| from __future__ import annotations |
|
|
| from random import Random |
|
|
| from .models import AgentState, GridConfig, OrderState, Scenario, ZonePhase |
|
|
|
|
| def _clamp(value: int, low: int, high: int) -> int: |
| return max(low, min(high, value)) |
|
|
|
|
| def _shift_point( |
| point: tuple[int, int], |
| width: int, |
| height: int, |
| rng: Random, |
| max_shift: int, |
| ) -> tuple[int, int]: |
| x, y = point |
| return ( |
| _clamp(x + rng.randint(-max_shift, max_shift), 0, width - 1), |
| _clamp(y + rng.randint(-max_shift, max_shift), 0, height - 1), |
| ) |
|
|
|
|
| def _unique_points(points: tuple[tuple[int, int], ...]) -> tuple[tuple[int, int], ...]: |
| seen: list[tuple[int, int]] = [] |
| for point in points: |
| if point not in seen: |
| seen.append(point) |
| return tuple(seen) |
|
|
|
|
| def _vary_order( |
| order: OrderState, |
| grid: GridConfig, |
| rng: Random, |
| *, |
| time_shift: int, |
| spatial_shift: int, |
| reward_shift: int, |
| deadline_shift: int, |
| ) -> OrderState: |
| created_at = max(0, order.created_at + rng.randint(-time_shift, time_shift)) |
| pickup = _shift_point(order.pickup_location, grid.width, grid.height, rng, spatial_shift) |
| drop = _shift_point(order.drop_location, grid.width, grid.height, rng, spatial_shift) |
| reward_value = max(4.0, order.reward_value + rng.randint(-reward_shift, reward_shift)) |
| deadline = max(created_at + 4, order.deadline + rng.randint(-deadline_shift, deadline_shift)) |
| return order.model_copy( |
| update={ |
| "created_at": created_at, |
| "pickup_location": pickup, |
| "drop_location": drop, |
| "reward_value": float(reward_value), |
| "deadline": deadline, |
| }, |
| deep=True, |
| ) |
|
|
|
|
| def _vary_scenario( |
| scenario: Scenario, |
| seed: int | None, |
| *, |
| hotspot_shift: int, |
| congestion_shift: int, |
| time_shift: int, |
| spatial_shift: int, |
| reward_shift: int, |
| deadline_shift: int, |
| ) -> Scenario: |
| if seed is None: |
| return scenario |
|
|
| rng = Random(f"{scenario.name}:{seed}") |
| varied_grid = scenario.grid.model_copy( |
| update={ |
| "hotspots": _unique_points(tuple( |
| _shift_point(point, scenario.grid.width, scenario.grid.height, rng, hotspot_shift) |
| for point in scenario.grid.hotspots |
| )) |
| }, |
| deep=True, |
| ) |
| varied_orders = tuple( |
| sorted( |
| ( |
| _vary_order( |
| order, |
| varied_grid, |
| rng, |
| time_shift=time_shift, |
| spatial_shift=spatial_shift, |
| reward_shift=reward_shift, |
| deadline_shift=deadline_shift, |
| ) |
| for order in scenario.orders |
| ), |
| key=lambda item: (item.created_at, item.order_id), |
| ) |
| ) |
| varied_hotspot_phases = tuple( |
| phase.model_copy( |
| update={ |
| "points": tuple( |
| _unique_points(tuple( |
| _shift_point(point, scenario.grid.width, scenario.grid.height, rng, hotspot_shift) |
| for point in phase.points |
| ))) |
| }, |
| deep=True, |
| ) |
| for phase in scenario.hotspot_phases |
| ) |
| varied_congestion_phases = tuple( |
| phase.model_copy( |
| update={ |
| "points": tuple( |
| _unique_points(tuple( |
| _shift_point(point, scenario.grid.width, scenario.grid.height, rng, congestion_shift) |
| for point in phase.points |
| ))) |
| }, |
| deep=True, |
| ) |
| for phase in scenario.congestion_phases |
| ) |
| return scenario.model_copy( |
| update={ |
| "grid": varied_grid, |
| "orders": varied_orders, |
| "hotspot_phases": varied_hotspot_phases, |
