Robustness problems and verified computations for computational geometry

This paper is concerned with robustness problems in computational geometry. To solve geometric problems, numerical computations by floating-point arithmetic are preferred in terms of computational performance. However, when the rounding errors of floating-point computations accumulate, meaningless results of the problems may be obtained. It is called robustness problem in computational geometry. Recently, several verified algorithms for a two-dimensional orientation problem have been developed. By applying these algorithms, we develop an algorithm which overcomes this problem. Finally, we present numerical examples for a convex hull for a set of points on two-dimensional spaces to illustrate an efficiency of the proposed algorithm.