mission_count + 1
summary = session.run(
'CREATE (m:Mission {mid: $mission_id, name: $name, description: $description})',
mission_id=mission_id,
name=f"Mission {mission_id} - Active Forest Fire Monitoring",
description='').consume()
print(summary.counters)
# Add the observations that need to be measured
now_time = datetime.now()
month_time = now_time + timedelta(days=14)
summary = session.run(
'MATCH (op1:ObservableProperty), (op2:ObservableProperty), (op3:ObservableProperty), '
'(op4:ObservableProperty), (m:Mission) '
'WHERE op1.name = "Land surface temperature" AND op2.name = "Fire temperature" '
'AND op3.name = "Cloud type" AND op4.name = "Trace gases (excluding ozone)" '
#'AND op5.name = "Atmospheric Chemistry - SO2 (column/profile)" '
'AND m.mid = $mission_id '
'CREATE (o1:Observation {name: $name1, startDate: $start_date, endDate: $end_date, accuracy: $acc1}), '
'(o2:Observation {name: $name2, startDate: $start_date, endDate: $end_date, accuracy: $acc2}), '
'(o3:Observation {name: $name3, startDate: $start_date, endDate: $end_date, accuracy: $acc3}), '
'(o4:Observation {name: $name4, startDate: $start_date, endDate: $end_date, accuracy: $acc4}), '
#'(o5:Observation {name: $name5, startDate: $start_date, endDate: $end_date, accuracy: $acc5}), '
'(m)-[:REQUIRES]->(o1), (m)-[:REQUIRES]->(o2), (m)-[:REQUIRES]->(o3), '
'(m)-[:REQUIRES]->(o4), '
#'(m)-[:REQUIRES]->(o5), '
'(o1)-[:OBSERVEDPROPERTY]->(op1), (o2)-[:OBSERVEDPROPERTY]->(op2), '
'(o3)-[:OBSERVEDPROPERTY]->(op3), (o4)-[:OBSERVEDPROPERTY]->(op4) ',
#'(o5)-[:OBSERVEDPROPERTY]->(op5)',
mission_id=mission_id,
start_date=now_time,
end_date=month_time,
name1='M1 - Temperature (TIR)',
acc1='1 K',
name2='M1 - Temperature (SWIR)',
acc2='1 K',
name3='M1 - Smoke',
acc3='10 % confidence',
name4='M1 - CO Trace',
acc4='10 cm',
#name5='M1 - Volcano Gases',
#acc5='0.1'
).consume()
print(summary.counters)
summary = session.run(
'MATCH (m:Mission), (l:Location) '
'WHERE m.mid = $mission_id AND l.name = $location '
'CREATE (m)-[:HASLOCATION]->(l)',
mission_id=mission_id,
location=location).consume()
print(summary.counters)
return mission_id
def display_simulation_results(simulation_probabilities):
cdf_line = None
cdf2_line = None
plt.ion()
figure = plt.figure(constrained_layout=True, figsize=(15, 8))
widths = [1, 4, 1]
heights = [2, 1]
gs = figure.add_gridspec(ncols=3,
nrows=2,
width_ratios=widths,
height_ratios=heights)
earth_axes = figure.add_subplot(gs[0, 1])
earth_axes.set_title('Eruption locations and sizes')
earth_axes.set_xlabel('Longitude (deg)')
earth_axes.set_ylabel('Latitude (deg)')
sim_info = figure.add_subplot(gs[0, 2])
sim_info.axis('off')
sim_text = sim_info.text(0.05,
0.95,
"",
transform=sim_info.transAxes,
fontsize=12,
verticalalignment='top')
cdf_axes = figure.add_subplot(gs[1, 1])
cdf_axes.set_title('Montecarlo Results')
cdf_axes.set_xlabel('Simulation number')
cdf_axes.set_ylabel('Probability of mission success')
cdf_info = figure.add_subplot(gs[1, 2])
cdf_info.axis('off')
cdf_text = cdf_info.text(0.05,
0.95,
"",
transform=cdf_info.transAxes,
fontsize=12,
verticalalignment='top')
mng = plt.get_current_fig_manager()
#mng.window.state('zoomed') # works fine on Windows!