| "congestion_phases": varied_congestion_phases, |
| }, |
| deep=True, |
| ) |
|
|
|
|
| def build_low_demand_scenario(seed: int | None = None) -> Scenario: |
| scenario = Scenario( |
| name="low_demand", |
| grid=GridConfig(width=8, height=8, hotspots=((5, 5), (6, 5), (5, 6))), |
| agents=( |
| AgentState(agent_id="a1", location=(1, 1)), |
| AgentState(agent_id="a2", location=(4, 2)), |
| AgentState(agent_id="a3", location=(6, 6)), |
| ), |
| orders=( |
| OrderState(order_id="o1", created_at=0, pickup_location=(1, 2), drop_location=(3, 3), reward_value=10, deadline=10), |
| OrderState(order_id="o2", created_at=2, pickup_location=(5, 2), drop_location=(6, 4), reward_value=12, deadline=13), |
| OrderState(order_id="o3", created_at=5, pickup_location=(2, 6), drop_location=(1, 7), reward_value=8, deadline=17), |
| OrderState(order_id="o4", created_at=8, pickup_location=(5, 5), drop_location=(7, 6), reward_value=11, deadline=19), |
| OrderState(order_id="o5", created_at=12, pickup_location=(2, 1), drop_location=(4, 1), reward_value=9, deadline=23), |
| OrderState(order_id="o6", created_at=16, pickup_location=(6, 5), drop_location=(7, 7), reward_value=13, deadline=28), |
| OrderState(order_id="o7", created_at=21, pickup_location=(3, 4), drop_location=(1, 5), reward_value=8, deadline=31), |
| OrderState(order_id="o8", created_at=26, pickup_location=(4, 6), drop_location=(6, 7), reward_value=14, deadline=36), |
| ), |
| episode_horizon=40, |
| default_max_decision_steps=20, |
| hotspot_phases=( |
| ZonePhase(start_time=0, points=((5, 5), (6, 5), (5, 6))), |
| ZonePhase(start_time=24, points=((5, 5), (6, 5), (6, 6))), |
| ), |
| briefing="Sparse city with generous deadlines. Most orders are feasible, so wasted idle time and unnecessary detours matter.", |
| dispatch_objective="Prefer clean on-time execution and avoid leaving easy nearby work untouched.", |
| known_future_signal="", |
| ) |
| return _vary_scenario( |
| scenario, |
| seed, |
| hotspot_shift=0, |
| congestion_shift=0, |
| time_shift=0, |
| spatial_shift=0, |
| reward_shift=1, |
| deadline_shift=0, |
| ) |
|
|
|
|
| def build_high_demand_scenario(seed: int | None = None) -> Scenario: |
| scenario = Scenario( |
| name="high_demand", |
| grid=GridConfig( |
| width=10, |
| height=10, |
| congested_zones=((4, 4), (4, 5), (5, 4), (5, 5), (6, 5)), |
| hotspots=((7, 7), (7, 8), (8, 7), (8, 8)), |
| ), |
| agents=( |
| AgentState(agent_id="a1", location=(1, 1)), |
| AgentState(agent_id="a2", location=(2, 8)), |
| AgentState(agent_id="a3", location=(8, 2)), |
| AgentState(agent_id="a4", location=(6, 6)), |
| ), |
| orders=( |
| OrderState(order_id="o1", created_at=0, pickup_location=(7, 7), drop_location=(9, 8), reward_value=16, deadline=12), |
| OrderState(order_id="o2", created_at=1, pickup_location=(6, 7), drop_location=(7, 9), reward_value=12, deadline=11), |
| OrderState(order_id="o3", created_at=2, pickup_location=(3, 3), drop_location=(1, 5), reward_value=9, deadline=10), |
| OrderState(order_id="o4", created_at=4, pickup_location=(8, 7), drop_location=(9, 9), reward_value=15, deadline=15), |
| OrderState(order_id="o5", created_at=6, pickup_location=(2, 8), drop_location=(4, 9), reward_value=10, deadline=16), |
| OrderState(order_id="o6", created_at=8, pickup_location=(7, 8), drop_location=(9, 6), reward_value=18, deadline=19), |
| OrderState(order_id="o7", created_at=11, pickup_location=(5, 2), drop_location=(2, 2), reward_value=11, deadline=20), |
