Adaptive Fixed-Time Control for an Uncertain Robot With Input Quantization: A Broad Learning System Approach

Abstract

In this paper, an adaptive fixed-time control approach is designed for a robot with dynamic uncertainty in the presence of input quantization by using the Broad learning system (BLS). The proposed BLS-based control algorithm is constructed by fusing the BLS with the radial basis function neural network, which is improved in terms of node selection rule with a self-adjusting Gaussian function center and enhancement layer. A hysteresis quantizer is applied to the requirement of a low transmission rate. For the nonlinearity occurring in the quantized input, a novel adaptive fixed-time method is developed such that 1) the adverse effect of quantization nonlinearity is removed in a finite interval; 2) the BLS-based approximation technique can improve the approximation accuracy, which enhances the robustness of the closed-loop system; and 3) via the Lyapunov stability method, the fixed-time convergence of the closed-loop system is proved. Finally, numerical simulations and experiments validate the effectiveness of the proposed control scheme.