Scalable Partial-ACO Applied to Fleet Optimisation: Sampling and Multi-Colony Approaches

Abstract

Fleet optimisation can significantly reduce vehicular traversal of road networks cutting costs and increasing capacity. Reduced road use also leads to lower emissions and improved air quality, an increasingly important issue. Partial-ACO has previously shown promise in optimising vehicle fleets but issues remain over scalability. This paper demonstrates that increased numbers of ants improves results but with a quadratic computational cost. Consequently, this paper addresses this issue introducing two enhancements, sampling and multi-colony approaches. Whilst these are shown to reduce computational costs by up to 75% solution quality is impacted. Hence, this paper concludes that small numbers of ants run for many more iterations provides the best scalability although with fewer ants a higher risk exists of becoming trapped in local optima. This setup yields fleet traversal reductions for real-world scenarios of over 50% for up to 45 vehicles and 437 jobs. Moreover, using Partial-ACO, emissions of CO2 are cut by 3. 9Kg per vehicle a day improving air quality.