Entropy-Aware Last-Mile Logistics: A Hyperheuristic Approach to Minimizing Operational Disorder

Kassem Danach1, Email

Charbel Chedrawi2,3

1Department of Computer Science and Information Technology, Faculty of Arts and Sciences, Lebanese International University, Beirut, 1001, Lebanon 
2USJ Business School, Saint Joseph University, Beirut, 1104 2020, Lebanon
3School of Business, University of Nicosia, Nicosia, CY-1700, Cyprus

 

Abstract

Last-mile delivery operations are characterized by high levels of execution uncertainty, where small disruptions can propagate along delivery routes and lead to service failures. Existing routing approaches primarily manage uncertainty through expected-value optimization or worst-case robustness, but often overlook how uncertainty is structured and accumulates during route execution. This paper introduces an entropy-aware framework for last-mile logistics that explicitly models operational disorder using information-theoretic measures. Service-time uncertainty at delivery nodes is represented through entropy, capturing both the magnitude and organization of variability. Route-level entropy is used to quantify execution fragility arising from uncertainty accumulation. Building on this formulation, the paper proposes an entropy-regularized routing problem and develops an entropy-aware hyperheuristic solution approach that balances routing cost and operational disorder through adaptive heuristic selection. Using real-world last-mile delivery data, the results show that route-level entropy is strongly and statistically significantly associated with delivery failure probability, even after controlling for route cost and size. Grounded in Shannon’s Information Theory, Complexity Theory, and Supply Chain Resilience Theory, our analysis of the structure of uncertainty demonstrates that route-level entropy predicts delivery failure and that moderate entropy penalties enhance last-mile execution resilience at minimal cost. Compared with deterministic, stochastic, and robust vehicle routing baselines, the proposed approach consistently produces routes with lower entropy, reduced delivery failures, and improved execution stability, while incurring only modest increases in routing cost. Sensitivity analyses further indicate that substantial reliability gains are achieved for moderate entropy penalties, with diminishing returns beyond this range. Overall, the results demonstrate that managing the structure of uncertainty, rather than its average magnitude alone, is critical for improving last-mile delivery reliability. The proposed entropy-aware framework provides a scalable and practical decision-support tool for last-mile logistics design in complex urban environments.

Entropy-Aware Last-Mile Logistics: A Hyperheuristic Approach to Minimizing Operational Disorder