| OrderState(order_id="o8", created_at=13, pickup_location=(8, 8), drop_location=(6, 9), reward_value=14, deadline=23), |
| OrderState(order_id="o9", created_at=17, pickup_location=(1, 7), drop_location=(3, 9), reward_value=9, deadline=26), |
| OrderState(order_id="o10", created_at=17, pickup_location=(7, 7), drop_location=(8, 9), reward_value=17, deadline=24), |
| OrderState(order_id="o11", created_at=18, pickup_location=(8, 8), drop_location=(9, 7), reward_value=15, deadline=24), |
| OrderState(order_id="o12", created_at=22, pickup_location=(6, 6), drop_location=(8, 4), reward_value=13, deadline=31), |
| OrderState(order_id="o13", created_at=22, pickup_location=(7, 9), drop_location=(9, 9), reward_value=16, deadline=29), |
| OrderState(order_id="o14", created_at=28, pickup_location=(7, 7), drop_location=(9, 7), reward_value=12, deadline=37), |
| OrderState(order_id="o15", created_at=34, pickup_location=(4, 1), drop_location=(5, 3), reward_value=8, deadline=42), |
| OrderState(order_id="o16", created_at=36, pickup_location=(8, 7), drop_location=(9, 5), reward_value=14, deadline=43), |
| OrderState(order_id="o17", created_at=19, pickup_location=(1, 1), drop_location=(9, 9), reward_value=5, deadline=25), |
| OrderState(order_id="o18", created_at=15, pickup_location=(0, 8), drop_location=(9, 0), reward_value=6, deadline=22), |
| ), |
| episode_horizon=60, |
| default_max_decision_steps=25, |
| hotspot_phases=( |
| ZonePhase(start_time=0, points=((7, 7), (7, 8), (8, 7), (8, 8))), |
| ZonePhase(start_time=26, points=((6, 7), (7, 7), (7, 8), (8, 8))), |
| ), |
| congestion_phases=( |
| ZonePhase(start_time=0, points=((4, 4), (4, 5), (5, 4), (5, 5), (6, 5))), |
| ZonePhase(start_time=38, points=((5, 5), (5, 6), (6, 5), (6, 6), (7, 6))), |
| ), |
| briefing="Demand arrives faster than the fleet can comfortably absorb, with clustered premium bursts and low-yield long-haul distractions.", |
| dispatch_objective="Maximize cumulative reward under sustained capacity pressure; serving everything greedily should no longer be obviously best.", |
| known_future_signal="", |
| ) |
| return scenario if seed is None else scenario |
|
|
|
|
| def build_hotspot_congestion_scenario(seed: int | None = None) -> Scenario: |
| scenario = Scenario( |
| name="hotspot_congestion", |
| grid=GridConfig( |
| width=15, |
| height=15, |
| congested_zones=( |
| (6, 6), (6, 7), (6, 8), |
| (7, 6), (7, 7), (7, 8), |
| (8, 6), (8, 7), (8, 8), |
| (10, 10), (10, 11), (11, 10), |
| ), |
| hotspots=((11, 11), (11, 12), (12, 11), (12, 12), (13, 12)), |
| ), |
| agents=( |
| AgentState(agent_id="a1", location=(2, 2)), |
| AgentState(agent_id="a2", location=(3, 10)), |
| AgentState(agent_id="a3", location=(10, 4)), |
| AgentState(agent_id="a4", location=(13, 13)), |
| AgentState(agent_id="a5", location=(8, 12)), |
| ), |
| orders=( |
| OrderState(order_id="o1", created_at=0, pickup_location=(11, 11), drop_location=(14, 14), reward_value=20, deadline=16), |
| OrderState(order_id="o2", created_at=1, pickup_location=(12, 11), drop_location=(10, 14), reward_value=18, deadline=15), |
| OrderState(order_id="o3", created_at=3, pickup_location=(4, 5), drop_location=(2, 9), reward_value=10, deadline=15), |
| OrderState(order_id="o4", created_at=5, pickup_location=(11, 12), drop_location=(13, 10), reward_value=17, deadline=18), |
| OrderState(order_id="o5", created_at=7, pickup_location=(9, 11), drop_location=(12, 13), reward_value=16, deadline=20), |
