A dynamic sampling technique for the simulation of probabilistic and generalized activity networks

Most probabilistic activity networks (e.g. PERT) of any reasomable size are practically impossible to analyse mathematically in an acceptable time. This problem is augmented when stochastic branching is introduced to form generalized activity networks. For this reason simulation has proved to be one of the more popular and 'accurate' techniques available for network attribute analysis. In this paper a dynamic sampling technique is introduced that improves on the standard simulation approach used in popular project management software tools. A comparison is also made between the simulation requirements of standard probabilistic activity networks and a finite sample set of generalized activity networks in which activities are assigned either dependent or independent probability generations.