Optimize Delivery Route and Order Dispatching for Next-day Delivery Service

Introduction

In logistics, deciding which vehicle should carry each order and in what sequence it should drive along real roads is a classic yet still challenging problem. You must simultaneously satisfy business constraints — vehicle capacity, customer priority, time windows, owned vs. contracted fleet — while keeping the total driving distance close to optimal.

This project is a sample application that tackles the problem.

• Problem domain: Order dispatching and route optimization for next-day delivery services.
• Approach
  • Solves the problem as a VRPTW-style multi-objective (hard / medium / soft) score model with OptaPlanner.
  • Computes real road-graph distances and travel times with GraphHopper on OpenStreetMap (OSM) data.
  • Runs the whole system on AWS — defined with AWS CDK, executed on serverless + ECS, and visualized in a React-based web UI.

After uploading the day's orders, an operator can immediately review the per-vehicle dispatch result and the actual driving route on the map.

---

Demo Scenario

The bundled sample data models a same-day medical-supply delivery to hospitals in Seoul, Korea.

Scenario

• Customers are hospitals in Seoul that order medical supplies every day.
• Orders are dispatched once per day as a batch.
• Every delivery starts from a single warehouse.
• Customers have delivery priorities.
• A single customer may place multiple orders.
• Orders are delivered by company-owned vehicles by default; on busy days, temporarily contracted vehicles are also used.

Business Considerations

• Orders are dispatched to vehicles with respect to vehicle capacity.
• Total driving distance must be minimized.
• An order must be delivered by a vehicle whose time group is earlier than the ordering customer's time group.
• Multiple orders from the same customer should preferably be handled by a single vehicle in one trip.
• Company-owned vehicles are assigned first; contracted vehicles are only used when all owned vehicles are saturated.

This scenario and its data (apps_infra/scripts/data/sample_order.csv) let you reproduce the optimization result in the deployed environment. See Quickstart §3 Run Demo for the walkthrough.

---

Documentation

• Quickstart Guide

  • Requirements and tech-stack versions
  • AWS credential setup, build order (Optimization Engine → Web → Infra), and deploy
  • Running the demo (master data → distance cache → order upload) and uninstall

• Architecture

  • Solution architecture diagram
  • Domain model

Project Governance

• CHANGELOG.md — Release notes
• CONTRIBUTING.md — Bug reports and pull-request workflow
• CODE_OF_CONDUCT.md — Amazon Open Source Code of Conduct
• LICENSE — MIT-0
• LICENSE_THIRDPARTY.txt — License list of every npm / Maven / Gradle dependency used across the workspaces

---

Repository Layout

.
├── apps_opt_engine/   # Optimization Engine (Java 21 + Spring Boot + OptaPlanner + GraphHopper)
├── apps_web/          # Web App (React 19 + Vite + Cloudscape + MapLibre)
├── apps_infra/        # Infrastructure-as-Code (AWS CDK + TypeScript, single pnpm package)
└── docs/              # Project documentation (quickstart, architecture, images)

---

References

This sample project refers to AWS Last Mile Delivery Hyperlocal — Last Mile Logistics.

License

This sample project is licensed under the MIT-0. See the LICENSE file.