The idea
A short route is not automatically a usable route. Deliveries also have capacities, service requirements and time windows. Wouda, Lan and Kool present PyVRP, a hybrid genetic-search solver with Python-level customisation and performance-critical C++ components.
What the authors established
Their paper reports computational comparisons on capacitated and time-window routing problems. The implementation offers a practical starting point for further routing work, rather than requiring every application team to build a solver from scratch.
Applications
Routing comparisons should use identical demand, travel times and service rules. Distance, missed deliveries and runtime together reveal whether a shorter route is operationally useful.
Where care is needed
A strong benchmark heuristic does not certify global optimality for every instance. Constraints and features must be checked against the solver version used.
Sources and credits
- PyVRP: A High-Performance VRP Solver Package
Niels A. Wouda, Leon Lan, Wouter Kool, published
INFORMS Journal on Computing 36(4), 943–955 (2024).
Original solver research by Wouda, Lan and Kool; our note explains its relevance. Software licensing is separate from article-figure rights.