THULab/defense_logistics_stochastic_simulation

Defense Logistics Simulation Suite (TsFile)

Apache TsFile version of AIMindTeams/defense-logistics-stochastic-simulation.

Overview

Mission-critical supply-chain telemetry engineered for zero-drift mathematical precision and high-stakes stochastic chaos.

This is a high-fidelity synthetic dataset representing a Tier-3 End Unit (Forward Operating Base) operating under continuous stress and operational surges. It is designed to bridge the gap between "clean" academic datasets and the messy reality of edge-node logistics. This 5,000-hour sample is a subset of the AI Mind Teams 50,000-Hour Premium Suite, engineered to stress-test Reinforcement Learning (RL) agents and forecasting models.

• Rows: 5,000 hourly records across 9 simulated episodes (Episode_ID 1–9, station Tier_3).
• Cadence: 1 record per simulated hour.
• Time span: Timestamp_hr runs from 0 to 4,999; each episode is a contiguous segment of that run (e.g. episode 1 = hours 0–24, episode 9 = hours 800–4,999).

Key Features

• 0.0 Mathematical Drift: Verified flow conservation physics: $I_t = \max(0, I_{t-1} + R_t - D_t)$.
• Route Severance Physics: Dynamic lead-time spikes (24h to 150h+).
• Stochastic Demand: Poisson-distributed consumption with periodic operational surges.
• Transit Pipeline Queue: Real-time tracking of orders in the "void."

Data dictionary

Column Name	Description
Current_Inventory_Pallets	Net physical inventory at the start of the hour.
Inbound_Transit_Pallets	Pipeline Arrival: Total supply physically arriving this hour.
Lead_Time_hrs	Current expected transit time for newly placed orders.
Stochastic_Demand	End-user consumption for the hour.
Holding_Cost_USD	$2.50 per pallet per hour.
Stockout_Penalty_USD	$1,000.00 per pallet shortfall.

Schema (TsFile structure)

• Time (INT64, milliseconds) — hours-since-simulation-start in milliseconds, computed as Time = Timestamp_hr × 3,600,000. The source hour index Timestamp_hr (0–4,999) is therefore recoverable as Time / 3,600,000; it is not duplicated as a separate FIELD.
• Episode_ID, Node_Tier (TAG) — device/series dimensions. Read one episode with WHERE Episode_ID='1' (all rows currently have Node_Tier='Tier_3').
• Current_Inventory_Pallets, Inbound_Transit_Pallets, Stochastic_Demand, Stockout_Penalty_USD (INT64) — hourly integer measurements.
• Lead_Time_hrs, Holding_Cost_USD (DOUBLE) — hourly floating-point measurements.

Licensing & Commercial Use

This 5,000-hour sample is provided under the Creative Commons Attribution-NonCommercial 4.0 (CC BY-NC 4.0) license. It is intended for academic research and non-commercial exploration.

For commercial research, interactive RL training, and full-scale benchmarking, the AI Mind Teams Enterprise Suite offers a 50,000-hour full dataset, a Farama Gymnasium environment, and pre-tuned PPO & (s, S) baseline scripts.

Contact: aimindteams@gmail.com

Usage

Install the Apache TsFile Python SDK (pip install tsfile) and read a converted file:

from pathlib import Path
from tsfile import TsFileReader

path = Path("defense_logistics_stochastic_simulation.tsfile")
with TsFileReader(str(path)) as reader:
    schemas = reader.get_all_table_schemas()
    print("tables:", list(schemas))
    table_name = next(iter(schemas))
    table = schemas[table_name]
    columns = [column.get_column_name() for column in table.get_columns()]
    print("columns:", columns)
    field_names = [
        column.get_column_name()
        for column in table.get_columns()
        if column.get_column_name() not in {"Time", "time"}
    ]
    if field_names:
        with reader.query_table(table_name, field_names[:3], batch_size=1024) as result:
            batch = result.read_arrow_batch()
            if batch is not None:
                print(batch.to_pandas().head())

Source & license

• Original dataset: https://huggingface.co/datasets/AIMindTeams/defense-logistics-stochastic-simulation
• Author / publisher: AI Mind Teams
• License: cc-by-nc-4.0