Learning Consensus for Multi-Agent Systems through Incremental Adaptive Neural Network Mechanism

In this article, we explore the adaptive learning approach of consensus protocols for multi-agent systems, employing an incremental adaptive neural network (NN) mechanism. Initially, we design an appropriate distributed control protocol utilizing a distributed error tracking approach. Furthermore, we introduce an integral Lyapunov function, which is instrumental in adeptly addressing the unknown gains associated with the system’s state. Subsequently, we propose an incremental adaptive mechanism for the adaptive estimation of parameters. This novel approach circumvents the need for numerical integration. Concurrently, we address the issue of complexity explosion by integrating a command filter technique within the consensus protocol design. It is proved that the novel control protocol not only effectively enables the tracking performance of the system but also ensures every signal in the closed-loop system remain bounded. Ultimately, efficacy of the adaptive learning control (ALC) method is substantiated through numerical example.