OptiFlow Logo

OptiFlow - Intelligent Production Scheduling System

Smarter Scheduling. Smoother Production.

OptiFlow is an intelligent production scheduling and workflow-management system developed for industrial print shops and light-manufacturing facilities. It empowers plant managers to create complex multi-task jobs, configure Directed Acyclic Graph (DAG) task dependencies, allocate specialized machinery and human minders, and automatically generate optimal, conflict-free schedules using constraint programming. With a dual-interface architecture—a feature-rich Manager Desktop Dashboard and a streamlined Mobile Worker Portal—OptiFlow bridges high-level optimization with real-time floor execution.

OptiFlow Command Center LIVE Dashboard
Figure 1: OptiFlow Command Center LIVE — Real-time shop-floor operational status, equipment utilization metrics, and activity tracking.

Team

Supervisor

Table of Contents

  1. Introduction
  2. Solution Architecture
  3. Software Design & UI Showcase
  4. Scheduling and Optimization
  5. System Usage and Setup
  6. Testing
  7. Project Achievements and Delivered Capabilities
  8. Conclusion
  9. Links

1. Introduction

1.1 Project Overview

Commercial printing and manufacturing environments handle dozens of concurrent, multi-stage production orders. A single product (such as a 5,000-unit sewn hardbound diary or school model exam papers) involves a strict sequence of dependent operations:

Order Intake ➔ High-Speed Printing ➔ [Ink Drying Wait] ➔ Folding ➔ Sewing / Binding ➔ Guillotine Cutting ➔ Quality Check

Each operation requires specific machine capabilities, certified machine minders, and precise processing windows. Manually scheduling these workflows across shared equipment inevitably causes bottlenecks, idle equipment, worker confusion, and missed delivery deadlines.

OptiFlow replaces error-prone whiteboards and spreadsheets with an automated, constraint-driven platform. It models production constraints mathematically and uses Google OR-Tools CP-SAT to calculate provably optimal or high-quality schedules in seconds.

1.2 Real-World Problem

Manual production scheduling in print shops suffers from critical operational friction:

1.3 Proposed Solution

OptiFlow solves these challenges by combining:

  1. Manager Command Center (Desktop): Complete oversight over active jobs, real-time fleet health, DAG workflow creation, machine and minder assignments, and historical analytics.
  2. 1-Click Mathematical Optimization Engine (Backend): CP-SAT constraint programming that slots tasks without overlaps, respects predecessor wait times, accounts for machine maintenance breaks, and prioritizes urgent orders.
  3. Floor Worker Mobile Portal (Mobile): A tailored mobile app for floor operators to view personalized shift queues (“My Tasks”), inspect machine assignments, claim unallocated work via the “Job Market”, and update task execution statuses in real time.

1.4 Main Features

2. Solution Architecture

OptiFlow is designed as a modular, decoupled four-tier architecture:

+-------------------------------------------------------------+
|                     PRESENTATION TIER                       |
|  Flutter Desktop (Manager Hub)  |  Flutter Mobile (Floor)   |
|  - Command Center Dashboard     |  - Worker Sign-In Portal  |
|  - DAG Order & Minder Builder   |  - Personal Task Stream   |
|  - Schedule & Analytics View    |  - Floor Job Market       |
+------------------------------+------------------------------+
                               |
                               | REST API / JSON (HTTP)
                               v
+-------------------------------------------------------------+
|                      APPLICATION TIER                       |
|                    FastAPI Backend (Python)                 |
|  - Pydantic Schema Validation & REST Endpoints              |
|  - Job & Task Lifecycle Management                          |
|  - Dual Resource & Capability Mapping                       |
|  - Optimizer Dispatch & Response Formatting                 |
+------------------------------+------------------------------+
                               |
         +---------------------+---------------------+
         |                                           |
         v                                           v
+-------------------------------+   +-------------------------------+
|         DATA TIER             |   |       OPTIMIZATION TIER       |
|      Supabase PostgreSQL      |   |    Google OR-Tools CP-SAT     |
| - Relational Schema & RLS     |   | - Constraint Programming      |
| - Auth & Role-Based Minders   |   | - No-Overlap 2D Intervals     |
| - Jobs, Tasks, DAG Relations  |   | - Precedence & Wait Windows   |
| - Resource Statuses & Logs    |   | - Makespan & Priority Tuning  |
+-------------------------------+   +-------------------------------+

2.1 Presentation Tier (Flutter & Dart)

The frontend is built with Flutter for unified cross-platform execution on Windows Desktop (managers) and Android/iOS (floor minders). It features custom glassmorphic dark-mode styling, real-time status badges, responsive layout builders, and direct REST/Supabase client connectivity.

