def __init__(self):
# Start knowledge and connect to the NLP pipeline
self.sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
try:
self.sock.connect(('localhost', DEFAULT_PORT))
except socket.error:
raise IOError("Could not connect to pipelinehost on port %d. "
"Make sure that pipelinehost is running." % DEFAULT_PORT)
self.world_knowledge = Knowledge()
# Sets of propositions, accessible to all methods
self.custom_props = set()
def run(self, aut_path=None):
"""Intialize all NL components."""
# pylint: disable=W0603
global WORLD_KNOWLEDGE
# Start the NL pipeline
if not aut_path:
print "Starting NL pipeline..."
init_pipes()
else:
print "Skipping loading nlpipeline because an automaton was loaded"
# Start the ROS node
print "Initializing ROS node..."
rospy.init_node('nlproxy')
# Set up the state mgr and its callbacks
print "Starting state manager..."
self.state_mgr = StateManager(self)
self.state_mgr.set_basedir(LTLGEN_BASE_DIR)
# Load the automaton if needed
if aut_path:
self.state_mgr.load_test_automaton(aut_path, False)
# Create world knowledge
print "Starting knowledge..."
WORLD_KNOWLEDGE = Knowledge(self)
self.state_mgr.world_knowledge = WORLD_KNOWLEDGE
# Wait a little for ROS to avoid timing startup issues.
print "Waiting for ROS node to settle..."
time.sleep(1)
# Set up ROS listening
print "Subscribing to notifications..."
# Set up state manager's sending and listening
pub = add_ltl_publisher()
self.state_mgr.set_publisher(pub)
add_subscriber(LTL_ENVIRONMENT_TOPIC,
self.state_mgr.process_sensor_data)
add_subscriber(LTL_ENVIRONMENT_TOPIC,
WORLD_KNOWLEDGE.process_sensor_data)
# Create the input comm_proxy and iPad connections
print "Starting comm_proxy..."
self.comm_proxy = CallbackSocket(self.text_port)
self.comm_proxy.register_callback(self.process_text)
self.ipad_conn = iPadConnection(self.map_port)
self.ipad_conn.register_rename_callback(rename_entity)
self.ipad_conn.register_text_callback(self.process_text)
# TODO Add highlight callback
self.map_proxy = MapProxy(self.ipad_conn)
add_subscriber(LTL_ENVIRONMENT_TOPIC, self.ipad_conn.add_icons)
WORLD_KNOWLEDGE.map_proxy = self.map_proxy
# Set up odometry forwarding to the ipad
tf = TransformListener()
while not tf.frameExists("/map"):
rospy.logwarn("Not ready for transforms yet")
rospy.sleep(1.0)
position_proxy = RobotPositionProxy(self.ipad_conn, tf)
rospy.Subscriber(ODOM_TOPIC, Odometry, position_proxy.forward_position)
# Set up getting directions
direction_proxy = DirectionProxy()
rospy.Subscriber(LOCATION_TOPIC, String, direction_proxy.set_location)
WORLD_KNOWLEDGE.direction_proxy = direction_proxy
print "NLMaster startup complete!"
print "*" * 80
# Finally, wait for input before quitting
try:
while True:
text = raw_input("").strip()
if text == "q":
break
except (KeyboardInterrupt, EOFError):
pass
except:
raise
finally:
# Only shutdown the pipeline if we actually were taking language input
if not aut_path:
print "Shutting down NL pipeline..."
close_pipes()
self.comm_proxy.shutdown()
self.ipad_conn.shutdown()
sys.exit()
def run(self, aut_path=None):
"""Intialize all NL components."""
# pylint: disable=W0603
global WORLD_KNOWLEDGE
# Start the NL pipeline
if not aut_path:
print "Starting NL pipeline..."
init_pipes()
else:
print "Skipping loading nlpipeline because an automaton was loaded"
# Start the ROS node
print "Initializing ROS node..."
rospy.init_node('nlproxy')
# Set up the state mgr and its callbacks
print "Starting state manager..."
self.state_mgr = StateManager(self)
self.state_mgr.set_basedir(LTLGEN_BASE_DIR)
# Load the automaton if needed
if aut_path:
self.state_mgr.load_test_automaton(aut_path, False)
# Create world knowledge
print "Starting knowledge..."
WORLD_KNOWLEDGE = Knowledge(self)
self.state_mgr.world_knowledge = WORLD_KNOWLEDGE
# Wait a little for ROS to avoid timing startup issues.
print "Waiting for ROS node to settle..."
time.sleep(1)
# Set up ROS listening
print "Subscribing to notifications..."
