The rotation minimizing frame is superior to the Frenet frame for mode
ling sweep surfaces [F. Klok, Computer Aided Geometric Design 3, 217-2
29 (1986)] However, the existing techniques for computing the rotation
minimizing frame either have low approximation degree or are unrobust
numerically. We present a method to compute an approximate rotation m
inimizing frame in a robust and efficient manner. The following proble
m is studied. Given an axial curve A(u) in space and a 2D cross-sectio
n curve C(v), generate a sweep surface S(u, v) = A(u) + F(u)C(v), wher
e F(u) is a rotation minimizing frame defined on A(u). Our method work
s by approximating A(u) with a G(1) circular-are spline curve and then
sweeping C(v) with a rotation minimizing frame along the approximatin
g circular-are spline curve; the sweep surface thus generated is an ap
proximation of S(u, v). The advantages of this method are: (1)the appr
oximate rotation minimizing frame is computed robustly, with its error
being much smaller than would be obtained by Klok's linear method wit
h the same number of segmentations; (2) the sweep surface generated is
a NURBS surface if the cross-section curve is a NURBS curve; (3) the
method is easily adapted to generating a smooth and closed sweep surfa
ce when A(u) is a closed smooth curve. (C) 1997 Elsevier Science Ltd.