Security Output Byzantine-Optimization of Heterogeneous Multiagent Systems via Adaptive Self-Triggered Detection Approach

Secure and communication-efficient distributed optimization has become increasingly important for multiagent systems (MASs) operating in the Internet of Things (IoT) environments, where open communication environments and compromised devices may severely degrade networked coordination. This paper investigates resilient output optimization for continuous-time linear heterogeneous MASs subject to Byzantine attacks. An edge-based adaptive self-triggered framework is developed to enable each agent to validate the information received from its neighbors while reducing unnecessary communication updates. Based on trigger-error thresholds, a Byzantine attack detection and isolation mechanism is proposed, under which suspicious neighbors are locally removed once abnormal behavior is identified. To cope with the hidden topology variations caused by such local isolation decisions, an adaptive distributed optimization algorithm is further designed to preserve the optimization performance of normal agents. It is shown that all normal agents can still converge to the global optimal value after Byzantine agents are completely isolated.