An improved estimate for the convergence of alternating sums to the alternating Pontryagin integral (Q799510): Difference between revisions

An estimate is known for the Hausdorff distance between the alternating integral (A.I.) and the alternating sum (A.S.), which is the most important part of the error estimate in the numerical method of constructing A.I., and implicitly uses a result on the equality of the A.I. and A.S. for a differential game with simple motion, i.e., when the matrix \(A=0\). This result was extended recently to the case \(A=\lambda E\), where \(\lambda\) is a number and E is the unit matrix, and moreover it was shown that this extension is definitive, in the sense that for a matrix \(A\neq\lambda E\) there is not equality between the A.I. and the A.S., irrespective of the time interval the sets P and Q and the terminal set [see \textit{P. B. Gusyatnikov} and \textit{E. S. Polovinkin}, Differ. Equations 16, 867-874 (1981; Zbl 0467.90096)]. In connection with this, it is possible to improve the estimate for the Hausdorff distance between the A.S. and the A.I. such that for the case \(A=\lambda E\), the improved estimate vanishes, i.e., the A.I. and A.S. coincide. The earlier estimate can be obtained from the improved one in the special case \(\lambda =0\).