diff --git a/overcooked_simulator/game_content/environment_config_rl.yaml b/overcooked_simulator/game_content/environment_config_rl.yaml
index 391481efe7e4190b561520e0bd8c1bca137c4ed5..1b32b457a53a361932ac4bcd419f9ccf29fbfe8e 100644
--- a/overcooked_simulator/game_content/environment_config_rl.yaml
+++ b/overcooked_simulator/game_content/environment_config_rl.yaml
@@ -6,7 +6,7 @@ plates:
# range of seconds until the dirty plate arrives.
game:
- time_limit_seconds: 300
+ time_limit_seconds: 400
meals:
all: true
@@ -67,18 +67,8 @@ orders:
b: 20
sample_on_serving: false
# Sample the delay for the next order only after a meal was served.
- score_calc_gen_func: !!python/name:overcooked_simulator.order.simple_score_calc_gen_func ''
- score_calc_gen_kwargs:
- # the kwargs for the score_calc_gen_func
- other: 0
- scores: [ ]
- expired_penalty_func: !!python/name:overcooked_simulator.order.simple_expired_penalty ''
- expired_penalty_kwargs:
- default: 0
- serving_not_ordered_meals: !!python/name:overcooked_simulator.order.serving_not_ordered_meals_with_five_score ''
- # a func that calcs a store for not ordered but served meals. Input: meal
- penalty_for_trash: !!python/name:overcooked_simulator.order.penalty_for_each_item ''
- # a func that calcs the penalty for items that the player puts into the trashcan.
+ serving_not_ordered_meals: true
+ # can meals that are not ordered be served / dropped on the serving window
player_config:
radius: 0.4
@@ -103,20 +93,15 @@ extra_setup_functions:
hooks: [ completed_order ]
callback_class: !!python/name:overcooked_simulator.scores.ScoreViaHooks ''
callback_class_kwargs:
- static_score: 20
- score_on_specific_kwarg: meal_name
- score_map:
- Burger: 15
- OnionSoup: 10
- Salad: 5
- TomatoSoup: 10
- not_ordered_meals:
+ static_score: 100
+
+ serve_not_ordered_meals:
func: !!python/name:overcooked_simulator.hooks.hooks_via_callback_class ''
kwargs:
hooks: [ serve_not_ordered_meal ]
callback_class: !!python/name:overcooked_simulator.scores.ScoreViaHooks ''
callback_class_kwargs:
- static_score: 2
+ static_score: 100
trashcan_usages:
func: !!python/name:overcooked_simulator.hooks.hooks_via_callback_class ''
kwargs:
@@ -124,13 +109,21 @@ extra_setup_functions:
callback_class: !!python/name:overcooked_simulator.scores.ScoreViaHooks ''
callback_class_kwargs:
static_score: -5
- expired_orders:
+ item_cut:
+ func: !!python/name:overcooked_simulator.hooks.hooks_via_callback_class ''
+ kwargs:
+ hooks: [ cutting_board_100 ]
+ callback_class: !!python/name:overcooked_simulator.scores.ScoreViaHooks ''
+ callback_class_kwargs:
+ static_score: 10
+ stepped:
func: !!python/name:overcooked_simulator.hooks.hooks_via_callback_class ''
kwargs:
- hooks: [ order_expired ]
+ hooks: [ post_step ]
callback_class: !!python/name:overcooked_simulator.scores.ScoreViaHooks ''
callback_class_kwargs:
- static_score: -10
+ static_score: -1
+
# json_states:
# func: !!python/name:overcooked_simulator.recording.class_recording_with_hooks ''
# kwargs:
diff --git a/overcooked_simulator/game_content/layouts/rl.layout b/overcooked_simulator/game_content/layouts/rl.layout
index 624af90b4e174a801bfd027e0c9c864450489eef..4b91262e0e78525486821ca6c18edd99097138d8 100644
--- a/overcooked_simulator/game_content/layouts/rl.layout
+++ b/overcooked_simulator/game_content/layouts/rl.layout
@@ -1,5 +1,4 @@
-#X###
-T___#
-#___#
-U___P
-##W##
+#X##
+T__#
+U__P
+##W#
diff --git a/overcooked_simulator/gym_env.py b/overcooked_simulator/gym_env.py
index 2e3f55d7c8cf504d108824beda2e02ac18313310..c5de18a97c01e7d5b21999c12ff7652f31545a05 100644
--- a/overcooked_simulator/gym_env.py
+++ b/overcooked_simulator/gym_env.py
@@ -196,17 +196,8 @@ class EnvGymWrapper(Env):
observation = self.get_vector_state()
- reward = -1
- if (
- self.env.score > self.prev_score
- and self.env.score != 0
- ):
- self.prev_score = self.env.score
- reward = 100
- elif self.env.score < self.prev_score:
- self.prev_score = 0
- reward = -1
-
+ reward = self.env.score - self.prev_score
+ self.prev_score = self.env.score
terminated = self.env.game_ended
truncated = self.env.game_ended
info = {}
@@ -283,7 +274,7 @@ def main():
# # save_code=True, # optional
# )
- env = make_vec_env(EnvGymWrapper, n_envs=64)
+ env = make_vec_env(EnvGymWrapper, n_envs=8)
# env = EnvGymWrapper()
model_classes = [A2C, DQN, PPO]
@@ -317,7 +308,7 @@ def main():
check_env(env)
obs, info = env.reset()
while True:
- time.sleep(1 / 30)
+ time.sleep(1 / 10)
action, _states = model.predict(obs, deterministic=False)
obs, reward, terminated, truncated, info = env.step(int(action))
env.render()
diff --git a/overcooked_simulator/overcooked_environment.py b/overcooked_simulator/overcooked_environment.py
index 601d876994d7fc4494ed1a93dd13a4dac8122986..54de3fb157e31f41d397185c379f8b93213f3203 100644
--- a/overcooked_simulator/overcooked_environment.py
+++ b/overcooked_simulator/overcooked_environment.py
@@ -48,7 +48,7 @@ from overcooked_simulator.hooks import (
ACTION_ON_NOT_REACHABLE_COUNTER,
ACTION_PUT,
ACTION_INTERACT_START,
- ITEM_INFO_CONFIG,
+ ITEM_INFO_CONFIG, POST_STEP,
)
from overcooked_simulator.order import (
OrderManager,
@@ -756,7 +756,7 @@ class Environment:
self.order_manager.progress(passed_time=passed_time, now=self.env_time)
for effect_manager in self.effect_manager.values():
effect_manager.progress(passed_time=passed_time, now=self.env_time)
- # self.hook(POST_STEP, passed_time=passed_time)
+ self.hook(POST_STEP, passed_time=passed_time)
def get_state(self):
"""Get the current state of the game environment. The state here is accessible by the current python objects.
diff --git a/overcooked_simulator/rl_agent_checkpoints/overcooked_PPO.zip b/overcooked_simulator/rl_agent_checkpoints/overcooked_PPO.zip
new file mode 100644
index 0000000000000000000000000000000000000000..b79af5605fa8773501888d028bca62b6845cb2d4
Binary files /dev/null and b/overcooked_simulator/rl_agent_checkpoints/overcooked_PPO.zip differ