An inversion procedure for nonlinear time-varying systems

From MaRDI portal





The authors consider the time-varying nonlinear system: \[ \begin{aligned} \dot x &= f(x,t)+ g(x,t) u(x,t),\\ y &= h(x,t), \end{aligned} \] where \(u\) and \(h\) are smooth maps to \({\mathfrak R}^ m\) and \(x\in {\mathfrak R}^ n\). They give a procedure to compute a control \(u\) so that \(y=0\). If the image of the (sufficiently smooth and regular) mapping \((\xi, t)\mapsto \pi(\xi, t)\) defined by \(h(\pi (\xi,t), t)=0\) is required to be controlled invariant, then the equation in the new \(\xi\) coordinate \[ f(\pi (\xi, t), t)+ g(\pi (\xi,t), t)u= {{\partial\pi} \over {\partial\xi}} \varphi( \xi,t)+ {{\partial\pi} \over {\partial t}} \] can be solved for \(u\) (resp. \(\varphi\)) when projecting it to the subspace spanned by (resp. perpendicular to) the control vector fields. The converse is true. An algorithm that computes these invariant submanifolds is proposed. When applied to Lagrange equations of a two-link manipulator, this algorithm does not require the inversion of matrices in contrast to previous procedures. For the entire collection see [Zbl 0778.00019].











This page was built for publication: An inversion procedure for nonlinear time-varying systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3140770)