Cyber-physical systems (CPS), integrating advanced computation, communication, and control technologies with the physical process, are widely applied in industry applications such as smart production and manufacturing systems, robotic and automotive control systems, and smart grids. Due to possible exposure to unreliable networks and complex physical environments, CPSs may simultaneously face multiple cyber and physical issues including cyber threats (e.g., malicious cyber attacks) and resource constraints (e.g., limited networking resources and physical constraints). As one of the essential topics in designing efficient CPSs, the controller design for CPSs, aiming to achieve secure and resource-aware control objectives under such cyber and physical issues, is very significant yet challenging. Emphasizing optimality and system constraint handling, model predictive control (MPC) is one of the most widely used control paradigms, notably famous for its successful applications in chemical process industry. However, the conventional MPC methods are not specifically tailored to tackle cyber threats and resource constraints, thus the corresponding theory and tools to design the secure and resource-aware controller are lacking and need to be developed. This dissertation focuses on developing MPC-based methodologies to address the i) secure control problem and ii) resource-aware control problem for CPSs subject to cyber threats and resource constraints. In the resource-aware control problem of CPSs, the nonlinear system with additive disturbance is considered. By using an integral-type event-triggered mechanism and an improved robustness constraint, we propose an integral-type event-triggered MPC so that smaller sampling frequency and robustness to the additive disturbance can be obtained. The sufficient conditions for guaranteeing the recursive feasibility and the closed-loop stability are established. For the secure control problem of CPSs, two aspects are considered. Firstly, to achieve the secure control objective, we design a secure dual-mode MPC framework, including a modified initial feasible set and a new positively invariant set, for constrained linear systems subject to Denial-of-Service (DoS) attacks. The exponential stability of the closed-loop system is guaranteed under several conditions. Secondly, to deal with cyber threats and take advantage of the cloud-edge computing technology, we propose a model predictive control as a secure service (MPCaaSS) framework, consisting of a double-layer controller architecture and a secure data transmission protocol, for constrained linear systems in the presence of both cyber threats and external disturbances. The rigorous recursive feasibility and robust stability conditions are established. To simultaneously address the secure and resource-aware control problems, an event-triggered robust nonlinear MPC framework is proposed, where a new robustness constraint is introduced to deal with additive disturbances, and a packet transmission strategy is designed to tackle DoS attacks. Then, an event-triggered mechanism, which accommodates DoS attacks occurring in the communication network, is proposed to reduce the communication cost for resource-constrained CPSs. The recursive feasibility and the closed-loop stability in the sense of input-to-state practical stable (ISpS) are guaranteed under the established sufficient conditions. / Graduate

The increasing concerns regarding users‘ data privacy leads to the infeasibility of distributed Machine Learning applications, which are usually data-hungry. Federated Learning has emerged as a privacy-preserving distributed machine learning paradigm, in which the client dataset is kept locally, and only the local model parameters are transmitted to the central server. However, adoption of the Federated Learning paradigm leads to new edge computing challenges, since it assumes computationally intensive tasks can be executed locally by each device. The diverse hardware resources in a population of edge devices (e.g., smartphone models) can negatively impact the performance of Federated Learning, at both the global and local levels. This thesis contributes to this context with the implementation of a hardware-aware Federated Learning platform, which provides comprehensive support regarding the impacts of hardware heterogeneity on Federated Learning performance metrics by modeling the costs associated with training tasks on aspects of computation and communication.