PyVRP is a high-performance, open-source solver for the Vehicle Routing Problem (VRP) developed by RoutingLab. It is designed to handle complex routing scenarios by utilizing state-of-the-art metaheuristic algorithms and efficient local search methods. The project offers a high-level Python modelling interface that allows users to define routing problems, including depots, clients, and heterogeneous vehicle types, without needing to perform manual implementation of low-level constraints.

Functionality of the library includes the ability to solve various VRP variants, such as those involving time windows, multiple depots, heterogeneous fleets, and reload operations. It is optimized to perform well on large-scale instances with thousands of clients, delivering high-quality solutions within a reasonable amount of time. The software combines C++ core components for computational speed with a flexible Python interface for usability.

Some of the key features are:

• Versatile Problem Support: Handles capacitied VRP, VRP with time windows, multi-depot VRP, multi-trip VRP, and prize-collecting variants.
• Heterogeneous Fleet Management: Supports vehicles with varying capacities, costs, shift durations, and routing profiles.
• Advanced Constraints: Provides support for client-specific service durations, release times, and mutually exclusive client groups.
• High-Level Modeling Interface: Simplifies the definition of routing problems using familiar concepts like locations, depots, and vehicle types.
• Enterprise Extensions: Offers additional modules for EU and US hours-of-service compliance, compatibility constraints, and fairness-based load and duration balancing.
• Optimized Performance: Leverages modern metaheuristics and efficient local search algorithms based on concatenation schemes to provide near-optimal results.

PyVRP operates through a clean, intuitive Python API where users construct a model by adding depots, clients, and vehicle types, then specify edges between locations. Once the model is defined, the solve method is triggered to initiate the optimization process using the solver's built-in algorithm. The tool allows for granular control over runtime and stopping criteria, making it suitable for both quick iterative development and robust production environments.

Some common use cases include:

• Last-Mile Delivery: Optimizing delivery routes for goods distribution to meet specific client time windows and vehicle capacities.
• Fleet Operations: Balancing shift durations and load assignments across a heterogeneous fleet for improved driver and vehicle utilization.
• Multi-Depot Distribution: Planning complex supply chains where multiple depots supply various clients and require reload operations.
• Prize Collection Planning: Selecting optimal subsets of clients to visit based on priority or prize value, commonly found in team orienteering scenarios.