Dataset for Reinforcement learning for joint scheduling in robotic mobile fulfillment systems
Description
This repository contains the comprehensive environment initialization files and datasets developed for the paper "Reinforcement learning for joint scheduling in robotic mobile fulfillment systems". It provides realistic layouts, initial pod inventory distributions, and long-term multi-batch order demand sequences for simulating Robotic Mobile Fulfillment Systems (RMFS) . The datasets are specifically designed to support research in Deep Reinforcement Learning (DRL), joint scheduling, and task allocation within automated warehouse environments. Dataset Variants The dataset is categorized into three distinct scenarios based on the scale of the warehouse and the occupancy density of the inventory: 1. Large RMFS Case (Large RMFS Case.zip) Scale: 4 Picking Stations, 200 Storage locations (20x10), 160 Pods, 20 Robots. Description: The baseline large-scale environment suitable for standard DRL algorithm testing, training, and evaluation. 2. Enlarged RMFS Case - 0.8 Occupancy (Enlarged RMFS Case with 0.8 occupancy.zip) Scale: 8 Picking Stations, 400 Storage locations (20x20), 320 Pods, 30 Robots. Description: A scaled-up environment with double the storage and station capacity. The initial pod inventory is generated with an 80% occupancy/density rate, presenting a balanced challenge for joint scheduling efficiency. 3. Enlarged RMFS Case - 0.9 Occupancy (Enlarged RMFS Case with 0.9 occupancy.zip) Scale: 8 Picking Stations, 400 Storage locations (20x20), 320 Pods, 30 Robots. Description: Shares the same physical footprint as the 0.8 case but features a higher inventory density (90% occupancy). This scenario is designed to test the robustness and efficiency of the DRL agent under heavy storage load and dense inventory distribution. Contents of Each Zip File Upon extracting any of the variant folders, you will find the following core components: RMFS_layout*.txt: The spatial coordinate dataset mapping the exact (X, Y) grid locations for picking stations, robot initial bases, and pod storage locations. Info_Pod_*.txt: The initial inventory matrix mapping cargo quantities and statuses to specific physical pods. Total_Info_long*.txt (Multiple files): The order demand sequences assigned to each picking station. Files are indexed (e.g., 1-25 for testing, 26-100 for training). README.md: Detailed technical documentation explaining the data structures, matrix reshaping logic, and environment parameters specific to that variant.