LeMur Scheduler - CP-SAT with EA

Overview

LeMur Scheduler is a Python-based optimization tool designed to solve scheduling problems using Google's OR-Tools CP-SAT solver. The project incorporates evolutionary algorithms (EAs) for refinement, focusing on job scheduling, machine assignments, and constraints like maintenance, operator availability, and production deadlines.

The solution supports both a graphical user interface (GUI) for managing inputs and outputs and a comprehensive backend for performing advanced optimization tasks.

---

Repository Structure

Main Folders and Files

• web_app/: Contains the Gradio-based web interface and essential logic.
  • app.py: The entry point for the web application.
  • Supporting Python files:
    • data_init.py: Handles data preprocessing and initialization.
    • solver.py: Contains the main optimization solver logic. You can also play aroung with the solver by directly running this file.
    • ga_refiner.py: Implements genetic algorithm-based refinement for schedules.
    • ga_utils.py: Utility functions for evolutionary algorithms.
    • utils.py: General helper functions.
• data/: Contains script to process the data provided from Lemur to generate the required input files and simulate orders.
• extras/: Additional files for testing, data pre-processing and performance benchmarking.

---

Features

1. Optimization Goals:

   • Minimize makespan while balancing workload.
   • Optimize machine and operator utilization.

2. Constraint Handling:

   • Machine compatibility.
   • Shift and operator scheduling.
   • Maintenance windows and holidays.

3. Evolutionary Algorithm Refinement:

   • Genetic algorithm to refine solutions.
   • Handles complex scenarios like overlapping operations.

4. Interactive GUI:

   • Upload and edit CSV/JSON input files.
   • Configure solver parameters.
   • Visualize schedules as Gantt charts.

---

Setup

Dependencies

Install the required Python packages:

pip install ortools gradio matplotlib pandas plotly inspyred intervaltree

Files

The following input files are required:

• new_orders.csv: Defines the list of jobs to be scheduled.
• lista_articoli.csv: Provides production statistics for each article.
• articoli_macchine.json: Defines the compatibility between articles and machines.
• macchine_info.json: Contains machine-specific details.

---

How to Use

Running the Web Interface

Run the following command to start the Gradio-based GUI while inside the web_app/ directory:

python app.py

GUI Features

1. Upload Input Files:

   • Upload and edit CSV/JSON files.
   • Save modified data back to the appropriate files.

2. Configure Solver Parameters:

   • Adjust scheduling horizon, shift times, and operator availability.
   • Define costs for setup, loading, and unloading operations.

3. Run the Solver:

   • Execute the scheduler and view outputs in real-time.
   • Visualize results as Gantt charts.

---

Inputs

Required Input Files

1. new_orders.csv:
   • Format:
     • cliente
     • cod_articolo
     • quantity
     • data inserimento
     • data consegna
   • Example: SCORTA, 4407DN, 472, 2025-04-20, 2025-08-06
2. running_orders.csv:
   • Format: - cliente - macchina - cod_articolo - quantity - fine operazione attuale - data consegna - levate rimanenti ciclo - tipo operazione attuale - operatore
   • Example: SCORTA,1,2277DN mini,124.37,2024-12-01 03:30:00,2025-11-19,1,2,0,0
3. lista_articoli.csv:
   • Production data for each article:
     • codarticolo
     • kg_ora (for a macine with 256 fusi)
     • no_cicli
     • ore_levata
   • Example: 10353ZF, 1.87, 2, 96.0
4. articoli_macchine.json:
   • Compatibility between articles and machines.
   • Example:

     {
        "2277DN mini": [1,2,4,25,70],
     }

5. macchine_info.json:
   • Contains the n_fusi for each machine.
   • Example:

   {
       "1": {
           "n_fusi": "392",
       },
   }

---

Outputs

1. Schedules:

   • output/schedule.txt: Initial schedule.
   • output/refined_schedule.txt: Refined schedule after GA optimization.

2. Visualizations:

   • Gantt charts for visualizing machine schedules.

---

Customization

Solver Parameters

Configure the solver by adjusting the following:

• Time Units:
  • Set time granularity (e.g., 24, 48, or 96 units per day).
• Operator Configuration:
  • Define the number of operator groups and members per group.
• Costs:
  • Adjust costs for setup, load, and unload phases.

Exceptions

• Define broken_machines, scheduled_maintenances, and festive_days to customize scheduling constraints.

---

Error Handling

• Missing Files: Ensure all input files are available before running the solver.
• Empty Rows in CSVs: The GUI automatically removes empty rows from uploaded files.

---

Example Usage

1. Upload the following:

   • new_orders.csv (Job list).
   • lista_articoli.csv (Article statistics).
   • articoli_macchine.json (Compatibility data).
   • macchine_info.json (Machine details).

2. Configure parameters:

   • Set scheduling horizon and operator availability.

3. Run the solver:

   • View results in the "Run Solver" tab.

---

Developers

This repository is created for LeMur for the Industrial AI challenge. Contributors:

• Davide Cavicchini
• Luca Cazzola
• Alessandro Lorenzi
• Emanuele Poiana
• Andrea Decarlo
• Silvano Maddonni