Neural Parameter-varying Data-enabled Predictive Control of Cold Atmospheric Pressure Plasma Jets

Abstract: Cold Atmospheric Pressure Plasma Jets (APPJs) show significant potential for biomedical applications, but their inherent complexity, characterized by nonlinear dynamics and strong sensitivity to operating conditions like tip-to-surface distance, presents considerable challenges for achieving robust and reliable real-time control. To address these issues, this paper presents the Neural Parameter-Varying Data-enabled Predictive Control (NPV-DeePC) framework. By integrating hyper neural networks (hypernets) into the neural Data-enabled Predictive Control (DeePC) paradigm, the proposed method adaptively captures system nonlinearities and parameter variations, updates the neural feature space accordingly, and enables efficient and accurate trajectory prediction and control. The NPV-DeePC framework is validated through extensive simulations involving surface temperature tracking and thermal dose delivery. The results highlight its ability to outperform existing controllers in terms of accuracy and adaptability. The computational efficiency of the NPV-DeePC approach makes it a viable candidate for real-time applications. These findings underscore its potential to advance the safe and precise control of APPJs and provide a scalable solution for other parameter-varying nonlinear systems.