mirror of
https://github.com/illiumst/marl-factory-grid.git
synced 2025-09-15 23:37:14 +02:00
fixed names and function, dirtspawn and item spawn
This commit is contained in:
@@ -60,7 +60,7 @@ Just define what your environment needs in a *yaml*-configfile like:
|
|||||||
done_at_collisions: !!bool True
|
done_at_collisions: !!bool True
|
||||||
ItemRespawn:
|
ItemRespawn:
|
||||||
spawn_freq: 5
|
spawn_freq: 5
|
||||||
DoDoorAutoClose: {}
|
DoorAutoClose: {}
|
||||||
|
|
||||||
Assets:
|
Assets:
|
||||||
- Defaults
|
- Defaults
|
||||||
|
@@ -1,10 +1,4 @@
|
|||||||
Agents:
|
Agents:
|
||||||
Eberhart:
|
|
||||||
Actions:
|
|
||||||
- Move8
|
|
||||||
- Noop
|
|
||||||
- ItemAction
|
|
||||||
Observations:
|
|
||||||
Wolfgang:
|
Wolfgang:
|
||||||
Actions:
|
Actions:
|
||||||
- Noop
|
- Noop
|
||||||
@@ -29,7 +23,9 @@ Agents:
|
|||||||
- DropOffLocations
|
- DropOffLocations
|
||||||
- Maintainers
|
- Maintainers
|
||||||
Entities:
|
Entities:
|
||||||
Batteries: {}
|
Batteries:
|
||||||
|
initial_charge: 0.8
|
||||||
|
per_action_costs: 0.02
|
||||||
ChargePods: {}
|
ChargePods: {}
|
||||||
Destinations: {}
|
Destinations: {}
|
||||||
DirtPiles:
|
DirtPiles:
|
||||||
@@ -56,25 +52,21 @@ General:
|
|||||||
verbose: false
|
verbose: false
|
||||||
|
|
||||||
Rules:
|
Rules:
|
||||||
Btry:
|
SpawnAgents: {}
|
||||||
initial_charge: 0.8
|
DoneAtBatteryDischarge: {}
|
||||||
per_action_costs: 0.02
|
|
||||||
BtryDoneAtDischarge: {}
|
|
||||||
Collision:
|
Collision:
|
||||||
done_at_collisions: false
|
done_at_collisions: false
|
||||||
AssignGlobalPositions: {}
|
AssignGlobalPositions: {}
|
||||||
DestinationReachAny: {}
|
DoneAtDestinationReachAny: {}
|
||||||
DestinationReach:
|
DestinationReachReward: {}
|
||||||
|
SpawnDestinations:
|
||||||
n_dests: 1
|
n_dests: 1
|
||||||
DestinationSpawn:
|
|
||||||
n_dests: 1
|
|
||||||
spawn_frequency: 5
|
|
||||||
spawn_mode: GROUPED
|
spawn_mode: GROUPED
|
||||||
DirtAllCleanDone: {}
|
DoneOnAllDirtCleaned: {}
|
||||||
DirtRespawnRule:
|
SpawnDirt:
|
||||||
spawn_freq: 15
|
spawn_freq: 15
|
||||||
DirtSmearOnMove:
|
EntitiesSmearDirtOnMove:
|
||||||
smear_amount: 0.2
|
smear_ratio: 0.2
|
||||||
DoorAutoClose:
|
DoorAutoClose:
|
||||||
close_frequency: 10
|
close_frequency: 10
|
||||||
ItemRules:
|
ItemRules:
|
||||||
|
@@ -12,7 +12,7 @@ class BoundEntityMixin:
|
|||||||
if self.bound_entity:
|
if self.bound_entity:
|
||||||
return f'{self.__class__.__name__}({self.bound_entity.name})'
|
return f'{self.__class__.__name__}({self.bound_entity.name})'
|
||||||
else:
|
else:
|
||||||
print()
|
pass
|
||||||
|
|
||||||
def belongs_to_entity(self, entity):
|
def belongs_to_entity(self, entity):
|
||||||
return entity == self.bound_entity
|
return entity == self.bound_entity
|
||||||
|
@@ -131,6 +131,7 @@ class EnvObject(Object):
|
|||||||
return c.VALUE_OCCUPIED_CELL
|
return c.VALUE_OCCUPIED_CELL
