Fixed-order Linear Parameter Varying Controller Design for a 2DOF Gyroscope

This paper presents an approach for fixed-order Linear Parameter Varying (LPV) controller design with application to a 2 Degree-of-Freedom (2DOF) gyroscope experimental setup. Inner convex approximation of the non-convex set of all stabilizing fixed-order LPV controllers is characterized through a set of Linear Matrix Inequalities (LMIs). This is achieved through the use of two slack matrices which enable decoupling of the controller and Lyapunov matrix parameters in the derivative of Lyapunov function. The LPV model obtained by the approximation of the nonlinear model of the 2DOF gyroscope is used for the design of a second- order LPV controller. Experimental results show good tracking performance in the presence of scheduling parameter variations.

From Fixed-Order Gain-Scheduling to Fixed-Structure LPV Controller Design

Zlatko Emedi

This thesis focuses on the development of some fixed-order controller design methods in the gain-scheduling/Linear Parameter Varying (LPV) framework. Gain-scheduled controllers designed using frequency-domain Single Input Single Output (SISO) models are considered first, followed by LPV controller design in the SISO transfer function setting and, finally, by Multiple Input Multiple Output (MIMO) LPV controller design in the state-space setting. In addition to the guarantee of closed-loop stability, each of the methods optimizes some classical performance measure, such as the

\mathscr{H}_\infty

\mathscr{H}_2

performance metrics. In the LPV state-space setting, the practical assumption of bounded scheduling parameter variations is taken into account in order to allow a higher performance level to be achieved. The fixed-order gain-scheduled controller design method is based on frequency-domain models dependent on the scheduling parameters. Based on the linearly parameterized gain-scheduled controllers and desired open-loop transfer functions, the

\mathscr{H}_\infty

performance of the weighted closed-loop transfer functions is presented in the Nyquist diagram as a set of convex constraints. No a posteriori interpolation is needed, so the stability and performance level are guaranteed for all values of scheduling parameters considered in the design. Controllers designed with this method are successfully applied to the international benchmark in adaptive regulation. These low-order controllers ensure good rejection of the multisinusoidal disturbance with time-varying frequencies on the active suspension testbed. One issue related to the gain-scheduled controller design using the frequency response model is the computational burden due to the constraint sampling in the frequency domain. The other is a guarantee of stability and performance for all the values of scheduling parameters, not just those treated in design. To overcome these issues, a method for the design of fixed-order LPV controllers with the transfer function representation is proposed. The LPV controller parameterization considered in this approach leads to design variables in both the numerator and denominator of the controller. Stability and

\mathscr{H}_\infty

performance conditions for all fixed values of scheduling parameters are presented in terms of Linear Matrix Inequalities (LMIs). With a problem of rejection of a multisinusoidal disturbance with time-varying frequencies in mind, LPV controller is designed for an LTI plant with a transfer function model. The extension of these methods from SISO to MIMO systems is far from trivial. The state-space setting is used for this reason, as there the transition from SISO to MIMO systems is natural. A method for fixed-order output-feedback LPV controller design for continuous-time state-space LPV plants with affine dependence on scheduling parameters is proposed. Bounds on the scheduling parameters and their variation rates are exploited in design through the use of affine Parameter Dependent Lyapunov Functions (PDLFs). The exponential decay rate, induced

\mathscr{L}_2

-norm and

\mathscr{H}_2

performance constraints are expressed through a set of LMIs. The proposed method is applied to the 2DOF gyroscope experimental setup. In practice control is performed using digital computers, so some effort needs to be put into the LPV controller discretization. If the discrete-time LPV model of the system is available [...]

EPFL2014

Fixed-Order H-infinity Controller Design via Convex Optimization Using an Alternative to Youla Prameterization

Alireza Karimi, Hamid Khatibi

All H-infinity controllers of a SISO LTI system are parameterized thanks to the relation between Bounded Real Lemma and Positive Real Lemma and a new concept of strict positive realness of two transfer functions with the same Lyapunov matrix in the matrix inequality of the Kalman-Yakubovic-Popov lemma. This new parameterization shares the same features with Youla parameterization, namely on the convexity of H-infinity norm constraints for the closed-loop transfer functions. However, in contrary to Youla parameterization, it can deal with any controller order and any controller structure such as e.g. PID. The main feature of the proposed method is that it can be extended easily for the systems with polytopic uncertainty. This way, a convex inner approximation of all H-infinity controllers for polytopic systems is given, which can be enlarged by increasing the controller order. In order to design a low-order robust H-infinity controller with less conservatism, rank of the k-th Sylvester resultant matrix of the controller is made to be deficient via a convex approximation of the rank minimization problem. The effectiveness of the proposed method is shown via simulation results.

Institute of Electrical and Electronics Engineers2010

From Fixed-Order Gain-Scheduling to Fixed-Structure LPV Controller Design

Zlatko Emedi

\mathscr{H}_\infty

\mathscr{H}_2

\mathscr{H}_\infty

\mathscr{H}_\infty

\mathscr{L}_2

-norm and

\mathscr{H}_2

EPFL2014