mirror of
https://github.com/illiumst/marl-factory-grid.git
synced 2025-09-15 23:37:14 +02:00
Redone the spawn procedute and destination objects
This commit is contained in:
0
marl_factory_grid/modules/aomas/__init__.py
Normal file
0
marl_factory_grid/modules/aomas/__init__.py
Normal file
28
marl_factory_grid/modules/aomas/narrow_corridor/rules.py
Normal file
28
marl_factory_grid/modules/aomas/narrow_corridor/rules.py
Normal file
@@ -0,0 +1,28 @@
|
||||
from random import shuffle
|
||||
from typing import List, Tuple
|
||||
|
||||
from marl_factory_grid.environment.rules import Rule
|
||||
from marl_factory_grid.environment import constants as c
|
||||
from marl_factory_grid.modules.destinations import constants as d
|
||||
from marl_factory_grid.modules.destinations.entitites import BoundDestination
|
||||
|
||||
|
||||
class NarrowCorridorSpawn(Rule):
|
||||
def __init__(self, positions: List[Tuple[int, int]], fixed: bool = False):
|
||||
super().__init__()
|
||||
self.fixed = fixed
|
||||
self.positions = positions
|
||||
|
||||
def on_init(self, state, lvl_map):
|
||||
if not self.fixed:
|
||||
shuffle(self.positions)
|
||||
for agent in state[c.AGENT]:
|
||||
pass
|
||||
|
||||
def trigger_destination_spawn(self, state):
|
||||
for (agent_name, position_list) in self.per_agent_positions.items():
|
||||
agent = state[c.AGENT][agent_name]
|
||||
destinations = [BoundDestination(agent, pos) for pos in position_list]
|
||||
state[d.DESTINATION].add_items(destinations)
|
||||
return c.VALID
|
||||
|
@@ -70,7 +70,7 @@ class PodRules(Rule):
|
||||
|
||||
def on_init(self, state, lvl_map):
|
||||
pod_collection = state[b.CHARGE_PODS]
|
||||
empty_tiles = state[c.FLOOR].empty_tiles[:self.n_pods]
|
||||
empty_tiles = state[c.FLOORS].empty_tiles[:self.n_pods]
|
||||
pods = pod_collection.from_tiles(empty_tiles, entity_kwargs=dict(
|
||||
multi_charge=self.multi_charge, charge_rate=self.charge_rate)
|
||||
)
|
||||
|
@@ -47,7 +47,7 @@ class DirtPiles(PositionMixin, EnvObjects):
|
||||
return c.VALID
|
||||
|
||||
def trigger_dirt_spawn(self, state, initial_spawn=False) -> bool:
|
||||
free_for_dirt = [x for x in state[c.FLOOR]
|
||||
free_for_dirt = [x for x in state[c.FLOORS]
|
||||
if len(x.guests) == 0 or (
|
||||
len(x.guests) == 1 and
|
||||
isinstance(next(y for y in x.guests), DirtPile))
|
||||
|
@@ -1,4 +1,4 @@
|
||||
from .actions import DestAction
|
||||
from .entitites import Destination
|
||||
from .groups import ReachedDestinations, Destinations
|
||||
from .rules import DestinationDone, DestinationReach, DestinationSpawn
|
||||
from .groups import Destinations, BoundDestinations
|
||||
from .rules import DestinationReachAll, DestinationSpawn
|
||||
|
@@ -3,8 +3,6 @@
|
||||
DESTINATION = 'Destinations'
|
||||
BOUNDDESTINATION = 'BoundDestinations'
|
||||
DEST_SYMBOL = 1
|
||||
DEST_REACHED_REWARD = 0.5
|
||||
DEST_REACHED = 'ReachedDestinations'
|
||||
|
||||
WAIT_ON_DEST = 'WAIT'
|
||||
|
||||
|
@@ -16,42 +16,31 @@ class Destination(Entity):
|
||||
var_is_blocking_pos = False
|
||||
var_is_blocking_light = False
|
||||
|
||||
@property
|
||||
def any_agent_has_dwelled(self):
|
||||
return bool(len(self._per_agent_times))
|
||||
|
||||
@property
|
||||
def currently_dwelling_names(self):
|
||||
return list(self._per_agent_times.keys())
|
||||
|
||||
@property
|
||||
def encoding(self):
|
||||
return d.DEST_SYMBOL
|
||||
|
||||
def __init__(self, *args, dwell_time: int = 0, **kwargs):
|
||||
def __init__(self, *args, action_counts=0, **kwargs):
|
||||
super(Destination, self).__init__(*args, **kwargs)
|
||||
self.dwell_time = dwell_time
|
||||
self._per_agent_times = defaultdict(lambda: dwell_time)
|
||||
self.action_counts = action_counts
|
||||
self._per_agent_actions = defaultdict(lambda: 0)
|
||||
|
||||
def do_wait_action(self, agent: Agent):
|
||||
self._per_agent_times[agent.name] -= 1
|
||||
