This dataset originates from a real-world footwear manufacturing warehouse and provides a comprehensive foundation for benchmarking research in warehouse order-picking operations. Data was collected via SQL queries on the company’s Warehouse Management System (WMS), resulting in diverse formats such as CSV files, CAD layouts, and Python scripts. The dataset includes geometric representations of the warehouse layout, with Cartesian-mapped storage locations, aisles, and central depots, detailed product classifications, storage positions, picking wave information, and routing paths. It supports evaluating various storage strategies, including Random, Class-Based, Dedicated, and Hybrid configurations, enabling the analysis of their impact on order-picking efficiency. Temporal data captures operational trends, including timestamps and operator-specific performance, offering insights into workflow efficiency and workload balancing. Anonymization and randomization techniques were applied while retaining realistic operational patterns to preserve confidentiality. This dataset is highly versatile and suitable for developing optimization algorithms for picker routing, order batching, wave generation, and intralogistics, as well as for advancing automation and robotics research through navigation-specific data for autonomous guided vehicles (AGVs) and robotic systems. This dataset significantly contributes to warehouse logistics research and operational optimization by supporting a wide range of applications.
Use case:
Warehouse Order-Picking Optimization and Intralogistics Analytics
| Example Observation | • Current picker position (x,y) – derivable via cumulative path in picker_routes.csv + layout file.• Remaining SKU locations of the current batch ( order_lines.csv).• Trolley load (items picked vs capacity). • Batch meta-data: priority, promised ship-time. • Warehouse congestion indicator – number of simultaneous pickers inside same aisle (from time-stamps). • Time of day / shift (from timestamp). |
| Example Action | Discrete: 1. Select next SKU location from the remaining list (≈ up to 30 options). 2. “Return-to-depot & drop-off” (if trolley close to full). 3. (Episode start) choose to merge two small orders into one batch (batching decision). |
| Example Reward | Step reward = − (travel time seconds) − λ ⋅ delay_penalty. Final reward bonus when the whole order is completed on/before due-time. 𝜆 is tuned so that finishing early dominates single-step speedups. |