Scheduling modular projects on a bottleneck resource

Project planning with alternative technologies in uncertain environments

“…In drug development, for example, when a certain module is needed to show the effectiveness of a drug, two precedence-related activities could represent the repeated measurement of the beneficial effects of the drug: the first test is performed after one week; the effects after two weeks will only be measured if first the effects after one week are inconclusive [9]. Alternatively, trials may be repeated in different doses and with different test subjects [19].…”

“…Precedence constraints within modules may also represent fallback options for project failure, as 'contingency plans': plans devised for an outcome different from expected. Comparable modeling choices were made in Coolen et al [9] and in Huysmans et al [19], but without discounting the cash flows, in which case durations become irrelevant and scheduling all activities sequentially is a dominant choice.…”

“…We borrow the datasets that were generated by Coolen et al [9]: these consist of 10 instances for each of various values of the number of activities n and for OS = 0.4, 0.6 and 0.8, with 'order strength' OS the number of comparable activity pairs according to the induced order B * , divided by the maximum possible number of such pairs (this value is only approximate). Average activity durations are not used by Coolen et al [9] and are additionally generated for each activity, for each instance separately; each such average duration is a uniform integer random variate between 1 and 15.…”

“…Average activity durations are not used by Coolen et al [9] and are additionally generated for each activity, for each instance separately; each such average duration is a uniform integer random variate between 1 and 15. In the generated instances, all activities apart from the final one have negative cash flows and the final activity has a positive payoff (which is also significantly larger in absolute value); we refer to the appendix of [9] for more details.…”

“…In order to examine the impact of duration variability on the value of a project in a more structured fashion, we have generated new instances in line with [9], with n ∈ {11, 21} and OS ∈ {0.4, 0.6, 0.8} but now we generate 100 instances per combina-tion of parameter settings, and there is no activity failure nor modular completion of the project (each activity constitutes a separate module). The payoff value C is (uniform) randomly chosen from interval [0.9C 0 ; 2C 0 ], where C 0 is the payoff value that associates objective value 0 (break-even) with the early-start policy Π ES for SCV = 1.…”