self._per_agent_actions[agent.name] += 1
|
||||
return c.VALID
|
||||
|
||||
def leave(self, agent: Agent):
|
||||
del self._per_agent_times[agent.name]
|
||||
|
||||
@property
|
||||
def is_considered_reached(self):
|
||||
agent_at_position = any(c.AGENT.lower() in x.name.lower() for x in self.tile.guests_that_can_collide)
|
||||
return (agent_at_position and not self.dwell_time) or any(x == 0 for x in self._per_agent_times.values())
|
||||
return agent_at_position or any(x >= self.action_counts for x in self._per_agent_actions.values())
|
||||
|
||||
def agent_is_dwelling(self, agent: Agent):
|
||||
return self._per_agent_times[agent.name] < self.dwell_time
|
||||
def agent_did_action(self, agent: Agent):
|
||||
return self._per_agent_actions[agent.name] >= self.action_counts
|
||||
|
||||
def summarize_state(self) -> dict:
|
||||
state_summary = super().summarize_state()
|
||||
state_summary.update(per_agent_times=[
|
||||
dict(belongs_to=key, time=val) for key, val in self._per_agent_times.items()], dwell_time=self.dwell_time)
|
||||
dict(belongs_to=key, time=val) for key, val in self._per_agent_actions.items()], counts=self.action_counts)
|
||||
return state_summary
|
||||
|
||||
def render(self):
|
||||
@@ -68,9 +57,8 @@ class BoundDestination(BoundEntityMixin, Destination):
|
||||
self.bind_to(entity)
|
||||
super().__init__(*args, **kwargs)
|
||||
|
||||
|
||||
@property
|
||||
def is_considered_reached(self):
|
||||
agent_at_position = any(self.bound_entity == x for x in self.tile.guests_that_can_collide)
|
||||
return (agent_at_position and not self.dwell_time) \
|
||||
or any(x == 0 for x in self._per_agent_times[self.bound_entity.name])
|
||||
return ((agent_at_position and not self.action_counts)
|
||||
or self._per_agent_actions[self.bound_entity.name] >= self.action_counts >= 1)
|
||||
|
@@ -23,14 +23,3 @@ class BoundDestinations(HasBoundMixin, Destinations):
|
||||
def __init__(self, *args, **kwargs):
|
||||
super().__init__(*args, **kwargs)
|
||||
|
||||
|
||||
class ReachedDestinations(Destinations):
|
||||
_entity = Destination
|
||||
is_blocking_light = False
|
||||
can_collide = False
|
||||
|
||||
def __init__(self, *args, **kwargs):
|
||||
super(ReachedDestinations, self).__init__(*args, **kwargs)
|
||||
|
||||
def __repr__(self):
|
||||
return super(ReachedDestinations, self).__repr__()
|
||||
|
@@ -1,84 +1,61 @@
|
||||
from typing import List, Union
|
||||
import ast
|
||||
from random import shuffle
|
||||
from typing import List, Union, Dict, Tuple
|
||||
from marl_factory_grid.environment.rules import Rule
|
||||
from marl_factory_grid.utils.results import TickResult, DoneResult
|
||||
from marl_factory_grid.environment import constants as c
|
||||
|
||||
from marl_factory_grid.modules.destinations import constants as d, rewards as r
|
||||
from marl_factory_grid.modules.destinations.entitites import Destination
|
||||
from marl_factory_grid.modules.destinations.entitites import Destination, BoundDestination
|
||||
|
||||
|
||||
class DestinationReach(Rule):
|
||||
class DestinationReachAll(Rule):
|
||||
|
||||
def __init__(self, n_dests: int = 1, tiles: Union[List, None] = None):
|
||||
super(DestinationReach, self).__init__()
|
||||
self.n_dests = n_dests or len(tiles)
|
||||
self._tiles = tiles
|
||||
def __init__(self):
|
||||
super(DestinationReachAll, self).__init__()
|
||||
|
||||
def tick_step(self, state) -> List[TickResult]:
|
||||
|
||||
for dest in list(state[d.DESTINATION].values()):
|
||||
results = []
|
||||
for dest in list(state[next(key for key in state.entities.names if d.DESTINATION in key)]):
|
||||
if dest.is_considered_reached:
|
||||
dest.change_parent_collection(state[d.DEST_REACHED])
|
||||
agent = state[c.AGENT].by_pos(dest.pos)
|
||||
results.append(TickResult(self.name, validity=c.VALID, reward=r.DEST_REACHED, entity=agent))
|
||||
state.print(f'{dest.name} is reached now, removing...')
