Adaptive fuzzy-wavelet neural network identification core for reinforced control of general arbitrarily switched nonlinear multi input-multi output dynamic systems
MR Homaeinezhad, S Yaqubi - Applied Soft Computing, 2020 - Elsevier
Applied Soft Computing, 2020•Elsevier
In control of switched systems with undetectable switching signals, robustness and precision
are on two different sides of a spectrum. In one hand, robustness to significant modeling
uncertainties arising from undetectability of switching modes can be attained by designing
the control scheme for worst-case switching configurations. On the other hand, such a
control scheme would potentially be overly conservative and imprecise. A natural solution to
this problem is pinpointing active switched dynamics at any given moment. However, this is …
are on two different sides of a spectrum. In one hand, robustness to significant modeling
uncertainties arising from undetectability of switching modes can be attained by designing
the control scheme for worst-case switching configurations. On the other hand, such a
control scheme would potentially be overly conservative and imprecise. A natural solution to
this problem is pinpointing active switched dynamics at any given moment. However, this is …
Abstract
In control of switched systems with undetectable switching signals, robustness and precision are on two different sides of a spectrum. In one hand, robustness to significant modeling uncertainties arising from undetectability of switching modes can be attained by designing the control scheme for worst-case switching configurations. On the other hand, such a control scheme would potentially be overly conservative and imprecise. A natural solution to this problem is pinpointing active switched dynamics at any given moment. However, this is no trivial task. In this study, we propose that the aforementioned problems can be overcome by designing a control scheme that prioritizes appropriate objectives according to operating conditions. In other words, the control scheme adjusts itself such that either of robustness to unknowable switching or increased tracking precision is selected as the primary control objective. As a result, the control model can be considered as dual-mode featuring a safe control mode and a precise control mode. In precise control mode, a model generation scheme using a modified Fuzzy-Wavelet Neural Network (FWNN) for Multi Input-Multi Output (MIMO) systems is incorporated for precise estimation of active dynamics which potentially features unknowable switching dynamics, external disturbances and parametric modeling uncertainty. However, this approximate model cannot be used immediately since convergence of the FWNN-based model to actual system dynamics takes place after a limited interval. In such periods (which often correspond to discontinuities in switching dynamics and references), using the FWNN scheme is perilous. Therefore, a robust discrete-time sliding mode control (DSMC) is used to ensure stabilization of closed-loop system in all potential modes of switched dynamics at the cost of reduced tracking precision. In combination, the dual-mode scheme ensures robust stabilization in safe control modes corresponding to transient-state stage and accurate tracking in steady-state stage of system response based on the proposed precise mode scheme. Numerical and experimental examples highlight the key features and improvements of the presented control algorithm.
Elsevier
以上显示的是最相近的搜索结果。 查看全部搜索结果