Nonlinear dynamics in adaptive control: Chaotic and periodic stabilization (Q1821071)

The nonlinear features of adaptive control algorithms are highlighted. A specific discrete-time example illustrates that in the presence of undermodelling errors, nonlinear phenomena in the feedback gain arise. The outline of contents is as follows: Section 1 contains an introduction. In Section 2 an explicitly nonlinear difference equation, which describes the complete closed-loop adaptive system, is derived. The dynamics of this difference equation fall into three distinct sets depending upon a single parameter b, measuring the unmodelled dynamics. Section 3 describes the properties of the adaptive scheme when \(b=0\) (no modelling errors). Section 4 considers the case \(b<0\) and proves that the feedback gain is asymptotically periodic. For a range of values of negative b this periodic gain stabilizes the plant. In Section 5 positive b is considered and the feedback gain is shown to be chaotic, again stability the plant for a range of values of b. Section 6 deals with the effects on the closed-loop stabilizing of the plant due to these different feedback gains and considers questions of performance and robustness of the adaptive control scheme. Section 7 concludes and draws together the threads of the previous sections to discuss the implication of the results of the paper.