plt.show()
path = geopandas.datasets.get_path('naturalearth_lowres')
earth_info = geopandas.read_file(path)
earth_info.plot(ax=earth_axes, facecolor='none', edgecolor='black')
simulations_path = Path("./int_files/simulations/")
# Connect to database, open session
driver = get_neo4j_driver()
# Updates
success_probs = []
success_probs_bench = []
x_axis = []
for simulation_idx, folder in enumerate(
[x for x in simulations_path.iterdir() if x.is_dir()]):
simulation_path = folder / "simulation_information.json"
with simulation_path.open("r") as simulation_file:
simulation_info = json.load(simulation_file)
with driver.session() as session:
result = session.run(
'MATCH (l:Location) '
'WHERE l.name=$name RETURN DISTINCT l;',
name=simulation_info["location"])
record = result.single()
location_info = {
"name": record["l"]["name"],
"latitude": record["l"]["latitude"],
"longitude": record["l"]["longitude"]
}
earth_axes.add_artist(
plt.Circle((location_info["longitude"], location_info["latitude"]),
simulation_info["size"] * 0.1 / 10000 * 300,
ec="red",
fill=True,
fc="orange"))
sim_text.set_text(generate_simulation_text_info(simulation_info))
# Compute probs for demo video
success_probs_bench.append(0.2)
success_probs_bench.sort()
success_probs.append(
simulation_probabilities["Full Pipeline"][simulation_idx])
success_probs.sort()
x_axis.append(simulation_idx)
if cdf_line is None:
cdf_line = cdf_axes.plot(x_axis,
success_probs,
marker='.',
linestyle='',
label="Full Pipeline")[0]
cdf_line.set_data(x_axis, success_probs)
if cdf2_line is None:
cdf2_line = cdf_axes.plot(x_axis,
success_probs_bench,
color="red",
label="Benchmark Team")[0]
cdf2_line.set_data(x_axis, success_probs_bench)
cdf_actualtext = '\n'.join([
f"Full Pipeline: {np.mean(success_probs):.5f}",
f"Benchmark Team: {np.mean(success_probs_bench):.5f}"
])
cdf_text.set_text(cdf_actualtext)
cdf_axes.legend()
cdf_axes.relim()
cdf_axes.autoscale_view()
# Animation
figure.canvas.draw_idle()
figure.canvas.start_event_loop(0.0001)
figure.canvas.start_event_loop(0)
def main():
simulation_path = Path('./int_files/simulations/simulation_0').resolve()
simulation_info_path = simulation_path / 'simulation_information.json'
with simulation_info_path.open() as simulation_info_file:
simulation_info = json.load(simulation_info_file)
# Method 1
# Full process (UniKER - Sensing - Verification)
location = simulation_info["location"]
mission_id = simulation_info["mission_id"]
access_intervals = read_access_times(location)
# ["Sentinel-1 A", "Sentinel-1 B", "GOES-13", "GOES-14", "GOES-15", "GOES-16", "GOES-17", "Aqua", "Terra"]
satellite_list = [
"Sentinel-1 A", "Sentinel-1 B", "GOES-15", "GOES-17", "Aqua", "Terra"
]
driver = get_neo4j_driver()
with driver.session() as session:
team = get_sensors_from_satellite_list(session, satellite_list)
team = run_sensor_planner(team, simulation_info)
team = construct_team_from_list(team)
team_time = find_time_bounds(team, location, access_intervals)
print(amy_team(team_time, "probabilities"))
print(amy_team(team_time, "times"))
entity_dict, inv_entity_dict = retrieve_entity_dict(driver)
prefix_list = ['a', 's', 'm']
a_prefix, s_prefix, m_prefix = prefix_list
team_time_id = generate_team_time_id(entity_dict, team_time, a_prefix,
s_prefix)
a_list, s_list = generate_as_lists(team, entity_dict, a_prefix, s_prefix)
m_list = generate_m_list(team,
simulation_path / "simulation_information.json",
entity_dict, prefix_list[2])
num_asm = [len(a_list), len(s_list), len(m_list)]
num_a, num_s, num_m = num_asm
print('# of agents, sensors, meas: ', num_asm)
check_time(team, team_time_id, m_list, entity_dict, s_prefix, m_prefix)
# relationship matrices
relation_as = construct_as_matrix(team, entity_dict, num_a, num_s,
a_prefix, s_prefix, a_list, s_list)
# modules for PRISM MDP
all_states = all_states_as(num_a, num_s, relation_as, a_list, s_list,
team_time_id)
num_states = len(all_states) # total number of states
prism_wsl = (os.environ.get("PRISM_WSL", "yes") == "yes")
# name of files for PRISM (saved to current directory)
mission_file = simulation_path / "prop1.txt" # specification
mdp_filename = "KG_MDP1.txt" # MDP
output_filename = "output1.txt" # output log
prism_path = Path(
os.environ.get("PRISM_PATH", 'D:/Dropbox/darpa_grant/prism/prism/bin'))
print(prism_path)
mission_length = 14
qout = mp.Queue()
processes = [
mp.Process(target=parallelize,
args=(team, team_time, entity_dict, inv_entity_dict,
mission_file, mdp_filename, output_filename,
simulation_path, prism_path, m_list, prefix_list, i,
qout, prism_wsl)) for i in range(mission_length)
]
for p in processes:
p.start()
for p in processes:
p.join()
result = []
teaming = []
times = []
for p in processes:
result_p, teaming_p, time_dict = qout.get()
result.append(result_p)
teaming.append(teaming_p)
times.append(time_dict)
# merge all teaming dictionaries into one
teams = {k: v for d in teaming for k, v in d.items()}
timestep_dict = {k: v for d in times for k, v in d.items()}
optimal_teaming = pareto_plot_all(result, teams, timestep_dict)
print('\n ===================== OPTIMAL TEAM ===================== ')
print(optimal_teaming)
print(result)
