Fixed-Time Self-Structuring Neuron Network-Based Adjustable Prescribed Performance Control for a Quadrotor UAV With Input Saturation

Abstract

This paper introduces a fixed-time self-structuring neural network (SSNN)-based adjustable prescribed performance control for quadrotors designed to handle mutational external disturbances and input saturation. To address the unpredictable nature of such disturbances during missions, SSNN is proposed for limited on-board resources of the quadrotors. This kind of neural network can adaptively adjust the number of neurons in real time to achieve more precise estimations while minimizing resource consumption. Second, to overcome the singularity problem that arises when the tracking error may exceed the boundary of the envelope when a mutation disturbance occurs, an adaptively adjustable prescribed performance function is proposed to constrain the tracking error, which is always within the envelope to avoid the singularity problem. A fixed-time adaptive command filter that can estimate an unknown upper bound on the derivative of the virtual control law with respect to time is proposed and improves the convergence rate. Under Lyapunov’s theorem, it is demonstrated that the closed-loop system can converge to the origin within a fixed time, showcasing the effectiveness of the proposed control strategy as evidenced by comparative simulation results.

Keywords

• Adjustable prescribed performance
• self-structuring neuron network
• fixed-time stability
• input saturation
• quadrotor UAV