Application of Computations with Calculation Error Exclusion for Computation Geometry Algorithms

Abstract

Calculation errors while using floating-point computation are considered as a source of computational geometry algorithms incorrect results. Methods of decreasing such errors influence on the algorithm are discussed. Computations with calculation error exclusion as a way of such errors elimination are proposed. Computations with calculation error exclusion defined as a kind of computation where results of certain arithmetic operations could be represented exactly (computations in rational numbers, for example). Algorithm of finding cases where floating-point computation causes incorrect results is proposed.