Multi-innovation Stochastic Gradient Algorithms for Input Nonlinear Time-Varying Systems Based on the Line Search Strategy
- 56 Downloads
Block-oriented nonlinear systems have attracted a considerable attention for their flexible structure and practicability. This study proposes a novel multi-innovation stochastic gradient (MISG) algorithm to address the identification problem in input nonlinear systems. This involves applying the inexact line search strategy to determine an appropriate convergence factor at each recursive step. The proposed algorithm tracks the nonlinear system dynamics faster than the conventional MISG algorithm. It is therefore suitable for online identification and can be applied to nonlinear time-varying systems. The concept of auxiliary model identification is also adopted for dealing with unmeasurable variables. The effectiveness of the proposed algorithm is verified through simulated examples.
KeywordsParameter identification Multi-innovation identification Nonlinear system Line search Time-varying system
This work was supported by the National Natural Science Foundation of China (No. 61403165), the Natural Science Foundation for Colleges and Universities in Jiangsu Province (No. 16KJB120006) and the Research Fund of Jinling Institute of Technology for Advanced Talents (No. jit-b-201805).