Discontinuous dependence of the controllability set of some nonlinear systems on parameter (Q1914129)

Let \(S(\alpha, \mu)\) be the single input \(n\)-dimensional system \(\dot x= Ax+ \alpha bu+ \mu f(x)\) such that \(\langle x, Ax\rangle =0\), \(S(\alpha, 0)\) is controllable, \(f\) is \(C^1\) with \(f(0) =0\), for \(\mu\) small enough, there is a time \(T(\mu)\) such that all trajectories of \(S(0, \mu)\) stay in a given ball for \(t\geq T(\mu)\) and the origin is the only equilibrium. Using a result on the continuous dependence on a parameter of the controllability set \(G(\alpha, \mu (\alpha)\), \(\tau (\alpha))\) in finite time, it is shown that, for \(\mu\) small enough, there exists \(0\leq \alpha_1 (\mu)\leq \alpha_2 (\mu)\) such that \(G(\alpha, \mu)= \mathbb{R}^n\) for \(\alpha\geq \alpha_2 (\mu)\), \(G(\alpha, \mu)\) is bounded for \(\alpha\leq \alpha_1 (\mu)\), and \(G(\alpha, \mu)\) is unbounded for other \(\alpha\). A lower bound for a constant \(L\) such that \(\alpha_2 (\mu)\leq L\mu\) is given, by showing that the feedback \(u(x)=- \text{sign} (\langle x, b\rangle)\) will send any initial condition in the ball given by the assumptions to the origin under the flow of \(S(L\mu, \mu)\). The text has many misprints, some of them are already in the original russian version. In particular, a factor two has been forgotten in the computation of a bound, and the sign of the control has to be corrected.