|
||||||
|
|
||||||
def __init__(self, **kwargs):
|
def __init__(self, **kwargs):
|
||||||
|
self._bound_entity = None
|
||||||
super(EnvObject, self).__init__(**kwargs)
|
super(EnvObject, self).__init__(**kwargs)
|
||||||
|
|
||||||
def change_parent_collection(self, other_collection):
|
def change_parent_collection(self, other_collection):
|
||||||
|
@@ -1,6 +1,6 @@
|
|||||||
from collections import defaultdict
|
from collections import defaultdict
|
||||||
from operator import itemgetter
|
from operator import itemgetter
|
||||||
from random import shuffle
|
from random import shuffle, random
|
||||||
from typing import Dict
|
from typing import Dict
|
||||||
|
|
||||||
from marl_factory_grid.environment.groups.objects import Objects
|
from marl_factory_grid.environment.groups.objects import Objects
|
||||||
@@ -26,6 +26,7 @@ class Entities(Objects):
|
|||||||
|
|
||||||
@property
|
@property
|
||||||
def floorlist(self):
|
def floorlist(self):
|
||||||
|
shuffle(self._floor_positions)
|
||||||
return self._floor_positions
|
return self._floor_positions
|
||||||
|
|
||||||
def __init__(self, floor_positions):
|
def __init__(self, floor_positions):
|
||||||
|
@@ -1,4 +1,4 @@
|
|||||||
from .actions import BtryCharge
|
from .actions import BtryCharge
|
||||||
from .entitites import Pod, Battery
|
from .entitites import Pod, Battery
|
||||||
from .groups import ChargePods, Batteries
|
from .groups import ChargePods, Batteries
|
||||||
from .rules import BtryDoneAtDischarge, BatteryDecharge
|
from .rules import DoneAtBatteryDischarge, BatteryDecharge
|
||||||
|
@@ -94,7 +94,7 @@ class BatteryDecharge(Rule):
|
|||||||
return results
|
return results
|
||||||
|
|
||||||
|
|
||||||
class BtryDoneAtDischarge(BatteryDecharge):
|
class DoneAtBatteryDischarge(BatteryDecharge):
|
||||||
|
|
||||||
def __init__(self, reward_discharge_done=b.REWARD_DISCHARGE_DONE, mode: str = b.SINGLE, **kwargs):
|
def __init__(self, reward_discharge_done=b.REWARD_DISCHARGE_DONE, mode: str = b.SINGLE, **kwargs):
|
||||||
f"""
|
f"""
|
||||||
|
@@ -30,6 +30,7 @@ class DirtPiles(PositionMixin, EnvObjects):
|
|||||||
if not self.amount > self.max_global_amount:
|
if not self.amount > self.max_global_amount:
|
||||||
amount = amount_s[idx] if isinstance(amount_s, list) else amount_s
|
amount = amount_s[idx] if isinstance(amount_s, list) else amount_s
|
||||||
if dirt := self.by_pos(pos):
|
if dirt := self.by_pos(pos):
|
||||||
|
dirt = next(dirt.iter())
|
||||||
new_value = dirt.amount + amount
|
new_value = dirt.amount + amount
|
||||||
dirt.set_new_amount(new_value)
|
dirt.set_new_amount(new_value)
|
||||||
else:
|
else:
|
||||||
@@ -42,8 +43,8 @@ class DirtPiles(PositionMixin, EnvObjects):
|
|||||||
return Result(identifier=f'{self.name}_spawn', validity=c.VALID, reward=0, value=spawn_counter)
|
