Input-to-state stability of PD-controlled robotic systems (Q1301442)

The author considers external stability for nonlinear systems of the form \[ \dot x=f(x,u),\quad y=h(x) \] with state \(x(t)\) in \(\mathbb{R}^n\), input \(u(t)\) in \(\mathbb{R}^m\), and output \(y(t)\) in \(\mathbb{R}^p\). In systems theory, it is important to be able to estimate the state variables in terms of past inputs. For stable linear systems, it is well known how to separate the effects of inputs from those of initial conditions. For nonlinear systems, Sontag introduced an analogous property to the free evolution/forced evolution concept used in linear systems. Sontag's formulation is called input-to-state stability (ISS). It is known that ISS is equivalent to the dissipation inequality \[ {\partial V\over\partial x}f(x,u)\leq- \alpha\bigl(|x|\bigr)+ \sigma\bigl( |u|\bigr) \] for an appropriate Lyapunov function \(V\) and comparison functions \(\alpha\) and \(\sigma\). The author's main result is to provide a sufficient condition for ISS when the inequality above is relaxed to \[ {\partial V\over\partial x}f(x,u)\leq- \alpha\bigl(|y|\bigr)+ \sigma\bigl(|u|\bigr). \] The additional requirement is the notion of detectability called input-output-to-state stability (IOSS) introduced by Sontag and Wang. The author illustrates his results with an example of mechanical manipulators -- specifically, of proportional/derivative (PD)-controlled robotic systems.