GRAN and its application

Computational experiments conducted in the field of activity network planning techniques are usually based on the use of the generator of random activity network (GRAN) to generate a set of random activity networks for the purpose of algorithm validation. After discussing the basic characteristics to be possessed by a GRAN, this paper gives: first, an analysis for the existing problems in developing GRAN; secondly, several useful conclusions of designing GRAN; thirdly, specific algorithms of designing GRAN with a given number of nodes and arcs; and last, an analysis for the effectiveness of heuristic criteria in networks scheduling with limited resources. Two regression equations are derived to estimate the duration of networks characterized by one and three resources respectively using the minimum slack time criterion