No label defined (Q4346147)

The tube of trajectories of the differential inclusion \[ \dot x\in A(t)x+P(t),\qquad x\in\mathbb{R}^n, \] with the initial condition \(x(t_0)\in X_0\) and the phase constraint \(x(t)\in Y(t)\), \(t_0\leq t\leq\theta\), is approximated by the tube of trajectories of a perturbed system of differential inclusions \[ \left\{\begin{aligned}\dot x\in A(t)+P(t), &\quad x(t_0)\in X_0 \\ R(t)\dot y\in -x +B(t)y+Y(t), &\quad y(t_0)\in Y_0.\end{aligned}\right. \] Here \(A(t)\) and \(B(t)\) are quadratic matrices of appropriate dimensions, \(R(t)\) is a nondegenerate matrix-valued function, \(P(t)\) and \(Y(t)\) are convex-valued multifunctions, \(X_0\) and \(Y_0\) are convex compact sets. The principal result is that under natural continuity assumptions imposed on \(A(t)\) and \(B(t)\) and for measurable mappings \(P(t)\) and \(Y(t)\) the \(x\)-projection of the intersection of all trajectory tubes of the perturbed systems, taken over all \(R(t)\), is equal to the trajectory tube of the original phase-constrained system. This proposition holds true either for \(B(t)\equiv 0\) and an arbitrary initial set \(Y_0\), or for an arbitrary continuous matrix function \(B(t)\) and an initial set containing the zero point. It is demonstrated on an example that the intersection over all matrix functions \(R(t)=\varepsilon Id\), where \(\varepsilon\) is a small number and \(Id\) denotes the identity operator, may not coincide with the intersection over all nondegenerate-valued matrix functions \(R(\cdot)\). The main result is applied to characterize upper ellipsoidal approximations of reachable sets of an uncertain control system. For other results on the application of small parameter methods in the theory of differential inclusions an interested reader may consult the works of \textit{V. A. Plotnikov} [Math. Notes 27, 456-459 (1980; Zbl 0473.34019)]. For the entire collection see [Zbl 0863.00015].