diff --git a/env_data/tests/Level_distance_map_shortest_path.pkl b/env_data/tests/Level_distance_map_shortest_path.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..938adf4660784976299a4be63c9d91667a27c924
Binary files /dev/null and b/env_data/tests/Level_distance_map_shortest_path.pkl differ
diff --git a/flatland/__init__.py b/flatland/__init__.py
index f6411ca099ebc8f211e3d639e8a873831e43496f..74f8368d8a19dd7eb1ad36a501c81805ef975133 100644
--- a/flatland/__init__.py
+++ b/flatland/__init__.py
@@ -4,4 +4,4 @@
__author__ = """S.P. Mohanty"""
__email__ = 'mohanty@aicrowd.com'
-__version__ = '2.1.6'
+__version__ = '2.1.7'
diff --git a/flatland/envs/distance_map.py b/flatland/envs/distance_map.py
index 2bc1a5117794959cca82d2edad821cb629397f78..fe546d1fdd368597f0a6263d61323ecfeb6dc31f 100644
--- a/flatland/envs/distance_map.py
+++ b/flatland/envs/distance_map.py
@@ -60,8 +60,16 @@ class DistanceMap:
self.env_height,
self.env_width,
4))
+
+ computed_targets = []
for i, agent in enumerate(agents):
- self._distance_map_walker(rail, agent.target, i)
+ if agent.target not in computed_targets:
+ self._distance_map_walker(rail, agent.target, i)
+ else:
+ # just copy the distance map form other agent with same target (performance)
+ self.distance_map[i, :, :, :] = np.copy(
+ self.distance_map[computed_targets.index(agent.target), :, :, :])
+ computed_targets.append(agent.target)
def _distance_map_walker(self, rail: GridTransitionMap, position, target_nr: int):
"""
diff --git a/flatland/envs/rail_env.py b/flatland/envs/rail_env.py
index 10908dddbc376f9493d5e00dcda2d48fd0d25163..991de23fe42d19c77d6003c403474609e27b44f9 100644
--- a/flatland/envs/rail_env.py
+++ b/flatland/envs/rail_env.py
@@ -115,6 +115,7 @@ class RailEnv(Environment):
schedule_generator: ScheduleGenerator = random_schedule_generator(),
number_of_agents=1,
obs_builder_object: ObservationBuilder = GlobalObsForRailEnv(),
+ max_episode_steps=None,
stochastic_data=None,
remove_agents_at_target=False,
random_seed=1
@@ -147,6 +148,7 @@ class RailEnv(Environment):
obs_builder_object: ObservationBuilder object
ObservationBuilder-derived object that takes builds observation
vectors for each agent.
+ max_episode_steps : int or None
remove_agents_at_target : bool
If remove_agents_at_target is set to true then the agents will be removed by placing to
RailEnv.DEPOT_POSITION when the agent has reach it's target position.
@@ -169,7 +171,7 @@ class RailEnv(Environment):
self.obs_builder = obs_builder_object
self.obs_builder.set_env(self)
- self._max_episode_steps = None
+ self._max_episode_steps = max_episode_steps
self._elapsed_steps = 0
self.dones = dict.fromkeys(list(range(number_of_agents)) + ["__all__"], False)
@@ -215,6 +217,9 @@ class RailEnv(Environment):
self.max_number_of_steps_broken = malfunction_max_duration
# Reset environment
+ self.reset()
+ self.num_resets = 0 # yes, set it to zero again!
+
self.valid_positions = None
def _seed(self, seed=None):
@@ -265,6 +270,11 @@ class RailEnv(Environment):
self.rail = rail
self.height, self.width = self.rail.grid.shape
+ # Do a new set_env call on the obs_builder to ensure
+ # that obs_builder specific instantiations are made according to the
+ # specifications of the current environment : like width, height, etc
+ self.obs_builder.set_env(self)
+
if optionals and 'distance_map' in optionals:
self.distance_map.set(optionals['distance_map'])
@@ -275,9 +285,8 @@ class RailEnv(Environment):
# TODO https://gitlab.aicrowd.com/flatland/flatland/issues/185
# why do we need static agents? could we it more elegantly?
- schedule = self.schedule_generator(self.rail, self.get_num_agents(), agents_hints, self.num_resets)
- self.agents_static = EnvAgentStatic.from_lists(schedule)
- self._max_episode_steps = schedule.max_episode_steps
+ self.agents_static = EnvAgentStatic.from_lists(
+ *self.schedule_generator(self.rail, self.get_num_agents(), agents_hints, self.num_resets))
self.restart_agents()
diff --git a/flatland/envs/rail_env_shortest_paths.py b/flatland/envs/rail_env_shortest_paths.py
index 7944a49daf388403c8a226dbf866ac810d1286a3..622fb6efd37ab07f59d575fa618cb34676fefcd1 100644
--- a/flatland/envs/rail_env_shortest_paths.py
+++ b/flatland/envs/rail_env_shortest_paths.py
@@ -72,7 +72,7 @@ def get_valid_move_actions_(agent_direction: Grid4TransitionsEnum,
# N.B. get_shortest_paths is not part of distance_map since it refers to RailEnvActions (would lead to circularity!)
-def get_shortest_paths(distance_map: DistanceMap, max_depth: Optional[int] = None) \
+def get_shortest_paths(distance_map: DistanceMap, max_depth: Optional[int] = None, agent_handle: Optional[int] = None) \
-> Dict[int, Optional[List[WalkingElement]]]:
"""
Computes the shortest path for each agent to its target and the action to be taken to do so.
