Multi-dimensional Taylor Network-Based Adaptive Output Feedback Control for Stochastic Nonlinear Systems with Input Delay

The article delves into the critical challenges posed by stochastic uncertainties and input delays in nonlinear systems, which can lead to performance degradation or instability. It highlights the limitations of existing control schemes, particularly when state variables are unmeasured or partially known. The core innovation presented is an adaptive output feedback control method based on multi-dimensional Taylor networks (MTN), which offers a simpler and more efficient structure compared to traditional neural networks. The article meticulously details the design of an MTN-based observer to estimate unmeasured states and an auxiliary system to mitigate the effects of input delays. Through the application of the backstepping technique and Lyapunov stability theory, the proposed controller ensures boundedness in probability of all closed-loop signals and successful tracking of the reference signal. Simulation results demonstrate the effectiveness of the method, showcasing its superior performance in handling stochastic disturbances, input delays, and unmeasured states. The article also compares the MTN-based approach with radial basis function neural networks (RBFNN), emphasizing the advantages of MTN in terms of real-time performance and numerical stability.