HYPAR ABC: A Parallel Hybrid Method for Enhancing the Artificial Bee Colony Algorithm on Large Scale Transportation Problems

Abstract

This study introduces HYPAR‑ABC, a parallel hybrid artificial bee colony algorithm designed to solve large‑scale vehicle routing problems efficiently. The solution population is partitioned into multiple “islands” that evolve independently, periodically exchanging information to balance exploration and exploitation. Dynamic adaptation mechanisms automatically adjust algorithm parameters based on search progress, while advanced migration strategies tune the frequency and rate of solution exchange. Selective local search further refines promising routes. Evaluated on benchmark datasets (Solomon, Gehring & Homberger, TSPLIB, Li & Lim) and simulated industrial logistics data, HYPAR‑ABC reduced optimality gaps by over 67%, achieved speedups up to 89× with 128 processors, and maintained parallel efficiencies between 0.70–0.95. Despite increased memory use with many islands, the method scales robustly and adapts to dynamic conditions, offering a practical framework for fleet routing in large distribution networks and paving the way for future extensions such as machine‑learning–driven parameter tuning and multi‑objective optimization.