Routing problem in cash transport: modeling, operational decomposition and comparison between PyVRP and PuLP

This article analyzes a case of daily planning in the transport of cash in banking operations, with synthetic data for security and confidentiality. The problem involves activating vehicles, defining exit bases and sequencing supply and collection orders, respecting service windows, physical and financial capacities and penalties for non-service. The case is modeled by a VRPTW-inspired common core with optional clients, broken down by operating class, and solved by PuLP/CBC and PyVRP in a scenario with 20 orders, 10 vehicles, and 5 bases. The solutions are evaluated by common index, feasibility, service, fleet, distance, duration and computational time. All orders were granted. In the full instance, PuLP obtained a lower rate, but required 800.6 s; PyVRP produced a viable solution in 2.0 s, indicating complementarity between mathematical reference and rapid heuristics.