EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Robust Carbon-Budget Screening for Offline Joint Inventory and Freight Control Under Emission-Factor Shift

Adrian Stewart, Eva Collins

International Journal of World Economic Research · 19.11.2026 · 4 (1) · 51-57

DOI: 10.54097/fd8ara21 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of World Economic Research, online erschienen am 19.11.2026. Verfasst von Adrian Stewart und 1 weiteren Autor:innen. Die Arbeit nennt 22 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Offline inventory and freight policies are increasingly evaluated under finite carbon budgets, yet a predicted-emissions filter does not itself prevent over-budget realized emissions when vehicle factors are uncertain. Building on the action-level pathwise carbon screen of carbon-aware conservative Q-learning (CA-CQL), which guarantees compliance for a predicted emissions ledger, we extend the screen to an uncertainty-bound feasibility projection that preserves behavior-supported replenishment proposals. The study constructs synthetic fixed logs for a supplier, distribution center, and three retailers and evaluates policy-independent demand and emission-error streams. Five offline training datasets contain 24,000 transitions in total, and each policy is evaluated over 250 held-out trajectories in each uncertainty condition. Under a 4,300 kg CO2e cap and emission multipliers in [0.8, 1.2], unfiltered behavior violates the cap in 30.0% of episodes, the nominal CA-CQL screen in 5.2%, and a 1.20 robust screen in 0%, while demand fill declines from 89.64% to 89.50%. Under mild [0.9, 1.3] and severe [1.1, 1.5] shifts, the 1.20 screen has 2.0% and 18.8% observed violations; thus the conditional guarantee does not extend past its assumed bound. A 1.35 factor reduces severe-shift violations to 6.4%. Comparisons with behavior cloning and support-regularized offline Q-learning separate constraint feasibility from policy quality, while common-random-number evaluation and seed-level paired tests quantify uncertainty. The result is a reproducible, auditable robust-dispatch interface rather than a claim of end-to-end optimal inventory reinforcement learning.

Zitieren

APA

Stewart, A., & Collins, E. (2026). Robust Carbon-Budget Screening for Offline Joint Inventory and Freight Control Under Emission-Factor Shift. International Journal of World Economic Research, 4(1), 51-57. https://doi.org/10.54097/fd8ara21

Vancouver

Stewart A, Collins E. Robust Carbon-Budget Screening for Offline Joint Inventory and Freight Control Under Emission-Factor Shift. International Journal of World Economic Research. 2026;4(1):51-57. doi:10.54097/fd8ara21

BibTeX
@article{Stewart2026,
  author = {Stewart, Adrian and Collins, Eva},
  title = {Robust Carbon-Budget Screening for Offline Joint Inventory and Freight Control Under Emission-Factor Shift},
  journal = {International Journal of World Economic Research},
  year = {2026},
  volume = {4},
  number = {1},
  pages = {51--57},
  doi = {10.54097/fd8ara21}
}
Kostenlos starten →