On set-valued differentiation and integration (Q1364812)

[Common review of the paper cited above and the present one.] In the first paper, the author announces some results on set-valued differentiation and integration, without any proofs. The second paper is an extensive presentation of this theory. The author deals with mappings from an interval into \(\text{co }\Omega(\mathbb{R}^n)\), the space of all nonempty convex and compact subsets of \(\mathbb{R}^n\). In his theory, the key role is played by quasi-affine mappings; both, differential and integral are defined via approximations by such mappings. A multifunction \(F:[0, T]\to\text{co } \Omega(\mathbb{R}^n)\) is quasi-affine if there exist \(F_0\in \text{co }\Omega(\mathbb{R}^n)\) and a positively homogeneous function \(c:\mathbb{R}^n\to \mathbb{R}\) such that \[ s(F(t),\psi)= [s(F_0,\cdot)+ tc(\cdot)]^{**}(\psi),\quad 0\leq t\leq T,\quad \psi\in\mathbb{R}^n, \] where \(s\) is the support function and asterisk denotes the Fenchel conjugate. The author studies basic properties of quasi-affine mappings and compares them with other multivalued generalizations of affine mappings, e.g., eclipsets and multi-affines. A multifunction \(F\) is differentiable at \(t_0\) if there exists a quasi-affine mapping \(G\) such that \(h(F(t_0+ t),G(t))= o(t)\), \(t>0\), where \(h\) is the Hausdorff metric. The author proves the basic rules of calculus for such defined differentials and studies relations between the differentiability of \(F\) and its support function \(t\to s(F(t),\psi)\). The final part of the paper is devoted to the set-valued integral, defined as the limit of the Riemann type integral sums being piecewise quasi-affine mappings. This concept of the integral includes the set-valued integral of Dinghas and Hukuhara, and Pontryagin's alternated integral. The author gives sufficient conditions for the existence of such an integral and estimates the convergence of integral sums. Then he proves that the differential of the integral with respect to the upper limit of integration can be expressed in terms of the integrand. Thus the differentiation of a multifunction can be treated as an inverse operation to the integration. The paper is endowed with many examples and counterexamples. For applications of proposed set-valued calculus, the author refers to his other papers.