Section: New Results
Stability and uncertain dynamics: new results
Let us list here our new results about stability and stabilization of control systems, on the properties of systems with uncertain dynamics.

In an open channel, a hydraulic jump is an abrupt transition between a torrential (supercritical) flow and a fluvial (subcritical) flow. In [9] hydraulic jumps are represented by discontinuous shock solutions of hyperbolic SaintVenant equations. Using a Lyapunov approach, we prove that we can stabilize the state of the system in ${H}^{2}$norm as well as the hydraulic jump location, with simple feedback boundary controls and an arbitrary decay rate, by appropriately choosing the gains of the feedback boundary controls.

In [10], we study the exponential stabilization of a shock steady state for the inviscid Burgers equation on a bounded interval. Our analysis relies on the construction of an explicit strict control Lyapunov function. We prove that by appropriately choosing the feedback boundary conditions, we can stabilize the state as well as the shock location to the desired steady state in ${H}^{2}$norm, with an arbitrary decay rate.

We develop in [19] a method ensuring robustness properties to bangbang strategies , for general nonlinear control systems. Our main idea is to add bang arcs in the form of needlelike variations of the control. With such bangbang controls having additional degrees of freedom, steering the control system to some given target amounts to solving an overdetermined nonlinear shooting problem, what we do by developing a leastsquare approach. In turn, we design a criterion to measure the quality of robustness of the bangbang strategy, based on the singular values of the endpoint mapping, and which we optimize. Our approach thus shows that redundancy implies robustness, and we show how to achieve some compromises in practice, by applying it to the attitude control of a 3d rigid body.

Partial stability characterizes dynamical systems for which only a part of the state variables exhibits a stable behavior. In his book on partial stability, Vorotnikov proposed a sufficient condition to establish this property through a Lyapunovlike function whose total derivative is upperbounded by a negative definite function involving only the substate of interest. In [20], we show with a simple twodimensional system that this statement is wrong in general. More precisely, we show that the convergence rate of the relevant state variables may not be uniform in the initial state. We also discuss the impact of this lack of uniformity on the connected issue of robustness with respect to exogenous disturbances.

The paper [21] elaborates control strategies to prevent clustering effects in opinion formation models. This is the exact opposite of numerous situations encountered in the literature where, on the contrary, one seeks controls promoting consensus. In order to promote declustering, instead of using the classical variance that does not capture well the phenomenon of dispersion, we introduce an entropytype functional that is adapted to measuring pairwise distances between agents. We then focus on a HegselmannKrausetype system and design declustering sparse controls both in finitedimensional and kinetic models. We provide general conditions characterizing whether clustering can be avoided as function of the initial data. Such results include the description of black holes (where complete collapse to consensus is not avoidable), safety zones (where the control can keep the system far from clustering), basins of attraction (attractive zones around the clustering set) and collapse prevention (when convergence to the clustering set can be avoided).

The goal of [23] is to compute a boundary control of reactiondiffusion partial differential equation. The boundary control is subject to a constant delay, whereas the equation may be unstable without any control. For this system equivalent to a parabolic equation coupled with a transport equation, a predictionbased control is explicitly computed. To do that we decompose the infinitedimensional system into two parts: one finitedimensional unstable part, and one stable infinitedimensional part. An finitedimensional delay controller is computed for the unstable part, and it is shown that this controller succeeds in stabilizing the whole partial differential equation. The proof is based on a an explicit form of the classical Artstein transformation, and an appropriate Lyapunov function. A numerical simulation illustrate the constructive design method.

Given a linear control system in a Hilbert space with a bounded control operator, we establish in [26] a characterization of exponential stabilizability in terms of an observability inequality. Such dual characterizations are well known for exact (null) controllability. Our approach exploits classical Fenchel duality arguments and, in turn, leads to characterizations in terms of observability inequalities of approximately null controllability and of $\alpha $null controllability. We comment on the relationships between those various concepts, at the light of the observability inequalities that characterize them.

In [37] we propose an extension of the theory of control sets to the case of inputs satisfying a dwelltime constraint. Although the class of such inputs is not closed under concatenation, we propose a suitably modified definition of control sets that allows to recover some important properties known in the concatenable case. In particular we apply the control set construction to dwelltime linear switched systems, characterizing their maximal Lyapunov exponent looking only at trajectories whose angular component is periodic. We also use such a construction to characterize supports of invariant measures for random switched systems with dwelltime constraints.

In [41] we study asymptotic stability of continuoustime systems with modedependent guaranteed dwell time. These systems are reformulated as special cases of a general class of mixed (discretecontinuous) linear switching systems on graphs, in which some modes correspond to discrete actions and some others correspond to continuoustime evolutions. Each discrete action has its own positive weight which accounts for its timeduration. We develop a theory of stability for the mixed systems; in particular, we prove the existence of an invariant Lyapunov norm for mixed systems on graphs and study its structure in various cases, including discretetime systems for which discrete actions have inhomogeneous time durations. This allows us to adapt recent methods for the joint spectral radius computation (Gripenberg's algorithm and the Invariant Polytope Algorithm) to compute the Lyapunov exponent of mixed systems on graphs.

Given a discretetime linear switched system associated with a finite set of matrices, we consider the measures of its asymptotic behavior given by, on the one hand, its deterministic joint spectral radius and, on the other hand, its probabilistic joint spectral radius for Markov random switching signals with given transition matrix and corresponding invariant probability. In [42], we investigate the cases of equality between the two measures.

