“…For the second stage one uses algorithms from linear programming (Bahr et al, 1968;Bollmann et al, 1981;Hodes, 1974;Rosen et al, 1991) or nonlinear, in particular quadratic, programming (Brahme, 1995;Censor et al, 1988;Chriss et al, 1995;Cooper, 1978;van Dalen et al, 2000;Holmes and Mackie, 1994;Hristov and Fallone, 1997;McDonald and Rubin, 1977;Redpath et al, 1976;Shu et al, 1998;Starkschall, 1984;Stein et al, 1997;de Wagter et al, 1998;Xing and Chen, 1996) and control theory (Hristov and Fallone, 1998) including multiple objective approaches (Cotrutz et al, 2001;Hamacher and Küfer, 2002;Yu, 1997). Also iterative dose reconstruction techniques (Bortfeld et al, 1990;Holmes et al, 1995;Pugachev et al, 2000;Söderström and Brahme, 1992) and methods from global optimization have been developped (Wu and Zhu, 2001b). Moreover, it must be mentioned that one-stage algorithms have been designed by means of mixed integer programming (Burkard et al, 1995;Langer et al, 1996;Lee et al, 2000) but they are time-and memory-consuming, too.…”