Fixed set iterations for relaxed Lipschitz multimaps. (Q1395839)

The authors study asymptotic properties of the solutions and attainable sets to the autonomous dynamical system \[ \dot x(t)\in F(x(t))\quad\text{for a.e. }t\in (0,\infty),\quad x(0)\in K,\tag{1} \] where \(K\) is a compact subset of \(\mathbb R^N\) and \(F: \mathbb R^N\to 2^{\mathbb R^N}\) is nonempty, convex and compact-valued, satisfies a relaxed one-sided Lipschitz condition (RL) with constant \(L< 0\) and is upper semicontinuous. (RL) weakens both Lipschitz and monotonicity conditions. (We omit here the precise definitions of (RL) and some of the other terms below for brevity.) It is shown there exists a nonempty compact fixed set \(A^\infty\) which is the limit of the attainable sets to (1) as \(t\to\infty\) and that \(A^\infty\) is strongly flow invariant. Corollary 3.9 shows that \(A^\infty\) is the unique minimal attractor for (1). Lemma 3.10 shows that the Euler map \(G_h(x)\equiv \{x+ hf: f\in F(x)\}\) is almost contractive. Corollary 3.12 gives a stability condition for the attainable set for stationary iterations. From Corollary 3.16 an algorithm is derived for approximating \(A^\infty\) by using stationary Euler iterations with a given error bound. Theorem 3.18 shows that \(A^\infty\) attracts not only the continuous trajectories, but also the discrete Euler trajectories, with a rate of at least \(O(\sqrt{h})\). Section 4 considers the relationship between the fixed set \(A^\infty\) and the fixed-point set \(F^*\) (the set of equilibrium points of \(F\)). The authors note that \(F^*\subseteq A^\infty\), but the two sets can be the same under certain conditions. Theorem 4.1 shows that \(F^*= A^\infty\) if \(F\) is single-valued on a dense set. Lemma 4.3 establishes for \(n= 1\) that \(F^*= A^\infty\) even if \(F\) is not single-valued. Example 4.4 shows, for \(n\geq 2\), that \(F^*\) might be a proper subset of \(A^\infty\), even for linear \(F\). Finally, Section 5 discusses the hypotheses and gives directions for further study.