return Result(identifier=f'{self.name}_spawn', validity=c.VALID, reward=0, value=spawn_counter)
|
||||||
|
|
||||||
def trigger_dirt_spawn(self, n, amount, state, n_var=0.2, amount_var=0.2) -> Result:
|
def trigger_dirt_spawn(self, n, amount, state, n_var=0.2, amount_var=0.2) -> Result:
|
||||||
free_for_dirt = [x for x in state.entities.floorlist if len(state.entities.pos_dict[x]) == 1 or (
|
free_for_dirt = [x for x in state.entities.floorlist if len(state.entities.pos_dict[x]) == 0 or (
|
||||||
len(state.entities.pos_dict[x]) == 2 and isinstance(next(y for y in x), DirtPile))]
|
len(state.entities.pos_dict[x]) >= 1 and isinstance(next(y for y in x), DirtPile))]
|
||||||
# free_for_dirt = [x for x in state[c.FLOOR]
|
# free_for_dirt = [x for x in state[c.FLOOR]
|
||||||
# if len(x.guests) == 0 or (
|
# if len(x.guests) == 0 or (
|
||||||
# len(x.guests) == 1 and
|
# len(x.guests) == 1 and
|
||||||
|
@@ -27,7 +27,8 @@ class DoneOnAllDirtCleaned(Rule):
|
|||||||
|
|
||||||
class SpawnDirt(Rule):
|
class SpawnDirt(Rule):
|
||||||
|
|
||||||
def __init__(self, initial_n: int, initial_amount: float, respawn_n: int, respawn_amount: float,
|
def __init__(self, initial_n: int = 5, initial_amount: float = 1.3,
|
||||||
|
respawn_n: int = 3, respawn_amount: float = 0.8,
|
||||||
n_var: float = 0.2, amount_var: float = 0.2, spawn_freq: int = 15):
|
n_var: float = 0.2, amount_var: float = 0.2, spawn_freq: int = 15):
|
||||||
"""
|
"""
|
||||||
Defines the spawn pattern of intial and additional 'Dirt'-entitites.
|
Defines the spawn pattern of intial and additional 'Dirt'-entitites.
|
||||||
|
@@ -13,7 +13,7 @@ from marl_factory_grid.modules.destinations.entitites import Destination
|
|||||||
|
|
||||||
class DestinationReachReward(Rule):
|
class DestinationReachReward(Rule):
|
||||||
|
|
||||||
def __init__(self, dest_reach_reward=marl_factory_grid.modules.destinations.constants.REWARD_DEST_REACHED):
|
def __init__(self, dest_reach_reward=d.REWARD_DEST_REACHED):
|
||||||
"""
|
"""
|
||||||
This rule introduces the basic functionality, so that targts (Destinations) can be reached and marked as such.
|
This rule introduces the basic functionality, so that targts (Destinations) can be reached and marked as such.
|
||||||
Additionally, rewards are reported.
|
Additionally, rewards are reported.
|
||||||
@@ -50,7 +50,7 @@ class DestinationReachReward(Rule):
|
|||||||
|
|
||||||
class DoneAtDestinationReachAll(DestinationReachReward):
|
class DoneAtDestinationReachAll(DestinationReachReward):
|
||||||
|
|
||||||
def __init__(self, reward_at_done=marl_factory_grid.modules.destinations.constants.REWARD_DEST_DONE, **kwargs):
|
def __init__(self, reward_at_done=d.REWARD_DEST_DONE, **kwargs):
|
||||||
"""
|
"""
|
||||||
This rule triggers and sets the done flag if ALL Destinations have been reached.
|
This rule triggers and sets the done flag if ALL Destinations have been reached.
