, Volume 21, Issue 1, pp 2–21

Fleet design optimisation from historical data using constraint programming and large neighbourhood search


DOI: 10.1007/s10601-015-9203-0

Cite this article as:
Kilby, P. & Urli, T. Constraints (2016) 21: 2. doi:10.1007/s10601-015-9203-0


We present an original approach to compute efficient mid-term fleet configurations, at the request of a Queensland-based long-haul trucking carrier. Our approach considers one year’s worth of demand data, and employs a constraint programming (CP) model and an adaptive large neighbourhood search (LNS) scheme to solve the underlying multi-day multi-commodity split delivery capacitated vehicle routing problem. Our solver is able to provide the decision maker with a set of Pareto-equivalent fleet setups trading off fleet efficiency against the likelihood of requiring on-hire vehicles and drivers. Moreover, the same solver can be used to solve the daily loading and routing problem. We carry out an extensive experimental analysis, comparing our approach with an equivalent mixed integer programming (MIP) formulation, and we show that our approach is a sound methodology to provide decision support for the mid- and short-term decisions of a long-haul carrier.


Vehicle routing problem Fleet size and mix Large neighbourhood search Mixed integer programming Constraint programming Pre-processing 

Copyright information

© Springer Science+Business Media New York 2015

Authors and Affiliations

  1. 1.Optimisation Research Group, NICTA Canberra Research LabCanberraAustralia

Personalised recommendations