On the choice of baseline schedules for the discrete time/resource trade-off problem under stochastic environment

Abstract

As a representative variant of the project scheduling problems, the discrete time/resource trade-off problem (DTRTP) is very important in both theory and practice. This paper investigates the stochastic DTRTP problem, and aims to give some principles about how to determine a baseline schedule to execute under stochastic environment (with an uncertain amount of work content of each activity), among a large number of candidate baseline schedules pre-obtained under deterministic environment (with a pre-estimated amount of work content of each activity), in order to obtain a statistically best performance (in terms of average project length). For this purpose, we do simulations on hundreds of randomly constructed DTRTP instances, based on which we further identify two characteristics, i.e. planned project length (PPL) and average parallel orders (APO), which seem to be closely related with the statistical performance. Simulation results show that, among all the candidate baseline schedules, it is much preferred to choose a baseline schedule with an as short PPL as possible. More interestingly, among baseline schedules corresponding to the same PPL, it seems more promising to choose a baseline schedule with an as large APO as possible. We hope these observations would be helpful for project managers to guide the choice of baseline schedules in practical applications.