|
||||||
|
|
||||||
|
@@ -1,4 +1,4 @@
|
|||||||
from .actions import DoorUse
|
from .actions import DoorUse
|
||||||
from .entitites import Door, DoorIndicator
|
from .entitites import Door, DoorIndicator
|
||||||
from .groups import Doors
|
from .groups import Doors
|
||||||
from .rules import DoDoorAutoClose, IndicateDoorAreaInObservation
|
from .rules import DoorAutoClose, IndicateDoorAreaInObservation
|
||||||
|
@@ -5,7 +5,7 @@ from . import constants as d
|
|||||||
from .entitites import DoorIndicator
|
from .entitites import DoorIndicator
|
||||||
|
|
||||||
|
|
||||||
class DoDoorAutoClose(Rule):
|
class DoorAutoClose(Rule):
|
||||||
|
|
||||||
def __init__(self, close_frequency: int = 10):
|
def __init__(self, close_frequency: int = 10):
|
||||||
"""
|
"""
|
||||||
|
@@ -104,7 +104,7 @@ class DropOffLocations(PositionMixin, EnvObjects):
|
|||||||
|
|
||||||
@staticmethod
|
@staticmethod
|
||||||
def trigger_drop_off_location_spawn(state, n_locations):
|
def trigger_drop_off_location_spawn(state, n_locations):
|
||||||
empty_positions = state.entities.empty_positions()[:n_locations]
|
empty_positions = state.entities.empty_positions[:n_locations]
|
||||||
do_entites = state[i.DROP_OFF]
|
do_entites = state[i.DROP_OFF]
|
||||||
drop_offs = [DropOffLocation(pos) for pos in empty_positions]
|
drop_offs = [DropOffLocation(pos) for pos in empty_positions]
|
||||||
do_entites.add_items(drop_offs)
|
do_entites.add_items(drop_offs)
|
||||||
|
@@ -8,6 +8,7 @@ import yaml
|
|||||||
|
|
||||||
from marl_factory_grid.environment.groups.agents import Agents
|
from marl_factory_grid.environment.groups.agents import Agents
|
||||||
from marl_factory_grid.environment.entity.agent import Agent
|
from marl_factory_grid.environment.entity.agent import Agent
|
||||||
|
from marl_factory_grid.environment.rules import Rule
|
||||||
from marl_factory_grid.utils.helpers import locate_and_import_class
|
from marl_factory_grid.utils.helpers import locate_and_import_class
|
||||||
from marl_factory_grid.environment import constants as c
|
from marl_factory_grid.environment import constants as c
|
||||||
|
|
||||||
@@ -81,7 +82,15 @@ class FactoryConfigParser(object):
|
|||||||
entity_class = locate_and_import_class(entity, folder_path)
|
entity_class = locate_and_import_class(entity, folder_path)
|
||||||
except AttributeError as e3:
|
except AttributeError as e3:
|
||||||
ents = [y for x in [e1.argss[1], e2.argss[1], e3.argss[1]] for y in x]
|
ents = [y for x in [e1.argss[1], e2.argss[1], e3.argss[1]] for y in x]
|
||||||
raise AttributeError(e1.argss[0], e2.argss[0], e3.argss[0], 'Possible Entitys are:', str(ents))
|
print('### Error ### Error ### Error ### Error ### Error ###')
|
||||||
|
print()
|
||||||
|
print(f'Class "{entity}" was not found in "{folder_path.name}"')
|
||||||
|
print('Possible Entitys are:', str(ents))
|
||||||
|
print()
|
||||||
|
print('Goodbye')
|
||||||
|
print()
|
||||||
|
exit()
|
||||||
|
# raise AttributeError(e1.argss[0], e2.argss[0], e3.argss[0], 'Possible Entitys are:', str(ents))
|
||||||
|
|
||||||
entity_kwargs = self.entities.get(entity, {})
|
entity_kwargs = self.entities.get(entity, {})
|
||||||
entity_symbol = entity_class.symbol if hasattr(entity_class, 'symbol') else None
|
entity_symbol = entity_class.symbol if hasattr(entity_class, 'symbol') else None
|
||||||
@@ -114,6 +123,7 @@ class FactoryConfigParser(object):
|
|||||||
|
|
||||||
# Observation
|
# Observation
|
||||||
observations = list()
|
observations = list()
|
||||||
|
assert self.agents[name]['Observations'] is not None, 'Did you specify any Observation?'