2.2 Application Tier (FastAPI & Python)

The backend provides high-performance asynchronous REST endpoints. It validates incoming order structures, enforces business rules, resolves candidate capabilities for machines and workers, builds constraint definitions for the optimizer, and persists results atomically.

2.3 Data Tier (Supabase PostgreSQL)

Supabase provides enterprise-grade PostgreSQL with real-time replication. Relational tables track:

2.4 Optimization Tier (Google OR-Tools CP-SAT)

The optimization engine formulates production scheduling as a Constraint Satisfaction and Optimization Problem (COP). It evaluates all valid machine candidates per operation, builds non-overlapping interval variables, enforces precedence inequalities, and minimizes the global makespan weighted by job priority.

3. Software Design & UI Showcase

3.1 Design Principles

3.2 Technology Stack

Component Technology Role
Manager Frontend Flutter (Desktop) Command center, DAG order creation, analytics
Worker Frontend Flutter (Mobile) Personal shift task stream, floor job market
Backend Framework FastAPI (Python 3.11+) Asynchronous API, validation, business logic
Optimization Solver Google OR-Tools CP-SAT Finite-domain constraint scheduling engine
Database & Auth Supabase (PostgreSQL) Relational persistence, Auth tokens, RLS
State Management Provider / Stateful Hooks Reactive UI updates across views
HTTP Communication HTTP / JSON REST Secure payload exchange

3.3 User Interface Walkthrough

3.3.1 Manager Command Center & Live Dashboard

The Command Center provides a high-level operational pulse of the entire manufacturing floor. Managers can monitor machine uptime gauges, active vs. offline equipment counts, total pending operations, task counts grouped by operation type, and a live activity audit feed.

Command Center Dashboard
Figure 2: Manager Command Center with fleet health badges (7 Active, 1 Offline), uptime gauge, and operation distributions.

3.3.2 Advanced Job Order Builder with DAG Sequencing

Creating a job order is divided into two intuitive sections:

  1. Order Details: Job title, client name, total print units, priority level (HIGH, MEDIUM, LOW), and delivery deadline date picker.
  2. Task Sequence (DAG): Managers can add multiple tasks, define predecessor dependencies (Depends On), enter duration in separate hours and minutes fields, restrict to specific machines or allow any capable machine, assign a certified human minder, and enable post-task maintenance cooldowns.

New Job Order Builder with DAG
Figure 3: Multi-task Job Order Creator supporting DAG dependencies, duration parsing, minder assignment, and cooling breaks.

3.3.3 Job Management & 1-Click CP-SAT Optimization

The Jobs screen lists all active and draft orders. Expanding an order reveals its sub-tasks, current execution state (DRAFT, PENDING, SCHEDULED, IN_PROGRESS, COMPLETED), allocated machine, and assigned minder. Clicking the purple Optimize button immediately invokes the backend CP-SAT solver, converting unscheduled tasks into timed machine allocations.

Job Management and Optimization Trigger
Figure 4: Job Management screen showing task statuses, assigned minder Sarah Chen, and the 1-click Optimize trigger.

3.3.4 Production Analytics & OEE Reporting

The Analytics & Reports module provides managers with deep insights into factory throughput. It calculates Overall Equipment Effectiveness (OEE), tracks historical lead time reductions, displays defect rates, and renders job status distribution charts over customized reporting windows (e.g., Last 30 Days).

Analytics and Reports Screen
Figure 5: Production Analytics tracking Overall Equipment Effectiveness (OEE) trends, average lead times, and order distribution.

3.3.5 Floor Worker & Minder Mobile Portal

Floor operators interact through a tailored mobile application that connects directly to their shift responsibilities:

Mobile Sign In   Mobile Worker Tasks   Mobile Job Market
Figure 6: Mobile Floor Portal — (Left) Worker Login, (Center) Personalized Task Queue for Sarah Chen, (Right) Floor Job Market.

