Abstract
Computed results by floating-point arithmetic may not be accurate due to accumulation of rounding errors. Therefore, verifying the accuracy of the approximate solutions has been frequently discussed. In this paper, we focus on 2D orientation problem (Orient2D) which is one of the basic problems in computational geometry. It is assumed in most previous works that given data is rigorously represented by floating-point numbers. We assume that given floating-point numbers are results of rounding from real numbers. We develop floating-point filters of this problem and apply them to the convex hull algorithm.