|
||||||
if c.DEFAULTS in self.agents[name]['Observations']:
|
if c.DEFAULTS in self.agents[name]['Observations']:
|
||||||
observations.extend(self.default_observations)
|
observations.extend(self.default_observations)
|
||||||
observations.extend(x for x in self.agents[name]['Observations'] if x != c.DEFAULTS)
|
observations.extend(x for x in self.agents[name]['Observations'] if x != c.DEFAULTS)
|
||||||
@@ -141,6 +151,8 @@ class FactoryConfigParser(object):
|
|||||||
rule_class = locate_and_import_class(rule, folder_path)
|
rule_class = locate_and_import_class(rule, folder_path)
|
||||||
except AttributeError:
|
except AttributeError:
|
||||||
rule_class = locate_and_import_class(rule, self.custom_modules_path)
|
rule_class = locate_and_import_class(rule, self.custom_modules_path)
|
||||||
|
# Fixme This check does not work!
|
||||||
|
# assert isinstance(rule_class, Rule), f'{rule_class.__name__} is no valid "Rule".'
|
||||||
rule_kwargs = self.rules.get(rule, {})
|
rule_kwargs = self.rules.get(rule, {})
|
||||||
rules_classes.update({rule: {'class': rule_class, 'kwargs': rule_kwargs}})
|
rules_classes.update({rule: {'class': rule_class, 'kwargs': rule_kwargs}})
|
||||||
return rules_classes
|
return rules_classes
|
||||||
|
@@ -148,18 +148,19 @@ class OBSBuilder(object):
|
|||||||
np.put(obs[idx], 0, v, mode='raise')
|
np.put(obs[idx], 0, v, mode='raise')
|
||||||
except IndexError:
|
except IndexError:
|
||||||
raise ValueError(f'Max(obs.size) for {e.name}: {obs[idx].size}, but was: {len(v)}.')
|
raise ValueError(f'Max(obs.size) for {e.name}: {obs[idx].size}, but was: {len(v)}.')
|
||||||
|
if self.pomdp_r:
|
||||||
try:
|
try:
|
||||||
light_map = np.zeros(self.obs_shape)
|
light_map = np.zeros(self.obs_shape)
|
||||||
visible_floor = set(self.ray_caster[agent.name].visible_entities(self._floortiles, reset_cache=False))
|
visible_floor = set(self.ray_caster[agent.name].visible_entities(self._floortiles, reset_cache=False))
|
||||||
if self.pomdp_r:
|
if self.pomdp_r:
|
||||||
coords = [((f.x - agent.x) + self.pomdp_r, (f.y - agent.y) + self.pomdp_r) for f in visible_floor]
|
# Fixme: This Sucks if the Map is too small!!
|
||||||
else:
|
coords = [((f.x - agent.x) + self.pomdp_r, (f.y - agent.y) + self.pomdp_r) for f in visible_floor]
|
||||||
coords = [x.pos for x in visible_floor]
|
else:
|
||||||
np.put(light_map, np.ravel_multi_index(np.asarray(coords).T, light_map.shape), 1)
|
coords = [x.pos for x in visible_floor]
|
||||||
self.curr_lightmaps[agent.name] = light_map
|
np.put(light_map, np.ravel_multi_index(np.asarray(coords).T, light_map.shape), 1)
|
||||||
except KeyError:
|
self.curr_lightmaps[agent.name] = light_map
|
||||||
print()
|
except (KeyError, ValueError):
|
||||||
|
pass
|
||||||
return obs, self.obs_layers[agent.name]
|
return obs, self.obs_layers[agent.name]
|
||||||
|
|
||||||
def _sort_and_name_observation_conf(self, agent):
|
def _sort_and_name_observation_conf(self, agent):
|
||||||
|
@@ -88,8 +88,8 @@ class Gamestate(object):
|
|||||||
results.extend(self.rules.tick_pre_step_all(self))
|
results.extend(self.rules.tick_pre_step_all(self))
|
||||||
|
|
||||||
for idx, action_int in enumerate(actions):
|
for idx, action_int in enumerate(actions):
|
||||||
|
agent = self[c.AGENT][idx].clear_temp_state()
|
||||||
if not agent.var_is_paralyzed:
|
if not agent.var_is_paralyzed:
|
||||||
agent = self[c.AGENT][idx].clear_temp_state()
|
|
||||||
action = agent.actions[action_int]
|
action = agent.actions[action_int]
|
||||||
action_result = action.do(agent, self)
|
action_result = action.do(agent, self)
|
||||||
results.append(action_result)
|
results.append(action_result)
|
||||||
|
Reference in New Issue
Block a user