SUPPRESSION OF ATTACKS ON CONSENSUS OF MULTI AGENT SYSTEM WITH EVENT-TRIGGERED IMPULSIVE CONTROL

• Main Author: Shadrina Amalina, Nur
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/52001
• Institution: Institut Teknologi Bandung

Distributed control in multi agent system has been successfully applied in various field, one of which is to achieve group objectives in the form of consensus. In consensus problems, each node submits a value to which all nodes must agree on the value. Distributed multi agent systems are vulnerable to cyber attacks because there is not any central entity to monitor the activities of all agents. Without the need for complete knowledge of agent network dynamics or topology, an attacker can inject a state-independent into the sensor or actuator of one of the agents. The state independent given by the attacker can be an impulse signal. With external interference in the form of impulses, the system can still reach consensus, but the consensus value of the system can shift or break. Unfortunately, multi agent systems sometimes do not recognize an attack on the system, which leads to silent attacks. In this study, the multi agent system is assumed to have homogeneous dynamics for each agent, in the form of double integrators. The system can detect any attacks or events that differ from normal conditions by means of an event-triggered function. Zeno behavior is guaranteed not to occur in the designed triggered event function. In order to return the system to its consensus value, an impulsive control system is used. Necessary and sufficient conditions for gain control are obtained from convergence analysis based on the stability theory of impulsive dynamical systems and properties of the network Laplacian matrix. With the proposed impulsive control law, the multi agent system remains towards its consensus value even though the system is under sensor and/or actuator attacks.