Resource Allocation Using Reinforcement Learning in Industrial Cyber-Physical Systems

Document Type

Article

Source of Publication

IEEE Transactions on Industrial Cyber Physical Systems

Publication Date

1-1-2026

Abstract

Efficient and reliable resource allocation is essential for Industrial Cyber-Physical Systems (ICPS) operating under dynamic workloads and uncertain environments. Existing approaches often struggle to adapt to uncertain, dynamic environments in Industrial Cyber-Physical Systems (ICPS), lacking robust and sequential decision-making mechanisms that integrate reinforcement learning with online resource allocationunder real-time constraints. This paper proposes RL-RA, a reinforcement learning-based online allocation framework that models the problem as a constrained sequential decision process and integrates feasibility-aware control, uncertainty-informed state design, and robustness against disturbances. Theoretical analysis establishes bounded regret, constraint satisfaction, and a competitive ratio guarantee. Experiments on three real-world datasets demonstrate that RL-RA consistently outperforms four representative baselines in cumulative reward, violation rate, and robustness under both nominal and noisy conditions, confirming its effectiveness for resilient ICPS resource management.

ISSN

2832-7004

Publisher

Institute of Electrical and Electronics Engineers (IEEE)

Volume

4

First Page

181

Last Page

189

Disciplines

Computer Sciences

Keywords

industrial cyber-physical systems, Reinforcement learning, resource allocation

Scopus ID

105035661430

Indexed in Scopus

yes

Open Access

no

Link to Full Text

Share

COinS