|
||||
assert dest.destroy(), f'{dest.name} could not be destroyed. Critical Error.'
|
||||
else:
|
||||
for agent_name in dest.currently_dwelling_names:
|
||||
agent = state[c.AGENT][agent_name]
|
||||
if agent.pos == dest.pos:
|
||||
state.print(f'{agent.name} is still waiting.')
|
||||
pass
|
||||
else:
|
||||
dest.leave(agent)
|
||||
state.print(f'{agent.name} left the destination early.')
|
||||
pass
|
||||
return [TickResult(self.name, validity=c.VALID, reward=0, entity=None)]
|
||||
|
||||
def tick_post_step(self, state) -> List[TickResult]:
|
||||
results = list()
|
||||
for reached_dest in state[d.DEST_REACHED]:
|
||||
for guest in reached_dest.tile.guests:
|
||||
if guest in state[c.AGENT]:
|
||||
state.print(f'{guest.name} just reached destination at {guest.pos}')
|
||||
state[d.DEST_REACHED].delete_env_object(reached_dest)
|
||||
results.append(TickResult(self.name, validity=c.VALID, reward=r.DEST_REACHED, entity=guest))
|
||||
return results
|
||||
|
||||
|
||||
class DestinationDone(Rule):
|
||||
|
||||
def __init__(self):
|
||||
super(DestinationDone, self).__init__()
|
||||
|
||||
def on_check_done(self, state) -> List[DoneResult]:
|
||||
if not len(state[d.DESTINATION]):
|
||||
return [DoneResult(self.name, validity=c.VALID, reward=r.DEST_REACHED)]
|
||||
return []
|
||||
|
||||
|
||||
class DoneOnReach(Rule):
|
||||
|
||||
def __init__(self):
|
||||
super(DoneOnReach, self).__init__()
|
||||
|
||||
def on_check_done(self, state) -> List[DoneResult]:
|
||||
dests = [x.pos for x in state[d.DESTINATION]]
|
||||
agents = [x.pos for x in state[c.AGENT]]
|
||||
|
||||
if any([x in dests for x in agents]):
|
||||
if not len(state[next(key for key in state.entities.names if d.DESTINATION in key)]):
|
||||
return [DoneResult(self.name, validity=c.VALID, reward=r.DEST_REACHED)]
|
||||
return [DoneResult(self.name, validity=c.NOT_VALID, reward=0)]
|
||||
|
||||
|
||||
class DestinationReachAny(DestinationReachAll):
|
||||
|
||||
def __init__(self):
|
||||
super(DestinationReachAny, self).__init__()
|
||||
|
||||
def on_check_done(self, state) -> List[DoneResult]:
|
||||
if not len(state[next(key for key in state.entities.names if d.DESTINATION in key)]):
|
||||
return [DoneResult(self.name, validity=c.VALID, reward=r.DEST_REACHED)]
|
||||
return []
|
||||
|
||||
|
||||
class DestinationSpawn(Rule):
|
||||
|
||||
def __init__(self, spawn_frequency: int = 5, n_dests: int = 1,
|
||||
def __init__(self, n_dests: int = 1,
|
||||
spawn_mode: str = d.MODE_GROUPED):
|
||||
super(DestinationSpawn, self).__init__()
|
||||
self.spawn_frequency = spawn_frequency
|
||||
self.n_dests = n_dests
|
||||
self.spawn_mode = spawn_mode
|
||||
|
||||
def on_init(self, state, lvl_map):
|
||||
# noinspection PyAttributeOutsideInit
|
||||
self._dest_spawn_timer = self.spawn_frequency
|
||||
self.trigger_destination_spawn(self.n_dests, state)
|
||||
pass
|
||||
|
||||
@@ -88,16 +65,40 @@ class DestinationSpawn(Rule):
|
||||
def tick_step(self, state) -> List[TickResult]:
|
||||
if n_dest_spawn := max(0, self.n_dests - len(state[d.DESTINATION])):