@@ -81,9 +81,15 @@ def get_shortest_paths(distance_map: DistanceMap, max_depth: Optional[int] = Non
If there is no path (rail disconnected), the path is given as None.
The agent state (moving or not) and its speed are not taken into account
+ example:
+ agent_fixed_travel_paths = get_shortest_paths(env.distance_map, None, agent.handle)
+ path = agent_fixed_travel_paths[agent.handle]
+
Parameters
----------
- distance_map
+ distance_map : reference to the distance_map
+ max_depth : max path length, if the shortest path is longer, it will be cutted
+ agent_handle : if set, the shortest for agent.handle will be returned , otherwise for all agents
Returns
-------
@@ -133,8 +139,11 @@ def get_shortest_paths(distance_map: DistanceMap, max_depth: Optional[int] = Non
WalkingElement(position, direction,
RailEnvNextAction(RailEnvActions.STOP_MOVING, position, direction)))
- for agent in distance_map.agents:
- _shortest_path_for_agent(agent)
+ if agent_handle is not None:
+ _shortest_path_for_agent(distance_map.agents[agent_handle])
+ else:
+ for agent in distance_map.agents:
+ _shortest_path_for_agent(agent)
return shortest_paths
diff --git a/flatland/evaluators/client.py b/flatland/evaluators/client.py
index e6ea1c500336b7592ba45480571ebbb9743d0bfe..9322b6a8530cd3c77c246a2e74e99a8e2bc91ab9 100644
--- a/flatland/evaluators/client.py
+++ b/flatland/evaluators/client.py
@@ -25,9 +25,9 @@ m.patch()
def are_dicts_equal(d1, d2):
""" return True if all keys and values are the same """
- return all(k in d2 and d1[k] == d2[k]
+ return all(k in d2 and np.isclose(d1[k], d2[k])
for k in d1) \
- and all(k in d1 and d1[k] == d2[k]
+ and all(k in d1 and np.isclose(d1[k], d2[k])
for k in d2)
@@ -80,6 +80,7 @@ class FlatlandRemoteClient(object):
'AICROWD_TESTS_FOLDER',
'/tmp/flatland_envs'
)
+ self.current_env_path = None
self.verbose = verbose
@@ -180,7 +181,9 @@ class FlatlandRemoteClient(object):
return observation, info
test_env_file_path = _response['payload']['env_file_path']
- print("Received Env : ", test_env_file_path)
+ if self.verbose:
+ print("Received Env : ", test_env_file_path)
+
test_env_file_path = os.path.join(
self.test_envs_root,
test_env_file_path
@@ -192,7 +195,10 @@ class FlatlandRemoteClient(object):
"to point to the location of the Tests folder ? \n"
"We are currently looking at `{}` for the tests".format(self.test_envs_root)
)
- print("Current env path : ", test_env_file_path)
+
+ if self.verbose:
+ print("Current env path : ", test_env_file_path)
+ self.current_env_path = test_env_file_path
self.env = RailEnv(
width=1,
height=1,
@@ -232,11 +238,13 @@ class FlatlandRemoteClient(object):
local_observation, local_reward, local_done, local_info = \
self.env.step(action)
- print(local_reward)
+ if self.verbose:
+ print(local_reward)
if not are_dicts_equal(remote_reward, local_reward):
+ print("Remote Reward : ", remote_reward, "Local Reward : ", local_reward)
raise Exception("local and remote `reward` are diverging")
- print(remote_reward, local_reward)
if not are_dicts_equal(remote_done, local_done):
+ print("Remote Done : ", remote_done, "Local Done : ", local_done)
raise Exception("local and remote `done` are diverging")
# Return local_observation instead of remote_observation
diff --git a/setup.cfg b/setup.cfg
index 0d5be947cbc3138fa48a9f6fdd4d7f86295dad2b..69077d856c45cf863dc0ff976c6b5ed32bf93330 100644
--- a/setup.cfg
+++ b/setup.cfg
@@ -1,5 +1,5 @@
[bumpversion]
-current_version = 2.1.6
+current_version = 2.1.7
commit = True
tag = True
diff --git a/setup.py b/setup.py
index b7dad95638a1d498016906ebfcd934e3c4a1b8e0..0b6cf7c5f0f04065ecab0d712c74b050f7069385 100644
--- a/setup.py
+++ b/setup.py
@@ -80,6 +80,6 @@ setup(
test_suite='tests',
tests_require=test_requirements,
url='https://gitlab.aicrowd.com/flatland/flatland',
- version='2.1.6',
+ version='2.1.7',
zip_safe=False,
)
diff --git a/tests/test_flatland_envs_rail_env_shortest_paths.py b/tests/test_flatland_envs_rail_env_shortest_paths.py
index ef18cf72d04a5e6ccfd89ad379bd30df77fc4075..0d863c281840839b2e80aa5e8fe08400c49ad873 100644
--- a/tests/test_flatland_envs_rail_env_shortest_paths.py
+++ b/tests/test_flatland_envs_rail_env_shortest_paths.py
@@ -1,3 +1,5 @@
+import sys
+
import numpy as np
from flatland.core.grid.grid4 import Grid4TransitionsEnum
@@ -196,3 +198,165 @@ def test_get_shortest_paths_max_depth():
for agent_handle in expected:
assert np.array_equal(actual[agent_handle], expected[agent_handle]), \
"[{}] actual={},expected={}".format(agent_handle, actual[agent_handle], expected[agent_handle])
+
+
+def test_get_shortest_paths_agent_handle():
+ env = load_flatland_environment_from_file('Level_distance_map_shortest_path.pkl', 'env_data.tests')
+ actual = get_shortest_paths(env.distance_map, agent_handle=6)
+
+ print(actual, file=sys.stderr)
