Convex geometry and semiflows in P/T nets. A comparative study of algorithms for computation of minimal p-semiflows

Abstract

P-semiflows are non-negative left anullers of a net's flow matrix. The importance of these vectors lies in their usefulness for analyzing net properties. The concept of minimal p-semiflow is known in the context of Mathematical Programming under the name ``extremal direction of a cone''. This connection highlights a parallelism between properties found in the domains of P/T nets and Mathematical Programming. The algorithms known in the domain of P/T nets for computing elementary semi-flows are basically a new rediscovery, with technical improvements with respect to type of problems involved, of the basic Fourier-Motzkin method. One of the fundamental problems of these algorithms is their complexity. Various methods and rules for mitigating this problem are examined. As a result, this paper presents two improved algorithms which are more efficient and robust when handling ``real-life'' Nets.