|
||||
if self.spawn_mode == d.MODE_GROUPED and n_dest_spawn == self.n_dests:
|
||||
validity = state.rules['DestinationReach'].trigger_destination_spawn(n_dest_spawn, state)
|
||||
validity = self.trigger_destination_spawn(n_dest_spawn, state)
|
||||
return [TickResult(self.name, validity=validity, entity=None, value=n_dest_spawn)]
|
||||
elif self.spawn_mode == d.MODE_SINGLE and n_dest_spawn:
|
||||
validity = self.trigger_destination_spawn(n_dest_spawn, state)
|
||||
return [TickResult(self.name, validity=validity, entity=None, value=n_dest_spawn)]
|
||||
else:
|
||||
pass
|
||||
|
||||
@staticmethod
|
||||
def trigger_destination_spawn(n_dests, state, tiles=None):
|
||||
tiles = tiles or state[c.FLOOR].empty_tiles[:n_dests]
|
||||
if destinations := [Destination(tile) for tile in tiles]:
|
||||
def trigger_destination_spawn(self, n_dests, state):
|
||||
empty_positions = state[c.FLOORS].empty_tiles[:n_dests]
|
||||
if destinations := [Destination(pos) for pos in empty_positions]:
|
||||
state[d.DESTINATION].add_items(destinations)
|
||||
state.print(f'{n_dests} new destinations have been spawned')
|
||||
return c.VALID
|
||||
else:
|
||||
state.print('No Destiantions are spawning, limit is reached.')
|
||||
return c.NOT_VALID
|
||||
|
||||
|
||||
class FixedDestinationSpawn(Rule):
|
||||
def __init__(self, per_agent_positions: Dict[str, List[Tuple[int, int]]]):
|
||||
super(Rule, self).__init__()
|
||||
self.per_agent_positions = {key: [ast.literal_eval(x) for x in val] for key, val in per_agent_positions.items()}
|
||||
|
||||
def on_init(self, state, lvl_map):
|
||||
for (agent_name, position_list) in self.per_agent_positions.items():
|
||||
agent = next(x for x in state[c.AGENT] if agent_name in x.name) # Fixme: Ugly AF
|
||||
shuffle(position_list)
|
||||
while True:
|
||||
pos = position_list.pop()
|
||||
if pos != agent.pos and not state[d.BOUNDDESTINATION].by_pos(pos):
|
||||
destination = BoundDestination(agent, state[c.FLOORS].by_pos(pos))
|
||||
break
|
||||
else:
|
||||
continue
|
||||
state[d.BOUNDDESTINATION].add_item(destination)
|
||||
pass
|
||||
|
@@ -21,7 +21,7 @@ class AgentSingleZonePlacementBeta(Rule):
|
||||
coordinates = random.choices(self.coordinates, k=len(agents))
|
||||
else:
|
||||
raise ValueError
|
||||
tiles = [state[c.FLOOR].by_pos(pos) for pos in coordinates]
|
||||
tiles = [state[c.FLOORS].by_pos(pos) for pos in coordinates]
|
||||
for agent, tile in zip(agents, tiles):
|
||||
agent.move(tile)
|
||||
|
||||
|
@@ -41,7 +41,7 @@ class ItemRules(Rule):
|
||||
|
||||
def trigger_item_spawn(self, state):
|
||||
if item_to_spawns := max(0, (self.n_items - len(state[i.ITEM]))):
|
||||
empty_tiles = state[c.FLOOR].empty_tiles[:item_to_spawns]
|
||||
empty_tiles = state[c.FLOORS].empty_tiles[:item_to_spawns]
|
||||
state[i.ITEM].spawn(empty_tiles)
|
||||
self._next_item_spawn = self.spawn_frequency
|
||||
state.print(f'{item_to_spawns} new items have been spawned; next spawn in {self._next_item_spawn}')
|
||||
@@ -73,7 +73,7 @@ class ItemRules(Rule):
|
||||
return []
|
||||
|
||||
