diff --git a/.idea/.gitignore b/.idea/.gitignore
index e7e9d11d4bf243bffe4bb60b4ac1f9392297f4bf..d6fd6f21b4f8d95e972cf75f62bebcdc4537a139 100644
--- a/.idea/.gitignore
+++ b/.idea/.gitignore
@@ -1,2 +1,2 @@
-# Default ignored files
-/workspace.xml
+# Default ignored files
+/workspace.xml
diff --git a/.idea/misc.xml b/.idea/misc.xml
index 65531ca992813bbfedbe43dfae5a5f4337168ed8..5417b684e7619b0b44d2f9d8be364ac8fb576783 100644
--- a/.idea/misc.xml
+++ b/.idea/misc.xml
@@ -1,4 +1,4 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<project version="4">
- <component name="ProjectRootManager" version="2" project-jdk-name="Python 3.6" project-jdk-type="Python SDK" />
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+ <component name="ProjectRootManager" version="2" project-jdk-name="Python 3.6" project-jdk-type="Python SDK" />
</project>
\ No newline at end of file
diff --git a/.idea/modules.xml b/.idea/modules.xml
index 925278cb163e3cc6c725cda433b8df8b625c3f0b..c9b6e2db97b91b3f14bb3798b10f26f334334bb6 100644
--- a/.idea/modules.xml
+++ b/.idea/modules.xml
@@ -1,8 +1,8 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<project version="4">
- <component name="ProjectModuleManager">
- <modules>
- <module fileurl="file://$PROJECT_DIR$/.idea/neurips2020-flatland-starter-kit.iml" filepath="$PROJECT_DIR$/.idea/neurips2020-flatland-starter-kit.iml" />
- </modules>
- </component>
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+ <component name="ProjectModuleManager">
+ <modules>
+ <module fileurl="file://$PROJECT_DIR$/.idea/neurips2020-flatland-starter-kit.iml" filepath="$PROJECT_DIR$/.idea/neurips2020-flatland-starter-kit.iml" />
+ </modules>
+ </component>
</project>
\ No newline at end of file
diff --git a/.idea/neurips2020-flatland-starter-kit.iml b/.idea/neurips2020-flatland-starter-kit.iml
index 8dc09e5476bcb840206461450ae44f23421d964a..951c9286734f053453803b04b5335bb575715344 100644
--- a/.idea/neurips2020-flatland-starter-kit.iml
+++ b/.idea/neurips2020-flatland-starter-kit.iml
@@ -1,11 +1,11 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<module type="PYTHON_MODULE" version="4">
- <component name="NewModuleRootManager">
- <content url="file://$MODULE_DIR$" />
- <orderEntry type="inheritedJdk" />
- <orderEntry type="sourceFolder" forTests="false" />
- </component>
- <component name="TestRunnerService">
- <option name="PROJECT_TEST_RUNNER" value="pytest" />
- </component>
+<?xml version="1.0" encoding="UTF-8"?>
+<module type="PYTHON_MODULE" version="4">
+ <component name="NewModuleRootManager">
+ <content url="file://$MODULE_DIR$" />
+ <orderEntry type="inheritedJdk" />
+ <orderEntry type="sourceFolder" forTests="false" />
+ </component>
+ <component name="TestRunnerService">
+ <option name="PROJECT_TEST_RUNNER" value="pytest" />
+ </component>
</module>
\ No newline at end of file
diff --git a/.idea/vcs.xml b/.idea/vcs.xml
index 94a25f7f4cb416c083d265558da75d457237d671..9661ac713428efbad557d3ba3a62216b5bb7d226 100644
--- a/.idea/vcs.xml
+++ b/.idea/vcs.xml
@@ -1,6 +1,6 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<project version="4">
- <component name="VcsDirectoryMappings">
- <mapping directory="$PROJECT_DIR$" vcs="Git" />
- </component>
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+ <component name="VcsDirectoryMappings">
+ <mapping directory="$PROJECT_DIR$" vcs="Git" />
+ </component>
</project>
\ No newline at end of file
diff --git a/README.md b/README.md
index 22846ae8c532fa537e79ca807fe71860100f1f23..8232bb1e569f62e96028faf3bc31adc12c9c201e 100644
--- a/README.md
+++ b/README.md
@@ -7,9 +7,28 @@
+
+# Round 1 - 3rd best RL solution
+
+## Used agent
+* [PPO Agent -> Mitchell Goff](https://github.com/mitchellgoffpc/flatland-training)
+
+## LICENCE for the Observation EXTRA.py
+
+The observation can be used freely and reused for further submissions. Only the author needs to be referred to
+/mentioned in any submissions - if the entire observation or parts, or the main idea is used.
