Event-based fuzzy neural network distributed optimization for complex multiagent networks with unknown directions

Abstract

This article focuses on the event-based adaptive distributed optimization problem of uncertain complex multiagent networks subject to unmodeled nonlinearities, uncertain disturbances, and unknown interaction direction. The measured gradient functions depending on the real output information of multi-agent systems (MASs) are employed, rather than explicit gradient information of the local objective functions. To reduce continuous communication and update requirements, an adaptive dynamic event-triggered strategy is applied. Under these multiple constraints, achieving asymptotic optimization in MASs presents a significant challenge. To this end, a novel event-based distributed optimization scheme is developed, incorporating a feedback tracking term, a neural network estimation term, and a disturbance compensating term. The fuzzy neural networks are used to approximate the nonlinear functions. Based on this framework, a novel adaptive distributed optimization controller is constructed to ensure optimal consensus, whose stability and convergence are rigorously established via Lyapunov-based analysis. Furthermore, the proposed neuroadaptive distributed optimization algorithm is extended to the practical classes of systems with unknown virtual control coefficients. Lastly, the proposed neuroadaptive distributed optimization scheme is illustrated using two examples.