(702a) Keynote 1 - Resilient Multi-Agent Estimation and Control of Complex Process Networks

Cyber-physical systems, representing complex integrations between physical and digital elements, are particularly vulnerable to cyber incidents that can cause significant operational disruptions and safety hazards [1,2]. The growing complexity and nonlinearity of these systems, combined with the necessity for extensive communication, often render centralized predictive models impractical for incident detection, recovery, and control [3]. Traditional centralized approaches for process incident detection, parameter estimation, and control demand significant computational resources. Conversely, decentralized methods, where independent agents manage each subsystem, face challenges such as communication overhead, frequent synchronization, and complexity, especially in highly integrated systems [4-7].

This study proposes a novel adaptive multi-agent architecture for distributed incident detection, state recovery, and regulation. By leveraging spectral community detection, the method identifies highly interactive state variables within the system [8]. It recursively decomposes the system into a predetermined number of subsystems in a single step, differing from traditional hierarchical methods that sequentially divide subsystems to achieve maximum modularity. For incident detection, community-based long short-term memory (LSTM) models are developed using data from each community to predict the system's reference dynamics. Each LSTM model learns the normal operations of its community while considering integrated dynamics, thereby maintaining high detection accuracy while mitigating issues related to data collection, communication, and computational constraints.

A combined spectral community detection-based distributed moving horizon estimation (DMHE) and distributed model predictive control (DMPC) architecture is then utilized to further enhance the resilience of integrated process systems in the presence of cyber incidents. The DMHE estimates unmeasured state variables and recovers the falsified variables required for output feedback regulations. The proposed architecture is evaluated through simulations of an integrated benzene alkylation benchmark under various process incident scenarios. These evaluations demonstrate the proposed multi-agent architecture's closed-loop performance and computational advantages in detecting and mitigating cyber incidents, particularly those targeting temperature measurement sensors.

References:
[1] Zhe Wu, Scarlett Chen, David Rincon, and Panagiotis Christofide, Post cyber-attack state reconstruction for nonlinear processes using machine learning, Chemical Engineering Research and Design, 159, 248-261, 2020

[2] Henrique Oyama and Helen Durand. Integrated cyberattack detection and resilient control strategies using Lyapunov‐based economic model predictive control. AIChE Journal, 66(12), e17084, 2020.

[3] Brayden Sundberg and Davood B. Pourkargar. Cyberattack awareness and resiliency of integrated moving horizon estimation and model predictive control of complex process networks. In Proceedings of the American Control Conference (ACC), 3815–3820, 2023.

[4] Scarlett Chen, Zhe Wu, and Panagiotis D. Christofides. Cyber-security of centralized, decentralized, and distributed control-detector architectures for nonlinear processes. Chemical Engineering Research and Design, 165, 25-39, 2021.

[5] Dana Alsagheer, Lei Xu, and Weidong Shi. Decentralized machine learning governance: Overview, opportunities, and challenges. IEEE Access, 2023.

[6] Yuwei Sun, Hideya Ochiai, and Hiroshi Esaki. Decentralized deep learning for multi-access edge computing: A survey on communication efficiency and trustworthiness. IEEE Transactions on Artificial Intelligence, 3(6):963–972, 2021.

[7] Prashant Mhaskar, Adiwinata Gani, Nael H. El-Farra, Charles McFall, Panagiotis D. Christofides, and James F Davis. Integrated fault-detection and fault-tolerant control of process systems. AIChE Journal, 52(6):2129–2148, 2006.

[8] AmirMohammad Ebrahimi and Davood B. Pourkargar. Distributed estimation and control of process networks using adaptive community detection. In Proceedings of 12^th IFAC International Symposium on Advanced Control of Chemical Processes, 2024.