In [45] we address the question of the exponential stability for the ${C}^{1}$ norm of general 1D quasilinear systems with source terms under boundary conditions. To reach this aim, we introduce the notion of basic ${C}^{1}$ Lyapunov functions, a generic kind of exponentially decreasing function whose existence ensures the exponential stability of the system for the ${C}^{1}$ norm. We show that the existence of a basic ${C}^{1}$ Lyapunov function is subject to two conditions: an interior condition, intrinsic to the system, and a condition on the boundary controls. We give explicit sufficient interior and boundary conditions such that the system is exponentially stable for the ${C}^{1}$ norm and we show that the interior condition is also necessary to the existence of a basic ${C}^{1}$ Lyapunov function. Finally, we show that the results conducted in this article are also true under the same conditions for the exponential stability in the ${C}^{p}$ norm, for any $p\ge 1$.

In [46] we study the exponential stability for the ${C}^{1}$ norm of general $2\times 2$ 1D quasilinear hyperbolic systems with source terms and boundary controls. When the eigenvalues of the system have the same sign, any nonuniform steadystate can be stabilized using boundary feedbacks that only depend on measurements at the boundaries and we give explicit conditions on the gain of the feedback. In other cases, we exhibit a simple numerical criterion for the existence of basic ${C}^{1}$ Lyapunov function, a natural candidate for a Lyapunov function to ensure exponential stability for the ${C}^{1}$ norm.

In [47] we study the exponential stability in the ${H}^{2}$ norm of the nonlinear SaintVenant (or shallow water) equations with arbitrary friction and slope using a single ProportionalIntegral (PI) control at one end of the channel. Using a local dissipative entropy we find a simple and explicit condition on the gain the PI control to ensure the exponential stability of any steadystates. This condition is independent of the slope, the friction, the length of the river, the inflow disturbance and, more surprisingly, the steadystate considered. When the inflow disturbance is timedependent and no steadystate exist, we still have the InputtoState stability of the system, and we show that changing slightly the PI control enables to recover the exponential stability of slowly varying trajectories.

In [48], we address the problem of the exponential stability of densityvelocity systems with boundary conditions. Densityvelocity systems are omnipresent in physics as they encompass all systems that consist in a flux conservation and a momentum equation. In this paper we show that any such system can be stabilized exponentially quickly in the ${H}^{2}$ norm using simple local feedbacks, provided a condition on the source term which holds for most physical systems, even when it is not dissipative. Besides, the feedback laws obtained only depends on the target values at the boundaries, which implies that they do not depend on the expression of the source term or the force applied on the system and makes them very easy to implement in practice and robust to model errors. For instance, for a river modeled by SaintVenant equations this means that the feedback laws do not require any information on the friction model, the slope or the shape of the channel considered. This feat is obtained by showing the existence of a basic ${H}^{2}$ Lyapunov functions and we apply it to numerous systems: the general SaintVenant equations, the isentropic Euler equations, the motion of water in rigidpipe, the osmosis phenomenon, etc.

The general context of [56] is the feedback control of an infinitedimensional system so that the closedloop system satisfies a fadingmemory property and achieves the setpoint tracking of a given reference signal. More specifically , this paper is concerned with the Proportional Integral (PI) regulation control of the left Neumann trace of a onedimensional reactiondiffusion equation with a delayed right Dirichlet boundary control. In this setting, the studied reactiondiffusion equation might be either openloop stable or unstable. The proposed control strategy goes as follows. First, a finitedimensional truncated model that captures the unstable dynamics of the original infinitedimensional system is obtained via spectral decomposition. The truncated model is then augmented by an integral component on the tracking error of the left Neumann trace. After resorting to the Artstein transformation to handle the control input delay, the PI controller is designed by pole shifting. Stability of the resulting closedloop infinitedimensional system, consisting of the original reactiondiffusion equation with the PI controller, is then established thanks to an adequate Lyapunov function. In the case of a timevarying reference input and a timevarying distributed disturbance, our stability result takes the form of an exponential InputtoState Stability (ISS) estimate with fading memory. Finally, another exponential ISS estimate with fading memory is established for the tracking performance of the reference signal by the system output. In particular, these results assess the setpoint regulation of the left Neumann trace in the presence of distributed perturbations that converge to a steadystate value and with a timederivative that converges to zero. Numerical simulations are carried out to illustrate the efficiency of our control strategy.

There exist many ways to stabilize an infinitedimensional linear autonomous control systems when it is possible. Anyway, finding an exponentially stabilizing feedback control that is as simple as possible may be a challenge. The Riccati theory provides a nice feedback control but may be computationally demanding when considering a discretization scheme. Proper Orthogonal Decomposition (POD) offers a popular way to reduce largedimensional systems. In [59], we establish that, under appropriate spectral assumptions, an exponentially stabilizing feedback Riccati control designed from a POD finitedimensional approximation of the system stabilizes as well the infinitedimensional control system.

In [60] we consider a 1D linear transport equation on the interval $(0,L)$, with an internal scalar control. We prove that if the system is controllable in a periodic Sobolev space of order greater than 1, then the system can be stabilized in finite time, and we give an explicit feedback law.

In [61] we use the backstepping method to study the stabilization of a 1D linear transport equation on the interval $(0,L)$, by controlling the scalar amplitude of a piecewise regular function of the space variable in the source term. We prove that if the system is controllable in a periodic Sobolev space of order greater than 1, then the system can be stabilized exponentially in that space and, for any given decay rate, we give an explicit feedback law that achieves that decay rate.