Mitigating exposure bias for reliable ANN-based control: The enhanced scheduled sampling framework
ISA Transactions · 2026-11 · 178 · 333-347
DOI: 10.1016/j.isatra.2026.09.013 ↗Kurz gefasst
Fachartikel in der Zeitschrift ISA Transactions, 2026-11. Verfasst von Pau Comas und 3 weiteren Autor:innen. Beteiligte Einrichtung: IDEAS Research Group, Universitat Autónoma de Barcelona (UAB) Campus UAB, 08193 Bellaterra, Barcelona, Spain. Electronic address: pau.comas@uab.cat. Die Arbeit nennt 37 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0). Gefördert von: State Agency of Research, European Commission.
Abstract (Originalsprache)
The design of Artificial Neural Networks (ANNs) as controllers for industrial processes offers significant potential but is often hindered by the challenge of exposure bias. Classic training methods, such as Batch Learning (BL), can suffer from it when trained controllers are deployed in closed-loop systems. This paper introduces the Enhanced Scheduled Sampling (ESS) framework, a novel and comprehensive training methodology designed to mitigate these limitations and produce reliable ANN-based controllers. Based on the Scheduled Sampling curriculum learning strategy, the ESS framework integrates an online simulation environment with key enhancements such as Weighted ANN Actuations to blend controller outputs, a sequence-based control error term and a controlled dynamic learning rate adjustment strategy. The framework is validated against the baseline learning approach on a conventional PID control scenario acting on First-Order Plus Dead-Time processes as well as on a non-linear Continuous Stirred-Tank Reactor exhibiting second-order dynamics. The PID benchmark provides an industrially relevant reference, enabling the proposed ANN-based controller to be compared with the established standard for process-level feedback control. The results demonstrate that controllers trained with ESS consistently outperform their BL counterparts in both control performance and robustness to parametric variations and training initialisation. By effectively mitigating exposure bias, the ESS framework represents a significant step towards developing dependable, high-performance ANN controllers ready for real-world industrial deployment or to advance research in transfer learning for control. The open-source implementation is shared to encourage further research.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- IDEAS Research Group, Universitat Autónoma de Barcelona (UAB) Campus UAB, 08193 Bellaterra, Barcelona, Spain. Electronic address: pau.comas@uab.cat.
- IDEAS Research Group, Universitat Autónoma de Barcelona (UAB) Campus UAB, 08193 Bellaterra, Barcelona, Spain. Electronic address: antoni.morell@uab.cat.
- IDEAS Research Group, Universitat Autónoma de Barcelona (UAB) Campus UAB, 08193 Bellaterra, Barcelona, Spain. Electronic address: jose.vicario@uab.cat.
- IDEAS Research Group, Universitat Autónoma de Barcelona (UAB) Campus UAB, 08193 Bellaterra, Barcelona, Spain. Electronic address: ramon.vilanova@uab.cat.
Förderung
- State Agency of Research
- European Commission
Zitieren
APA
Comas, P., Morell, A., Lopez Vicario, J., & Vilanova, R. (2026). Mitigating exposure bias for reliable ANN-based control: The enhanced scheduled sampling framework. ISA Transactions, 178, 333-347. https://doi.org/10.1016/j.isatra.2026.09.013
Vancouver
Comas P, Morell A, Lopez Vicario J, Vilanova R. Mitigating exposure bias for reliable ANN-based control: The enhanced scheduled sampling framework. ISA Transactions. 2026;178:333-347. doi:10.1016/j.isatra.2026.09.013
BibTeX
@article{Comas2026,
author = {Comas, Pau and Morell, Antoni and Lopez Vicario, Jose and Vilanova, Ramon},
title = {Mitigating exposure bias for reliable ANN-based control: The enhanced scheduled sampling framework},
journal = {ISA Transactions},
year = {2026},
volume = {178},
pages = {333--347},
doi = {10.1016/j.isatra.2026.09.013}
}
EUVIMED