# Set up state manager's sending and listening
pub = add_ltl_publisher()
self.state_mgr.set_publisher(pub)
add_subscriber(LTL_ENVIRONMENT_TOPIC, self.state_mgr.process_sensor_data)
add_subscriber(LTL_ENVIRONMENT_TOPIC, WORLD_KNOWLEDGE.process_sensor_data)
# Create the input comm_proxy and iPad connections
print "Starting comm_proxy..."
self.comm_proxy = CallbackSocket(self.text_port)
self.comm_proxy.register_callback(self.process_text)
self.ipad_conn = iPadConnection(self.map_port)
self.ipad_conn.register_rename_callback(rename_entity)
self.ipad_conn.register_text_callback(self.process_text)
# TODO Add highlight callback
self.map_proxy = MapProxy(self.ipad_conn)
add_subscriber(LTL_ENVIRONMENT_TOPIC, self.ipad_conn.add_icons)
WORLD_KNOWLEDGE.map_proxy = self.map_proxy
# Set up odometry forwarding to the ipad
tf = TransformListener()
while not tf.frameExists("/map"):
rospy.logwarn("Not ready for transforms yet")
rospy.sleep(1.0)
position_proxy = RobotPositionProxy(self.ipad_conn,tf)
rospy.Subscriber(ODOM_TOPIC, Odometry, position_proxy.forward_position)
# Set up getting directions
direction_proxy = DirectionProxy()
rospy.Subscriber(LOCATION_TOPIC, String, direction_proxy.set_location)
WORLD_KNOWLEDGE.direction_proxy = direction_proxy
print "NLMaster startup complete!"
print "*" * 80
# Finally, wait for input before quitting
try:
while True:
text = raw_input("").strip()
if text == "q":
break
except (KeyboardInterrupt, EOFError):
pass
except:
raise
finally:
# Only shutdown the pipeline if we actually were taking language input
if not aut_path:
print "Shutting down NL pipeline..."
close_pipes()
self.comm_proxy.shutdown()
self.ipad_conn.shutdown()
sys.exit()
class SpecGenerator(object):
"""Enables specification generation using natural language."""
def __init__(self):
# Start knowledge and connect to the NLP pipeline
self.sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
try:
self.sock.connect(('localhost', DEFAULT_PORT))
except socket.error:
raise IOError("Could not connect to pipelinehost on port %d. "
"Make sure that pipelinehost is running." % DEFAULT_PORT)
self.world_knowledge = Knowledge()
# Sets of propositions, accessible to all methods
self.custom_props = set()
def generate(self, text, sensors, regions, props):
"""Generate a logical specification from natural language and propositions."""
# Clean unicode out of everything
text = text.encode('ascii', 'ignore')
sensors = [astr.encode('ascii', 'ignore') for astr in sensors]
regions = [astr.encode('ascii', 'ignore') for astr in regions]
props = [astr.encode('ascii', 'ignore') for astr in props]
print "NL->LTL Generation called on:"
print "Sensors:", sensors
print "Props:", props
print "Regions:", regions
print "Text:", repr(text)
# Make lists for POS conversions, including the metapar keywords
force_nouns = list(regions) + list(sensors)
force_verbs = list(props) + METAPARS.keys()
responses = []
system_lines = []
env_lines = []
env_lines_set = set() # Used to track duplicates
custom_props = set()
generation_trees = []
for line in text.split('\n'):
if not line:
# Blank lines are counted as being processed correctly but are skipped
responses.append(True)
continue
# Init the generation tree to the empty result
generation_tree = [line.strip(), []]
# Lowercase and strip the text before using it
line = line.strip().lower()
print "Sending to remote parser:", repr(line)
parse = socket_parse(asocket=self.sock, text=line, force_nouns=force_nouns,
force_verbs=force_verbs)
print parse
user_response, semantics_result, semantics_response, new_commands = \
self.world_knowledge.process_parse_tree(parse, line)
# Build the metapars
failure = False
for command in new_commands:
try:
new_sys_lines, new_env_lines, new_custom_props = _apply_metapar(command)
except KeyError:
failure = True
continue
system_lines.extend(new_sys_lines)
# We need to ensure no duplicates are inserted into env_lines, so we keep
# an redundant set. If we were requiring 2.7 we would use OrderedDict.
for env_line in new_env_lines:
if env_line not in env_lines_set:
env_lines.append(env_line)
env_lines_set.add(env_line)
custom_props.update(new_custom_props)
# Add the statements as the children of the generation tree
generation_tree[1].append([str(command), [new_env_lines, new_sys_lines]])
# Add a true response if there were commands and no failures
responses.append(new_commands and not failure)
# Add this line's stuff to the generation tree
generation_trees.append(generation_tree)
# Convert custom props into a list
custom_props = list(custom_props)
print "Spec generation complete."
print "Responses:", responses
print "Environment lines:", env_lines
print "System lines:", system_lines
print "Custom props:", custom_props
print "Generation tree:", generation_trees
return env_lines, system_lines, custom_props, responses, generation_trees