+
+Author: Adrian Egli (adrian.egli@gmail.com)
+
+[Linkedin](https://www.researchgate.net/profile/Adrian_Egli2)
+[Researchgate](https://www.linkedin.com/in/adrian-egli-733a9544/)
+
+
+
+
Main links
---
-
+* [Submit in 10 minutes](https://flatland.aicrowd.com/getting-started/first-submission.html?_ga=2.175036450.1456714032.1596434204-43124944.1552486604)
* [Flatland documentation](https://flatland.aicrowd.com/)
* [NeurIPS 2020 Challenge](https://www.aicrowd.com/challenges/neurips-2020-flatland-challenge/)
diff --git a/checkpoints/ppo/model_checkpoint.meta b/checkpoints/ppo/model_checkpoint.meta
index e998bcda155cbd11e6f1b70e77966ed92812930c..31959a4680ed59613594bfb9418b1e3497ad6ce1 100644
Binary files a/checkpoints/ppo/model_checkpoint.meta and b/checkpoints/ppo/model_checkpoint.meta differ
diff --git a/checkpoints/ppo/model_checkpoint.optimizer b/checkpoints/ppo/model_checkpoint.optimizer
index 0fbd49e2e1fa62c34dff663c9d47bd53f61128f3..89630787199743db98bebfd8f3132c681a60a099 100644
Binary files a/checkpoints/ppo/model_checkpoint.optimizer and b/checkpoints/ppo/model_checkpoint.optimizer differ
diff --git a/checkpoints/ppo/model_checkpoint.policy b/checkpoints/ppo/model_checkpoint.policy
index 4868691cf2f12669df395b845b2a903c1d917336..6049b699289690113a6eed1bca3545cd12bf4c4e 100644
Binary files a/checkpoints/ppo/model_checkpoint.policy and b/checkpoints/ppo/model_checkpoint.policy differ
diff --git a/dump.rdb b/dump.rdb
index ba7e97e6ff27c20f75b47463e01777453f724a57..bcb60c2ec208cac6ff4ea41cc5bd2d73a8e3e945 100644
Binary files a/dump.rdb and b/dump.rdb differ
diff --git a/run.py b/run.py
index 27c3107896fde363315d78eadd778ffe2ac99f7e..08f32e4b1c10916b44ed0a5363ea3d8470ffa5f0 100644
--- a/run.py
+++ b/run.py
@@ -30,7 +30,7 @@ def my_controller(extra: Extra, observation, info):
# the example here :
# https://gitlab.aicrowd.com/flatland/flatland/blob/master/flatland/envs/observations.py#L14
#####################################################################
-my_observation_builder = Extra(max_depth=2)
+my_observation_builder = Extra(max_depth=20)
# Or if you want to use your own approach to build the observation from the env_step,
# please feel free to pass a DummyObservationBuilder() object as mentioned below,
@@ -112,6 +112,8 @@ while True:
print("w : ", extra.env.width)
print("h : ", extra.env.height)
+ old_total_done = 0
+ old_total_active = 0
while True:
#####################################################################
# Evaluation of a single episode
@@ -141,7 +143,11 @@ while True:
x = (local_env.agents[a].status in [RailAgentStatus.DONE, RailAgentStatus.DONE_REMOVED])
total_done += int(x)
total_active += int(local_env.agents[a].status == RailAgentStatus.ACTIVE)
- # print("total_done:", total_done, "\ttotal_active", total_active, "\t num agents", local_env.get_num_agents())
+ if old_total_done != total_done or old_total_active != total_active:
+ print("total_done:", total_done, "\ttotal_active", total_active, "\t num agents",
+ local_env.get_num_agents())
+ old_total_done = total_done
+ old_total_active = total_active
if done['__all__']:
print("Reward : ", sum(list(all_rewards.values())))
diff --git a/src/extra.py b/src/extra.py
index e5a75c8d6eb82997aff52f7b14263e78edc39693..0d948bd4796888fdbcbe361d14379f5db31b3b41 100644
--- a/src/extra.py
+++ b/src/extra.py
@@ -19,7 +19,7 @@
#
# Private submission
# http://gitlab.aicrowd.com/adrian_egli/neurips2020-flatland-starter-kit/issues/8
-
+#
import numpy as np
from flatland.core.env_observation_builder import ObservationBuilder
from flatland.core.grid.grid4_utils import get_new_position
@@ -66,9 +66,7 @@ def fast_count_nonzero(possible_transitions: (int, int, int, int)):
class Extra(ObservationBuilder):