def trigger_drop_off_location_spawn(self, state):
|
||||
empty_tiles = state[c.FLOOR].empty_tiles[:self.n_locations]
|
||||
empty_tiles = state[c.FLOORS].empty_tiles[:self.n_locations]
|
||||
do_entites = state[i.DROP_OFF]
|
||||
drop_offs = [DropOffLocation(tile) for tile in empty_tiles]
|
||||
do_entites.add_items(drop_offs)
|
||||
|
5
marl_factory_grid/modules/levels/narrow_corridor.txt
Normal file
5
marl_factory_grid/modules/levels/narrow_corridor.txt
Normal file
@@ -0,0 +1,5 @@
|
||||
#######
|
||||
###-###
|
||||
#1---2#
|
||||
###-###
|
||||
#######
|
@@ -13,7 +13,7 @@ class MachineRule(Rule):
|
||||
self.n_machines = n_machines
|
||||
|
||||
def on_init(self, state, lvl_map):
|
||||
empty_tiles = state[c.FLOOR].empty_tiles[:self.n_machines]
|
||||
empty_tiles = state[c.FLOORS].empty_tiles[:self.n_machines]
|
||||
state[m.MACHINES].add_items(Machine(tile) for tile in empty_tiles)
|
||||
|
||||
def tick_pre_step(self, state) -> List[TickResult]:
|
||||
|
@@ -39,7 +39,7 @@ class Maintainer(Entity):
|
||||
self._next = []
|
||||
self._last = []
|
||||
self._last_serviced = 'None'
|
||||
self._floortile_graph = points_to_graph(state[c.FLOOR].positions)
|
||||
self._floortile_graph = points_to_graph(state[c.FLOORS].positions)
|
||||
|
||||
def tick(self, state):
|
||||
if found_objective := state[self.objective].by_pos(self.pos):
|
||||
@@ -89,7 +89,7 @@ class Maintainer(Entity):
|
||||
|
||||
def _predict_move(self, state):
|
||||
next_pos = self._path[0]
|
||||
if len(state[c.FLOOR].by_pos(next_pos).guests_that_can_collide) > 0:
|
||||
if len(state[c.FLOORS].by_pos(next_pos).guests_that_can_collide) > 0:
|
||||
action = c.NOOP
|
||||
else:
|
||||
next_pos = self._path.pop(0)
|
||||
|
@@ -5,11 +5,9 @@ from marl_factory_grid.environment.entity.wall_floor import Floor
|
||||
from marl_factory_grid.environment.groups.env_objects import EnvObjects
|
||||
from marl_factory_grid.environment.groups.mixins import PositionMixin
|
||||
from ..machines.actions import MachineAction
|
||||
from ...utils.render import RenderEntity
|
||||
from ...utils.states import Gamestate
|
||||
|
||||
from ..machines import constants as mc
|
||||
from . import constants as mi
|
||||
|
||||
|
||||
class Maintainers(PositionMixin, EnvObjects):
|
||||
|
@@ -14,7 +14,7 @@ class MaintenanceRule(Rule):
|
||||
self.n_maintainer = n_maintainer
|
||||
|
||||
def on_init(self, state: Gamestate, lvl_map):
|
||||
state[M.MAINTAINERS].spawn(state[c.FLOOR].empty_tiles[:self.n_maintainer], state)
|
||||
state[M.MAINTAINERS].spawn(state[c.FLOORS].empty_tiles[:self.n_maintainer], state)
|
||||
pass
|
||||
|
||||
def tick_pre_step(self, state) -> List[TickResult]:
|
||||
|
@@ -19,7 +19,7 @@ class ZoneInit(Rule):
|
||||
while z_idx:
|
||||
zone_positions = lvl_map.get_coordinates_for_symbol(z_idx)
|
||||
if len(zone_positions):
|
||||
zones.append(Zone([state[c.FLOOR].by_pos(pos) for pos in zone_positions]))
|
||||
zones.append(Zone([state[c.FLOORS].by_pos(pos) for pos in zone_positions]))
|
||||
z_idx += 1
|
||||
else:
|
||||
z_idx = 0
|
||||
|
Reference in New Issue
Block a user