Modular Reservoir Echo State Network-Based Adaptive Consensus Control of Nonlinear Multi-Agent Systems Under Time Varying State Constraints, Delays, and Disturbances

This paper focuses on the adaptive neural network based leader follower consensus (LFC) tracking problem for a class of nonlinear multi-agent systems (MAS) with unknown dynamics, in the presence of time varying state constraints, unknown time varying delays and disturbances in a unified framework. A hybrid Modular Reservoir Echo State Network (MRESN) based adaptive control law is developed to achieve LFC with reduced tracking error and faster convergence. Here, MRESN is used to approximate the unknown agent dynamics and system uncertainties. Distinct from the existing studies, in this work, a disagreement function based backstepping-free control design is developed that utilizes a Time Varying Barrier Lyapunov Function (TVBLF) to constrain the states within prescribed bounds. The unknown time varying delays are compensated by utilizing an Exponential-type Lyapunov-Krasovskii functional (LKF) in the design process, resulting in the proposed control law that does not necessitate prior knowledge of the maximum time delay. Lyapunov analysis confirms the bound-edness of all closed-loop signals and guarantees that the system states are within defined bounds. The effectiveness and practicality of the developed control law are demonstrated through illustrative simulations and hardware-in-the-loop tests along with comparative studies against state-of-the-art controllers.