Abstract

Over-actuated systems are crucial for advanced control functionalities in various industries, with Dynamic Positioning (DP) systems serving as a prime example. DP systems, commonly found in marine vessels and offshore platforms, maintain fixed positions using multiple thrusters. As Cyber-Physical Systems (CPS), DP systems integrate computation, communication, and physical components to achieve control objectives. However, their reliance on communication networks makes them vulnerable to False Data Injection (FDI) cyber-attacks, where adversaries can compromise signals sent from the controller to the thrusters. This thesis presents methods for secure estimation, attack reconstruction, isolation, and compensation within a centralized thrust allocation framework. Key contributions include achieving these goals without relying on strict input-output matrix conditions, and allowing thrusters to remain operational even when affected by FDI attacks. Furthermore, the thesis addresses the challenge of reducing and shifting the attack surface, particularly in over-actuated systems, where numerous communication links increase vulnerability. To mitigate this, the allocation scheme is transformed from a centralized to a decentralized framework. In this approach, control signals are sent to a randomly chosen thruster, periodically switching to prevent successful attacks. Thrusters then coordinate through a consensus network to realize the control commands, while the consensus protocol is resilient under attacks on the communication channels between the thrusters. The decentralized protocol is effective in both closed-loop and open-loop operations. Finally, the estimation and compensation techniques developed earlier in the thesis are also applied to enhance the resilience of the decentralized architecture.