def __init__(self, max_depth):
- self.max_depth = max_depthmodel_checkpoint.meta
-model_checkpoint.optimizer
-model_checkpoint.policy
+ self.max_depth = max_depth
self.observation_dim = 22
self.agent = None
@@ -190,6 +188,9 @@ model_checkpoint.policy
return obsData
def is_collision(self, obsData):
+ return False
+
+ def intern_is_collision(self, obsData):
if np.sum(obsData[10:14]) == 0:
return False
if np.sum(obsData[10:14]) == np.sum(obsData[14:18]):
@@ -244,15 +245,20 @@ model_checkpoint.policy
return has_opp_agent, has_same_agent, visited
if agents_on_switch:
+ pt = 0
for dir_loop in range(4):
if possible_transitions[dir_loop] == 1:
+ pt += 1
hoa, hsa, v = self._explore(handle,
get_new_position(new_position, dir_loop),
dir_loop,
depth + 1)
visited.append(v)
- has_opp_agent = 0.5 * (has_opp_agent + hoa)
- has_same_agent = 0.5 * (has_same_agent + hsa)
+ has_opp_agent += hoa
+ has_same_agent + hsa
+ if pt > 0:
+ has_opp_agent /= pt
+ has_same_agent /= pt
return has_opp_agent, has_same_agent, visited
else:
new_direction = fast_argmax(possible_transitions)
@@ -267,7 +273,7 @@ model_checkpoint.policy
# observation[3] : 1 path towards target (direction 3) / otherwise 0 -> path is longer or there is no path
# observation[4] : int(agent.status == RailAgentStatus.READY_TO_DEPART)
# observation[5] : int(agent.status == RailAgentStatus.ACTIVE)
- # observation[6] : int(agent.status == RailAgentStatus.DONE or agent.status == RailAgentStatus.DONE_REMOVED)
+ # observation[6] : deadlock estimated (collision) 1 otherwise 0
# observation[7] : current agent is located at a switch, where it can take a routing decision
# observation[8] : current agent is located at a cell, where it has to take a stop-or-go decision
# observation[9] : current agent is located one step before/after a switch
@@ -284,7 +290,7 @@ model_checkpoint.policy
# observation[20] : If there is a path with step (direction 2) and there is a agent with same direction -> 1
# observation[21] : If there is a path with step (direction 3) and there is a agent with same direction -> 1
- observation = np.zeros(self.observation_dim)
+ observation = np.zeros(self.observation_dim) - 1
visited = []
agent = self.env.agents[handle]
@@ -296,7 +302,6 @@ model_checkpoint.policy
agent_virtual_position = agent.position
observation[5] = 1
else:
- observation[6] = 1
agent_virtual_position = (-1, -1)
agent_done = True
@@ -335,6 +340,8 @@ model_checkpoint.policy
observation[8] = int(agents_near_to_switch)
observation[9] = int(agents_near_to_switch_all)
+ observation[6] = int(self.intern_is_collision(observation))
+
self.env.dev_obs_dict.update({handle: visited})
return observation
@@ -351,6 +358,26 @@ model_checkpoint.policy
return action_dict
+ def rl_agent_act_X(self, observation, info, eps=0.0):
+ self.loadAgent()
+ action_dict = {}
+ active_cnt = 0
+ for a in range(self.env.get_num_agents()):
+ if active_cnt < 1 or self.env.agents[a].status == RailAgentStatus.ACTIVE:
+ if observation[a][6] == 0:
+ active_cnt += int(self.env.agents[a].status < RailAgentStatus.DONE)
+ if info['action_required'][a]:
+ action_dict[a] = self.agent.act(observation[a], eps=eps)
+ # action_dict[a] = np.random.randint(5)
+ else:
+ action_dict[a] = RailEnvActions.MOVE_FORWARD
+ else:
+ action_dict[a] = RailEnvActions.STOP_MOVING
+ else:
+ action_dict[a] = RailEnvActions.STOP_MOVING
+
+ return action_dict
+
def loadAgent(self):
if self.agent is not None:
return
@@ -359,4 +386,4 @@ model_checkpoint.policy
print("action_size: ", self.action_size)
print("state_size: ", self.state_size)
self.agent = Agent(self.state_size, self.action_size, 0)
- self.agent.load('./checkpoints/', 0, 1.0)
\ No newline at end of file
+ self.agent.load('./checkpoints/', 0, 1.0)