Spaces:
Sleeping
Sleeping
File size: 6,586 Bytes
c71bf62 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 | """Public models for the Dispatch Arena environment."""
from __future__ import annotations
from enum import Enum
from typing import Any, Dict, List, Mapping, Optional
from pydantic import BaseModel, ConfigDict, Field, model_validator
class DispatchArenaModel(BaseModel):
"""Base model with JSON-friendly helpers used by server and client."""
model_config = ConfigDict(use_enum_values=False)
def to_dict(self) -> Dict[str, Any]:
return self.model_dump(mode="json", exclude_none=True)
class Mode(str, Enum):
MINI = "mini"
NORMAL = "normal"
class CourierStatus(str, Enum):
IDLE = "idle"
TO_PICKUP = "to_pickup"
WAITING_PICKUP = "waiting_pickup"
TO_DROPOFF = "to_dropoff"
REPOSITIONING = "repositioning"
class OrderStatus(str, Enum):
QUEUED = "queued"
READY = "ready"
PICKED = "picked"
DELIVERED = "delivered"
EXPIRED = "expired"
class VerifierVerdict(str, Enum):
IN_PROGRESS = "in_progress"
DELIVERED_SUCCESSFULLY = "delivered_successfully"
TIMEOUT_FAILURE = "timeout_failure"
PARTIAL_SUCCESS = "partial_success"
class MiniActionType(str, Enum):
WAIT = "wait"
GO_PICKUP = "go_pickup"
GO_DROPOFF = "go_dropoff"
PICKUP = "pickup"
DROPOFF = "dropoff"
class NormalActionType(str, Enum):
ASSIGN = "assign"
REPOSITION = "reposition"
HOLD = "hold"
PRIORITIZE = "prioritize"
class Action(DispatchArenaModel):
"""OpenEnv-facing action payload."""
action_type: str
courier_id: Optional[str] = None
order_id: Optional[str] = None
node_id: Optional[str] = None
@property
def name(self) -> str:
return self.action_type
@classmethod
def from_dict(cls, data: Mapping[str, Any]) -> "Action":
return cls.model_validate(dict(data))
class Node(DispatchArenaModel):
id: str
kind: str
label: str
class Courier(DispatchArenaModel):
id: str
node_id: str
status: CourierStatus = CourierStatus.IDLE
eta_remaining: int = 0
assigned_order_id: Optional[str] = None
load: Optional[str] = None
target_node_id: Optional[str] = None
@model_validator(mode="after")
def _validate_eta(self) -> "Courier":
if self.eta_remaining < 0:
raise ValueError("eta_remaining must be >= 0")
return self
class Order(DispatchArenaModel):
id: str
kind: str = "food"
pickup_node_id: str
dropoff_node_id: str
created_tick: int = 0
arrival_tick: int = 0
prep_remaining: Optional[int] = None
deadline_tick: int = 20
status: OrderStatus = OrderStatus.QUEUED
assigned_courier_id: Optional[str] = None
ready_now: Optional[bool] = None
delivered_tick: Optional[int] = None
class Config(DispatchArenaModel):
mode: Mode = Mode.MINI
max_ticks: int = 12
visible_prep: bool = False
num_couriers: int = 1
num_orders: int = 1
scenario_bucket: str = "easy"
progress_shaping: bool = True
rolling_arrivals: bool = False
traffic_noise: float = 0.0
@model_validator(mode="after")
def _validate_config(self) -> "Config":
if self.max_ticks <= 0:
raise ValueError("max_ticks must be > 0")
if self.traffic_noise < 0.0 or self.traffic_noise > 2.0:
raise ValueError("traffic_noise must be in [0.0, 2.0]")
if self.mode == Mode.MINI:
self.num_couriers = 1
self.num_orders = 1
if self.mode == Mode.NORMAL:
self.num_couriers = min(max(self.num_couriers, 2), 5)
self.num_orders = min(max(self.num_orders, 3), 10)
return self
class RewardBreakdown(DispatchArenaModel):
"""Machine-readable reward decomposition for a transition."""
step_cost: float = 0.0
progress_reward: float = 0.0
invalid_penalty: float = 0.0
success_reward: float = 0.0
timeout_penalty: float = 0.0
on_time_bonus: float = 0.0
late_penalty: float = 0.0
idle_penalty: float = 0.0
route_churn_penalty: float = 0.0
fairness_penalty: float = 0.0
total_reward: float = 0.0
class State(DispatchArenaModel):
"""Sanitized public environment state.
Hidden `prep_remaining` values are excluded unless visible prep mode is
explicitly enabled by the scenario config.
"""
episode_id: Optional[str] = None
tick: int = 0
max_ticks: int = 12
seed: Optional[int] = None
mode: Mode = Mode.MINI
nodes: List[Node] = Field(default_factory=list)
travel_time_matrix: Dict[str, Dict[str, int]] = Field(default_factory=dict)
couriers: List[Courier] = Field(default_factory=list)
orders: List[Order] = Field(default_factory=list)
reward_breakdown: RewardBreakdown = Field(default_factory=RewardBreakdown)
done: bool = False
truncated: bool = False
verifier_status: VerifierVerdict = VerifierVerdict.IN_PROGRESS
last_action: Optional[Action] = None
event_log: List[str] = Field(default_factory=list)
invalid_actions: int = 0
total_reward: float = 0.0
backlog: int = 0
sla_pressure: float = 0.0
@model_validator(mode="after")
def _validate_ticks(self) -> "State":
if self.tick < 0:
raise ValueError("tick must be >= 0")
if self.max_ticks <= 0:
raise ValueError("max_ticks must be > 0")
return self
class Observation(DispatchArenaModel):
"""Public observation returned by reset and step."""
state: State = Field(default_factory=State)
reward: float = 0.0
done: bool = False
truncated: bool = False
verifier_status: VerifierVerdict = VerifierVerdict.IN_PROGRESS
reward_breakdown: RewardBreakdown = Field(default_factory=RewardBreakdown)
legal_actions: List[str] = Field(default_factory=list)
action_mask: List[int] = Field(default_factory=list)
summary_text: str = "Awaiting dispatch."
info: Dict[str, Any] = Field(default_factory=dict)
@classmethod
def from_dict(cls, data: Mapping[str, Any]) -> "Observation":
return cls.model_validate(dict(data))
class EpisodeSummary(DispatchArenaModel):
episode_id: Optional[str] = None
seed: Optional[int] = None
mode: Mode = Mode.MINI
max_ticks: int = 12
ticks_taken: int = 0
invalid_actions: int = 0
total_reward: float = 0.0
final_verdict: VerifierVerdict = VerifierVerdict.IN_PROGRESS
action_trace: List[Action] = Field(default_factory=list)
delivered_orders: int = 0
expired_orders: int = 0
DispatchArenaAction = Action
DispatchArenaObservation = Observation
DispatchArenaState = State
|