# Logistics Intelligent

> A dispatch and supply-planning prototype for delivery routes, fleet disruptions, and retail replenishment across multiple warehouses.

Plan deliveries and stock around a changing day.

- URL: https://weaveward.com/work/last-mile-delivery-fleet-optimization
- Stage: Prototype
- Year: 2026
- Industry: Logistics
- Context: Internal product
- Capabilities: AI solutions, Business systems, Web apps, Mobile apps, Data & dashboards
- Technologies: Python, FastAPI, Google OR-Tools, LightGBM, H3, PostgreSQL + PostGIS, React, TypeScript, MapLibre, Flutter

## The challenge

A delivery plan changes throughout the day. Urgent orders arrive, traffic builds, vehicles break down and a stop can take longer than expected. Dispatchers need to adjust the plan while keeping drivers informed and customer commitments visible.

Retail supply adds another layer: which warehouse should serve each store, what stock will run out, and when an emergency delivery or purchase order is needed.

## What we built

Logistics Intelligent connects daily delivery execution with retail stock and supply planning.

- **A planner for real operating rules.** Orders, delivery windows, vehicle capacity, driver shifts, breaks and skills feed a route-planning engine. Dispatchers can review plans and lock decisions that should be preserved
- **A control tower for the day ahead.** A map, fleet status, alerts and at-risk deliveries help the team spot problems and choose the next action
- **Changes with approval.** Urgent orders, cancellations, delays and breakdowns produce concrete route-change proposals. Small changes can be applied automatically; larger changes need dispatcher approval. Version checks prevent an older proposal from overwriting a newer plan
- **Driver workflows.** The Flutter app code includes route caching, queued offline events, proof of delivery and synchronisation handling
- **Multi-warehouse replenishment.** Forecasts, inventory freshness, store-to-warehouse selection and truck routing work together to plan deliveries across the network
- **Stock and procurement.** Stock-out risk, emergency requests, transfer suggestions and peak-season preparation connect to buyer-approved purchase orders. Supplier receipts update inventory and lead-time estimates

## How it works

### Plan and execution together

The platform keeps delivery planning, live exceptions and plan-versus-actual analytics together. Operations can also compare network scenarios and investigate why a route or sourcing choice was proposed.

### Integrations with fallback paths

Maps, notifications, telematics and model providers are integrations with fallback paths. The core demo can run without those external services.

## Current stage

This is an internal engineering prototype with synthetic route and network benchmarks. A documented development-laptop run planned 1,000 stops across 60 vehicles in 45.6 seconds, although 221 visits missed their soft deadlines. A separate 5,000-store replenishment run took 15.6 minutes; 10% of visits arrived after the receiving window but within its configured tolerance.

These are historical development measurements, not customer savings or production fleet guarantees. The documented throughput checks used a single API process on SQLite. The native driver app, real map interactions and external-service deployments still need validation, followed by a live operational pilot.

## Where this fits

Teams coordinating vehicles, stores and warehouses benefit from a shared planning workflow and clear change control. See our [business systems](/services/business-systems), [AI solutions](/services/ai-solutions) and [logistics work](/industries/logistics), or [talk through your operation](/contact).