No label defined (Q4939726)

The authors present a survey of results on the development of A.M. Samojlenko's numerical-analytic method (NAM). Initially this method was elaborated to construct periodic solutions to nonlinear systems of ODEs. But in the paper of \textit{A. M. Samojlenko} and \textit{V. A. Ronto} [Ukr. Math. J. 33, 356-362 (1982; Zbl 0479.65046)] it was shown that the NAM can be applied to study two-point boundary value problems of the form \[ \dot x=f(t,x) \quad Ax(0)+Cx(T)=d,\tag{1} \] with \(f\in C([0,T]\times D \mapsto {\mathbb{R}}^n)\), \(D\) is a domain in \({\mathbb{R}}^n\), \(A, C\) are given \(n\times n\)-matrices and \(d\in{\mathbb{R}}^n\) is a given vector. Later problems with multipoint boundary conditions \(\sum_{i=1}^{m+1}A_ix(t_i)=d\) and even with functional ones of the form (2) \(\ell(x):=\int_{0}^{T}x(s) ds=d\) were considered. When applied to the periodic problem the numerical part of the NAM consists of solving certain transcendental determining equation. The authors show that in the case of problems (1)--(2) one can organize an iterative process in such a way that the determining equation does not arise. Namely, it is sufficient to determine successive approximations \(x_m(t)\) by means of the formula \[ x_{m+1}(t)=d/T-\int_{0}^{T}(1-s/T)f(s,x_m(s)) ds+\int_{0}^{t} f(s,x_m(s)) ds, \] \(m=0,1,2,\ldots, x_0=d/T\). Some generalizations of this procedure are considered and a number of sufficient conditions for the convergence of the method are analyzed [see also the authors, Ukr. Mat. Zh. 50, No. 1, 102-117 (1998; Zbl 0897.34014)].