Transient performance, resetting and filtering in nonlinear multiple model adaptive control

Abstract

Multiple model adaptive control (MMAC) is a well established approach for implementing adaptive systems with fast transient response. This paper considers a recently developed MMAC method based on adaptive nonlinear backstepping control where the parameter estimate may be discontinuously reset based on a criterion that requires a negative jump in the associated control Lyapunov function. Particular attention is paid to transient effects due to data filtering, which must be introduced in any practical implementation of the MMAC algorithm in order to reduce the sensitivity to noise, disturbances and model uncertainty. The main contribution of this paper is insight into the robust behavior of the adaptive system resulting from the filtering, and an investigation into the trade-offs between high transient performance and robustness to uncertainty. We also suggest data filter tuning guidelines and illustrate the results using a simulation example.

Transient performance, estimator resetting and filtering in nonlinear multiple model adaptive backstepping control

“…The point of interest is located in the difference between the jump in the adaptive control Lyapunov function (13), and its estimated value (42) being used in the reset criterion ∆V i,est (t) < 0.…”

“…The contribution of the present paper is an analysis of the effect of such filters. Preliminary results were reported in [13].…”

Increased transient performance for the adaptive control of feedback linearizable systems using multiple models

“…Therefore the filters should be tuned carefully to address the trade-off between response time and sensitivity to uncertainties in the implementation. These issues are discussed in greater detail in a recent paper (Hansen and Johansen 2004). A simple methodology (suggested in Kalkkuhl et al (2002)) to prevent undesirable switching due to noise and uncertainties would be to include a negative threshold () in the resetting condition as,…”

Icing detection and identification for unmanned aerial vehicles using adaptive nested multiple models

“…Detection and isolation schemes using unknown input observers have been proposed for lateral dynamics and, recently, for longitudinal dynamics . In this paper, we adopt a multiple‐model (MM) approach . Defining a bank of possible models, corresponding to distinct admissible values of the icing severity factor, the proposed algorithm guarantees the identification of the closest model to the true system as well as the estimation of the icing severity factor, provided that a distinguishability condition among the models is fulfilled or, more generally, persistency of excitation is enforced.…”