Dynamic gain-scheduled control and extended linearisation: extensions, explicit formulae and stability

Abstract

A controller designed for linearisations at various trim/operating points of a nonlinear system using linear approaches is not necessarily performing well or stable once scheduled with a state under dynamic conditions; the key idea of using this scheduled control law design is to retain states close to the current, usually dynamically varying, operating point. Dynamic gain scheduling (DGS) is a technique aimed to resolve this controller scheduling issue for rapidly changing dynamics and states. It entails scheduling the control law gains with a fast-varying state variable rather than with a slowly varying state. It has been successfully applied to the aircraft system models, allowing also for nested loops. The aim of this paper is to extend DGS to a more general setting allowing for more complex multi-input dynamics incorporating the more simple static scheduling within the same controller. Hence, given a linear design, suitable transformations are provided which allow fast scheduling of multi-variable controllers. Important parallels to the approach of extended linearisation are drawn. Theoretical results are shown, providing explicit formulae related to nonlinear dynamic inversion control. Several examples of varying nonlinear complexity are presented in order to emphasise the characteristics of the approach.

Transient Dynamics Assessment of a Gain-Scheduled Aircraft Controller Using Nonlinear Frequency Approach

“…Dynamic gain scheduling in this case involves determining the gains using eigenstructure assignment (effectively pole placement in this model) and scheduling them against the fast-varying states rather than the slow-varying ones or the input while accounting for the effect of this on the local stability (Jacobian matrix). This novel method of nonlinear control design shows superior performance to conventional input gain scheduling [19][20][21][22], and is further examined here to ensure that our proposed method does not generate false-positive results (i.e. indicating a problem when there is none).…”