Ryan O'Neil - Build on-demand logistics apps with Python, OR-Tools, and DecisionOps | PyData Global

ORTools is a versatile optimization library that can handle forecasting, scheduling, and routing problems through different approaches (linear programming, mixed integer programming, constraint programming)

Three core models for on-demand logistics:

• Demand forecasting using LAD (Least Absolute Deviations) regression
• Shift scheduling using mixed integer programming
• Route planning using constraint programming/local search

LAD regression is preferable to least squares for forecasting because it’s more robust to outliers and can be customized with additional constraints

Scheduling optimization typically involves:

• Balancing oversupply vs undersupply of workers
• Different penalties for understaffing vs overstaffing
• Worker availability constraints
• Economic targets and service levels

Vehicle routing optimization focuses on:

• Minimizing drive time and costs
• Vehicle capacity constraints
• Start/end location requirements
• Sequencing of deliveries
• Driver assignments

Python has become the de facto language for optimization modeling, with growing adoption in data science and operations research

Optimization underlies many decision-making tools in logistics, even when hidden behind APIs

Different business cases require different cadences (daily, weekly, monthly) and customization of models

Moving from siloed spreadsheets to integrated optimization platforms can improve operational efficiency

The combination of forecasting, scheduling, and routing models creates a complete system for managing on-demand logistics operations