Adaptive nonlinear control with constrained parallel parameter estimates
Introduction
A basic and desirable property for any adaptive control design is to guarantee converging parameter estimation errors, when the parameter vector is identifiable either in certain operating conditions or under persistently exciting exogenous signals. In these cases exponential stability and, consequently, robustness for the closed loop system hold; in addition the performance of the known parameter controller can be recovered. Hence, it is preferable to parametrize the adaptive control in terms of the minimum number of parameter estimates so that overparametrization is avoided. There are, however, cases in which an overparametrized control has some advantages [1], [2]. A second order example in parametric strict-feedback form is presented in [1] showing that the overparametrized adaptive control proposed in [3] gives better transient performance than the non-overparametrized control in [4]. In [2] an alternative adaptive control employing overparametrization is presented which has the advantage of improved performance with respect to the tuning functions approach in [5]. The overparametrization problem will be addressed for feedback linearizable nonlinear systems, depending linearly on an unknown parameter vector. They have been widely studied (see [6], [7]) with the aim of designing adaptive controls under matching [8] or extended matching [4] conditions. Adaptive controls for the more general case of systems in parametric strict-feedback form were designed in [3] via adaptive backstepping and overparametrization. Tuning functions were introduced in [5] to avoid overparametrization. A robust adaptive design is proposed in [9] following [5]: it allows for parameter projections and improved transient performance at the expense of introducing nonlinear high-gain damping feedback terms. In the presence of unknown nonlinear functions, smooth projections and sliding mode controls are used in [10] to design adaptive robust controls. Additive disturbances are dealt with in [11] by incorporating robust control design techniques which guarantee asymptotic reference tracking, provided that the control gains are sufficiently large with respect to the system initial conditions. Dynamic uncertainties are addressed in [12] by robust modifications. Actuator saturations are considered in [13] for bounded input-bounded state systems. If the nonlinearities are either unknown or not linearly parametrizable, robust approaches such as [14] may be followed. However, the robust approaches in [10], [12], [14] do not guarantee asymptotic reference tracking.
In this paper it is shown how the adaptive backstepping technique in [3], which requires an overparametrization, can be simply equipped by constrained parallel estimates so that the same stability results presented in [5] via the tuning functions design can be reobtained. The key idea is to constrain the parallel estimates differences to tend asymptotically to zero. This alternative approach has the advantage of allowing for simple parameter projection algorithms as in [15], without resorting to complex smooth projection operators (see [10], [11], [16], [17]), when a priori parameter bounds are available. However, it has the drawback of requiring parallel estimates for each unknown parameter. The same third order system reported in [5] is considered in Section 3 as a comparative example showing that the proposed constrained parallel estimation approach is simpler to design. Numerical simulations are carried out with a nonlinearity consisting of a quadratic term. Better transient performance are obtained with respect to the controller designed following [5], since the tuning functions approach leads to high frequency oscillations and higher amplitude of the control signal. A simpler second order example illustrates that the extended matching approach in [4] leads to a controller amplitude which is one order of magnitude larger than the amplitude of the controller proposed in this paper. Finally, a third order linear system is considered for which the approaches followed in [5] and in this paper lead to linear closed loop dynamics whose eigenvalues can be easily compared.
Section snippets
Main results
Consider the nonlinear system in parametric strict-feedback form in which is the state vector, is the vector of unknown parameters, , , , are known smooth nonlinear functions with , . Let be an equilibrium point (around which the closed loop system must be stabilized) given by
Examples
In this section three examples are discussed and numerically simulated to illustrate the advantages of the proposed technique in comparison with existing adaptive control algorithms.
Example 1 Consider the third-order nonlinear system (see [5]) which must be stabilized around the equilibrium point We set , . The control algorithm (3) in this case becomes
Conclusions
A new approach involving constrained parallel parameter estimators has been proposed to design adaptive controls for nonlinear systems in parametric strict-feedback form. The differences between the multiple parallel parameter estimators are constrained to tend asymptotically to zero. An alternative to the tuning functions or the extended matching designs has been obtained which guarantees the same closed loop stability properties, with the advantages of allowing for a simpler design procedure
CRediT authorship contribution statement
Patrizio Tomei: Conceptualization, Methodology, Software, Writing - review & editing. Riccardo Marino: Conceptualization, Methodology, Data curation, Writing - original draft, Writing - review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (18)
- et al.
Overparameterised adaptive controllers can reduce non-singular costs
Systems Control Lett.
(2003) - et al.
Nonlinear adaptive control of systems in feedback form: an alternative to adaptive backstepping
Systems Control Lett.
(2008) - et al.
An extended direct scheme for robust adaptive nonlinear control
Automatica
(1991) - et al.
Adaptive nonlinear control without overparametrization
Systems Control Lett.
(1992) - et al.
Adaptive robust control of SISO nonlinear systems in a semi-strict feedback form
Automatica
(1997) - et al.
Design of robust adaptive controllers for nonlinear systems with dynamic uncertainties
Automatica
(1998) - et al.
A low-complexity global approximation-free control scheme with prescribed performance for unknown pure feedback systems
Automatica
(2014) - et al.
A novel parameter projection mechanism for smooth and stable adaptive control
Systems Control Lett.
(2005) - et al.
Systematic design of adaptive controllers for feedback linearizable systems
IEEE Trans. Automat. Control
(1991)
Cited by (4)
Parameter Convergence in Adaptive Control in Presence of Unmatched Uncertainty
2022, IFAC-PapersOnLinePI-generalizing saturated repetitive learning control for a class of nonlinear uncertain systems: Robustness w.r.t. relative degree zero or one
2022, Systems and Control LettersCitation Excerpt :A rather interesting problem is indeed represented by the output tracking control problem for uncertain nonlinear systems. When uncertainties satisfy structural conditions, such as matching, extended matching or triangularity, and are linearly parameterized, adaptive state or output feedback controls (including adaptive linearizing controls) guarantee asymptotic tracking of smooth bounded reference signals (see for instance [2,3] and references therein). Exponential convergence and related robustness properties are actually achieved, as soon as persistency of excitation (PE) conditions are satisfied (see [4,5]), with PE even allowing for the exponential estimation of the uncertain constant parameters.
A Bounded e-Modification Applied to Adaptive Control of Servo Systems
2023, 9th 2023 International Conference on Control, Decision and Information Technologies, CoDIT 2023