Robust cooperative output regulation of heterogeneous Lur'e networks

Abstract

Summary In this paper, we study robust cooperative output regulation problems for a directed network of Lur'e systems that consist of a nominal linear dynamics with an unknown static nonlinearity around it through negative feedback. We assume that the linear part of each agent is identical, but the nonlinearities are allowed to be different for distinct agents. In this sense, the network is heterogeneous. As is common in the context of Lur'e systems, the unknown nonlinearities are assumed to be sector bounded within one given sector. The interconnection graph among these agents is assumed to contain a directed spanning tree. Similar to classical output regulation problems, there is a virtual exosystem generating a reference signal in which all the agents are required to track cooperatively. Our designed distributed dynamic state/output feedback protocol makes a copy of the reference signal at each agent asymptotically, and then the robust cooperative output regulation problem becomes a robust tracking problem that can be handled by each agent via local information. It turns out that our cooperative protocols are fully distributed. Sufficient conditions on the existence of output synchronization protocols are given along with some discussions on these conditions. Finally, two simulation examples illustrate our design. Copyright © 2016 John Wiley & Sons, Ltd.

PID Passivity-Based Control of Port-Hamiltonian Systems

“…where g ⊥ : R n → R (n −m )×n is a full rank, left annihilator of g(x), that is, g ⊥ (x)g(x) = 0 and rank{g ⊥ (x)} = n − m. See [35] for a parameterization of all full rank left annihilators of a matrix. We restrict our attention to PID controllers of the form…”

Practical dynamic consensus of Stuart–Landau oscillators over heterogeneous networks

“…where w and v are normalized left and right eigenvectors of the unperturbed matrix A 0 associated to λ 1 hence, |w| = |v| = 1. The statement also applies if the multiplicity of λ 1 is larger than one, provided that there exists a complete set of eigenvectors for the associated eigenspace Moro et al (1997), Wilkinson (1965). Now, for the system (17a) the Laplacian matrix L is symmetric and corresponds to a connected graph hence, it is diagonalizable and there exists a real orthogonal matrix U such that…”

“…in which the parameter ε = 1/γ may be rendered arbitrarily small by design. That is, for sufficiently large values of γ, we may use results on perturbation theory for matrices (see, e.g., Horn and Johnson (1985); Moro et al (1997)) to characterize the eigenvalues and eigenvectors of A γ in terms of ε and the eigenvalues and eigenvectors of the Laplacian L. In particular, (Moro et al, 1997, Theorem 2.1) as well as Horn and Johnson (1985), Wilkinson (1965) allow to estimate the eigenvalues of A γ in terms of those of L, M and ε. In general, a small perturbation of a generic matrix A is denoted by…”

“…In its simplest form, consensus consists in a group of linear interconnected systems synchronizing their trajectories and asymptotically reaching a common equilibrium point determined by the systems' intial conditions (Olfati-Saber & Murray, 2004;Ren et al, 2007). Furthermore, when the network is heterogeneous, that is, when it comprises systems with different dynamics and/or different parameters, it is possible that consensus is not reached, but a steady-state error may prevail -see e.g., Qin et al (2018); Steur et al (2016); Wang et al (2015); Zhang et al (2016). This may be referred to as practical consensus.…”