Anti-Disturbance Proximal Neural Networks for Composite Resource Allocation.

Abstract

Composite resource allocation problems with general constraints are studied in this article, which frequently arise in networked systems such as smart grids and multiagent coordination. To address the challenges posed by multivalued differential inclusions resulting from nonsmooth objective functions, two anti-disturbance proximal neural networks are proposed, each tailored to handle structured and unstructured disturbances, respectively. For structured disturbances, the neural network is developed based on the internal model principle, which exploits the underlying structure of the known dynamics. To further improve resilience against unstructured disturbances, another observer-based neural network is designed. The asymptotic convergences of both neural networks are rigorously established using Lyapunov stability theory. Finally, numerical simulations validate the effectiveness and robustness of the proposed neural networks under different types of disturbances.