4. Scheduling and Optimization

4.1 Production Routing Workflow

OptiFlow is designed around standard manufacturing workflows such as sewn bookbinding and model paper production:

[ Job Order Intake ]
        |
        v
[ 1. Printing ] ----------> Heidelberg Speedmaster / HP Indigo
        |
        v (Mandatory Ink Drying Wait: 15-30 mins)
[ 2. Folding ] -----------> MBO Folding Machine
        |
        v
[ 3. Sewing / Binding ] --> Horizon BQ-470 / Aster Sewing
        |
        v (Glue Curing Break: 10 mins)
[ 4. Cutting ] -----------> Polar 115 Guillotine
        |
        v
[ Completed & QC ]

4.2 Mathematical Formulation (CP-SAT)

The optimizer models the production floor as a discrete set of Jobs, Tasks, and Resources (Machines & Minders). Each operational constraint is formulated in Google OR-Tools CP-SAT as follows:

1. Precedence Constraint with Mandatory Wait / Drying Times

A successor task cannot start until its predecessor has finished and any required ink-drying or glue-curing interval has elapsed:

Start(Task B)  ≥  End(Task A) + Mandatory_Wait_Duration

2. Machine Disjunctive Constraint (No Overlapping Work)

A machine can process only one task at any given time. If two tasks are assigned to the same machine, their time intervals cannot overlap:

Interval(Task i) ∩ Interval(Task j) = ∅    (for all i ≠ j on Machine k)

3. Task-Level Machine Maintenance / Cooldown Break

Heavy print runs or thermal binding cycles require post-task machine maintenance or cooldown before the machine can accept new work:

Machine_Release_Time = Task_End_Time + Break_Duration

4. Dual-Resource Coupling (Machine + Certified Minder)

Operations requiring both a machine and a human operator ensure both resources are booked simultaneously for the exact same interval:

Start(Machine) = Start(Minder)   AND   End(Machine) = End(Minder)

5. Multi-Objective Function

The optimizer minimizes total makespan while applying heavy penalties for late completion against customer deadlines:

Minimize: [ α × Total_Makespan ] + Σ [ Priority_Weight(j) × Max(0, Completion_Time(j) - Deadline(j)) ]

5. System Usage and Setup

5.1 Prerequisites

5.2 Clone Repository

git clone https://github.com/cepdnaclk/e22-co2060-OptiFlow.git
cd e22-co2060-OptiFlow

5.3 Backend Setup

  1. Navigate to the backend directory: 
    cd optiflow_back
  2. Create and activate a Python virtual environment: 
    python -m venv venv
.\venv\Scripts\Activate.ps1
  3. Install dependencies: 
    pip install -r requirements.txt
  4. Configure environment variables (optiflow_back/.env): 
    SUPABASE_URL=https://your-project.supabase.co
SUPABASE_KEY=your-supabase-anon-or-service-key
  5. Seed demo datasets (Optional but recommended for demonstration): 
    # Step 1: Wipe and seed machines, human minders, and capabilities
python seed_pitch_data.py

# Step 2: Register Supabase Auth credentials for minders (Sarah, Marcus, Elena)
python create_minders.py

# Step 3: Insert sample un-optimized job orders for live demo
python seed_demo_final.py
  6. Start FastAPI server: 
    uvicorn main:app --reload --port 8000
    • Interactive Swagger Docs: http://127.0.0.1:8000/docs

5.4 Frontend Setup

  1. Open a new terminal from repository root: 
    cd optiflow_front
flutter pub get
  2. Start the desktop or mobile application: 
    # Run Desktop Manager App (Windows)
flutter run -d windows

# Run Mobile Worker App (Connected Android Device)
flutter run -d <device_id>
  3. 1-Click Startup: On Windows, you can launch both backend and frontends simultaneously by double-clicking: 
    .\run.bat

6. Testing

6.1 Backend Test Suite

The backend includes comprehensive test coverage for optimization constraints, input validation, and API routing:

cd optiflow_back
$env:PYTHONPATH="."
pytest -v tests/

Key verification areas include:

6.2 Frontend & Static Analysis

cd optiflow_front
flutter test
flutter analyze

7. Project Achievements and Delivered Capabilities

7.1 Key Completed Deliverables

7.2 System Evaluation and Operational Impact

8. Conclusion

OptiFlow demonstrates how modern constraint programming and cross-platform UI engineering can solve complex industrial production challenges. By combining Google OR-Tools CP-SAT with a responsive Flutter architecture and Supabase real-time storage, the system eliminates scheduling conflicts, respects physical operational constraints, optimizes equipment utilization, and seamlessly connects plant managers with floor workers.

Department of Computer Engineering
Faculty of Engineering
University of Peradeniya