Abstract
Sliding-mode extremum seeking (SMES) control has been widely used for real-time optimization of perturbed nonlinear systems. However, the baseline SMES schemes often experience degraded accuracy, slow transient response, and even divergence when perturbations exceed certain levels. This motivates a deeper examination of their robustness properties. This article provides a deep analysis of the baseline SMES architecture, with particular focus on the coupling between practical controllers and the extremum-seeking loop. The effects of residual tracking errors from the practical controller on both stability and performance of the baseline SMES are thoroughly investigated, revealing critical limitations under perturbations. To this end, based on this analysis, a robust extremum-seeking framework is proposed by combining barrier function-based integral sliding mode control (BF-ISMC) with a modified SMES. The disturbance rejection capabilities of the BF-ISMC enhanced the robustness of this approach and alleviated the performance limits imposed by perturbations, without requiring an estimate of their upper bound. Simultaneously, the SMES is enhanced through two additional mechanisms: (i) a damping term that introduces an exponential attenuation of steady-state oscillations, and (ii) a recovery mechanism that accelerates convergence following abrupt shifts in the unknown optimal operating point. These modifications jointly improve robustness, accuracy, and transient performance. The proposed method is applied to an anti-lock braking system (ABS), and simulation results demonstrate superior performance relative to a recent benchmark approach, particularly under significant perturbations and sudden variations in optimal conditions. The findings confirm the effectiveness of the proposed approach for perturbed nonlinear systems.