A sampled normal form for feedback linearization

From MaRDI portal





Analytic, single input, \(n\)-dimensional affine control systems \[ dz/dt=f(z)+g(z)u,\tag{*} \] are considered whose \(r\)-dimensional subsystem is feedback linearizable after applying a suitably chosen linearizable output (so the relative degree of the system becomes equal to \(r\)). The main result proved in this paper is that the continuous system (*) subject to a sampling possesses a linearizing output, dependent on the sampling period \(\delta\), such that the relative degree of the sampled system is preserved. More specifically, it has been proved that if the system (*) has a relative degree \(r\) with respect to an output function \(y(t)\), then this degree is preserved under sampling the system (*) up to order \(r\) in \(\delta\), provided that a new output function is defined as \[ y^\delta=\sum^{r-1}_{j=0} \delta^j\gamma_r^jy^{(j)}, \] for appropriately defined coefficients \(\gamma^j_r\). Furthermore, a Nonlinear Sampled Normal Form \[ \zeta_a(k+1)=M\xi_a(k)+N(f_a(\xi(k))+g_a(\xi(k))u(k))+O(r,\delta), \] \[ \xi_b(k+1)=\xi_b(k)+\delta f_b(\xi(k))+O(\delta^2), \] \[ y^\delta=\xi_1(k), \] where \(M\), \(N\) denote some matrices, is introduced to which the sampled system with output can be transformed in a well defined approximate sense using a \(\delta\)-dependent coordinate change. This normal form can further be partially linearized to an arbitrary order of approximation by a \(\delta\)-dependent digital feedback, preserving stability of the zero-dynamics in the first approximation.











This page was built for publication: A sampled normal form for feedback linearization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2563994)