Problem
Dispatchers at container logistics companies plan routes manually every morning. No tooling for constraint-aware optimization (time windows, vehicle capacity, driver shifts). Scheduling takes 1-2 hours and produces suboptimal routes.
Solution
Web app with FastAPI backend and Streamlit UI. PyVRP solves the Vehicle Routing Problem with real-world constraints. OpenRouteService/Nominatim for geocoding and road-network routing. Supabase for order and fleet data. Docker for deployment on Hetzner.
Result
PoC built and validated with DINO Containerdienst. Route quality and time savings measured against current manual process.
Lessons Learned
- To be filled after completion.
Deep Dive
VRP formulation: capacity constraints, time windows, multi-depot. PyVRP chosen over OR-Tools for Python-native API and active maintenance. Geocoding pipeline: address normalization + ORS matrix for travel times.