Dynamic Route Rationalization Model using AI & ML

Category: Software (AI / Logistics)
Difficulty: Intermediate
Time to Build: 4–6 weeks
Prerequisites: Python & basic ML, SQL or NoSQL basics, familiarity with Streamlit/Mapbox/Azure services
Deliverables: SRS, database schema, architecture diagram, Streamlit + Mapbox working prototype, test report.

Table of Contents

Problem Statement & Expected Outcome

Problem Statement
Delivery and distribution fleets face the challenge of minimizing distance, time, and cost while meeting vehicle-capacity and demand constraints. Traditional manual scheduling cannot keep pace with dynamic city logistics.

Expected Outcome
An AI/ML-powered web tool that dynamically clusters delivery points, assigns vehicles, and computes cost-optimal routes. Users can visualize and export complete distribution plans with minimal effort.

Abstract

This project provides an AI-based route optimization and visualization platform for sales and delivery vehicles. Using clustering, heuristic search, and Travelling Salesman Problem (TSP) optimization, it recommends distribution hubs, assigns vehicles, and plots efficient routes on an interactive map. The kit includes synthetic datasets and a Streamlit + Mapbox visualization for immediate testing and demonstration.

Details: User Stories & Acceptance Criteria

As a fleet planner, I can upload delivery locations and constraints to generate optimized routes and vehicle assignments.
As a distribution manager, I can see a map view with distance, time, and cost estimates and export the schedule to CSV.
Acceptance: Solution computes routes covering 100+ delivery points in under 2 minutes and loads the visualization within 3 seconds on broadband.

Scope & Modules

Module 1: Data Input & Preprocessing

Upload or fetch delivery locations, constraints (capacity, time windows), and sales potential.

Module 2: Clustering & Hub Assignment

Modified k-means to group destinations around source nodes.

Module 3: Vehicle Recommendation

Determine number and types of vehicles required, considering constraints.

Module 4: Route Optimization

Solve the Travelling Salesman Problem for each vehicle using heuristic algorithms.

Module 5: Visualization & Export

Streamlit + Mapbox dashboard shows all routes, delivery points, and estimated cost; export to CSV/PDF.

Module 6: Recommendation Mode

Suggests optimal distribution hubs or vehicle counts to cover a given percentage of potential sales.

Proposed Architecture & Tech Stack

Option A (Fast Deployment)

Frontend: Streamlit + Mapbox
Backend: Python (Flask/FastAPI)
Optimization: K-means clustering, heuristic TSP
Cloud: Azure App Service, Azure Maps (distance and routing APIs)

Option B (Enterprise Ready)

Backend: Django/DRF
Frontend: React or Next.js with Mapbox
Database: PostgreSQL/Supabase
Cloud: Azure or AWS for large-scale deployments

KPIs & Analytics

Average computation time per route set
Percentage reduction in travel distance vs. baseline
Fuel cost savings
Number of successful exports and route updates

Milestones & Timeline

Week 1: Gather requirements, finalize SRS, prepare sample datasets
Week 2: Implement data ingestion and clustering
Week 3: Add vehicle recommendation and preliminary routing
Week 4: Build Streamlit + Mapbox visualization
Week 5: Integrate export features and test large datasets
Week 6: Final testing, documentation, and presentation

Who It’s For

Students and Capstone Teams working on AI/ML or logistics optimization
Instructors seeking a deployable project for supply-chain or smart-city courses
Logistics and E-commerce Startups wanting a ready pilot for fleet optimization