+
+ expected = {6:
+ [WalkingElement(position=(5, 5),
+ direction=0,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(4, 5), next_direction=0)),
+ WalkingElement(position=(4, 5),
+ direction=0,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(3, 5), next_direction=0)),
+ WalkingElement(position=(3, 5),
+ direction=0,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(2, 5), next_direction=0)),
+ WalkingElement(position=(2, 5),
+ direction=0,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(1, 5), next_direction=0)),
+ WalkingElement(position=(1, 5),
+ direction=0,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(0, 5), next_direction=0)),
+ WalkingElement(position=(0, 5),
+ direction=0,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(0, 6), next_direction=1)),
+ WalkingElement(position=(0, 6),
+ direction=1,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(0, 7), next_direction=1)),
+ WalkingElement(position=(0, 7),
+ direction=1,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(0, 8), next_direction=1)),
+ WalkingElement(position=(0, 8),
+ direction=1,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(0, 9), next_direction=1)),
+ WalkingElement(position=(0, 9),
+ direction=1,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(0, 10), next_direction=1)),
+ WalkingElement(position=(0, 10),
+ direction=1,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(1, 10), next_direction=2)),
+ WalkingElement(position=(1, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(2, 10), next_direction=2)),
+ WalkingElement(position=(2, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(3, 10), next_direction=2)),
+ WalkingElement(position=(3, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(4, 10), next_direction=2)),
+ WalkingElement(position=(4, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(5, 10), next_direction=2)),
+ WalkingElement(position=(5, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(6, 10), next_direction=2)),
+ WalkingElement(position=(6, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(7, 10), next_direction=2)),
+ WalkingElement(position=(7, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(8, 10), next_direction=2)),
+ WalkingElement(position=(8, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(9, 10), next_direction=2)),
+ WalkingElement(position=(9, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(10, 10), next_direction=2)),
+ WalkingElement(position=(10, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(11, 10), next_direction=2)),
+ WalkingElement(position=(11, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(12, 10), next_direction=2)),
+ WalkingElement(position=(12, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(13, 10), next_direction=2)),
+ WalkingElement(position=(13, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(14, 10), next_direction=2)),
+ WalkingElement(position=(14, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(15, 10), next_direction=2)),
+ WalkingElement(position=(15, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(16, 10), next_direction=2)),
+ WalkingElement(position=(16, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(17, 10), next_direction=2)),
+ WalkingElement(position=(17, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(18, 10), next_direction=2)),
+ WalkingElement(position=(18, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(19, 10), next_direction=2)),
+ WalkingElement(position=(19, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(20, 10), next_direction=2)),
+ WalkingElement(position=(20, 10),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(20, 9), next_direction=3)),
+ WalkingElement(position=(20, 9),
+ direction=3,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(20, 8), next_direction=3)),
+ WalkingElement(position=(20, 8),
+ direction=3, next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_LEFT,
+ next_position=(21, 8),
+ next_direction=2)),
+ WalkingElement(position=(21, 8),
+ direction=2,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(21, 7), next_direction=3)),
+ WalkingElement(position=(21, 7),
+ direction=3,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(21, 6), next_direction=3)),
+ WalkingElement(position=(21, 6),
+ direction=3,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.MOVE_FORWARD,
+ next_position=(21, 5), next_direction=3)),
+ WalkingElement(position=(21, 5),
+ direction=3,
+ next_action_element=RailEnvNextAction(action=RailEnvActions.STOP_MOVING,
+ next_position=(21, 5), next_direction=3))
+ ]}
+
+ for agent_handle in expected:
+ assert np.array_equal(actual[agent_handle], expected[agent_handle]), \
+ "[{}] actual={},expected={}".format(agent_handle, actual[agent_handle], expected[agent_handle])