| OrderState(order_id="o6", created_at=10, pickup_location=(6, 4), drop_location=(3, 2), reward_value=11, deadline=21), |
| OrderState(order_id="o7", created_at=14, pickup_location=(13, 12), drop_location=(14, 9), reward_value=19, deadline=25), |
| OrderState(order_id="o8", created_at=18, pickup_location=(10, 11), drop_location=(7, 13), reward_value=14, deadline=28), |
| OrderState(order_id="o9", created_at=18, pickup_location=(12, 12), drop_location=(14, 10), reward_value=22, deadline=27), |
| OrderState(order_id="o10", created_at=19, pickup_location=(11, 11), drop_location=(13, 14), reward_value=18, deadline=26), |
| OrderState(order_id="o11", created_at=24, pickup_location=(3, 12), drop_location=(1, 14), reward_value=9, deadline=34), |
| OrderState(order_id="o12", created_at=31, pickup_location=(12, 12), drop_location=(14, 8), reward_value=20, deadline=41), |
| OrderState(order_id="o13", created_at=31, pickup_location=(10, 12), drop_location=(13, 13), reward_value=17, deadline=39), |
| OrderState(order_id="o14", created_at=39, pickup_location=(11, 10), drop_location=(8, 11), reward_value=13, deadline=48), |
| OrderState(order_id="o15", created_at=47, pickup_location=(5, 11), drop_location=(8, 14), reward_value=12, deadline=58), |
| OrderState(order_id="o16", created_at=47, pickup_location=(12, 11), drop_location=(14, 14), reward_value=21, deadline=56), |
| OrderState(order_id="o17", created_at=56, pickup_location=(13, 11), drop_location=(14, 13), reward_value=21, deadline=66), |
| OrderState(order_id="o18", created_at=63, pickup_location=(10, 3), drop_location=(6, 1), reward_value=10, deadline=74), |
| OrderState(order_id="o19", created_at=32, pickup_location=(1, 2), drop_location=(14, 14), reward_value=6, deadline=40), |
| OrderState(order_id="o20", created_at=26, pickup_location=(0, 0), drop_location=(14, 13), reward_value=7, deadline=35), |
| ), |
| episode_horizon=80, |
| default_max_decision_steps=30, |
| hotspot_phases=( |
| ZonePhase(start_time=0, points=((11, 11), (11, 12), (12, 11), (12, 12), (13, 12))), |
| ZonePhase(start_time=18, points=((9, 10), (10, 10), (10, 11), (11, 11), (11, 12))), |
| ZonePhase(start_time=38, points=((7, 10), (7, 11), (8, 11), (8, 12), (9, 12))), |
| ZonePhase(start_time=54, points=((5, 11), (6, 11), (7, 12), (8, 12), (8, 13))), |
| ), |
| congestion_phases=( |
| ZonePhase( |
| start_time=0, |
| points=((6, 6), (6, 7), (6, 8), (7, 6), (7, 7), (7, 8), (8, 6), (8, 7), (8, 8)), |
| ), |
| ZonePhase( |
| start_time=22, |
| points=((8, 8), (8, 9), (9, 8), (9, 9), (10, 9), (10, 10), (10, 11), (11, 10)), |
| ), |
| ZonePhase( |
| start_time=46, |
| points=((5, 10), (5, 11), (6, 10), (6, 11), (7, 11), (8, 11), (8, 12)), |
| ), |
| ), |
| briefing="Large city with moving hotspot pressure, congestion pockets, premium spikes, and multiple low-yield long-haul traps.", |
| dispatch_objective="Force strategic tradeoffs: a strong policy should balance premium hotspot work against detours, congestion, and selective rejection.", |
| known_future_signal="", |
| ) |
| return _vary_scenario( |
| scenario, |
| seed, |
| hotspot_shift=1, |
| congestion_shift=1, |
| time_shift=2, |
| spatial_shift=1, |
| reward_shift=2, |
| deadline_shift=2, |
| ) |
|
|
|
|
| SCENARIO_BUILDERS = { |
| "low_demand": build_low_demand_scenario, |
| "high_demand": build_high_demand_scenario, |
| "hotspot_congestion": build_hotspot_congestion_scenario, |
| } |
|
|
