Resource-Constrained Project Scheduling for Timely Project Completion with Stochastic Activity Durations

Abstract

Resource-constrained project scheduling for timely project completion with stochastic activity durationsWe investigate resource-constrained project scheduling with stochastic activity durations. Various objective functions related to timely project completion are examined, as well as the correlation between these objectives. We develop a GRASP-heuristic to produce high-quality solutions, using so-called descriptive sampling. The algorithm outperforms existing algorithms for expected-makespan minimization. The distribution of the possible makespan realizations for a given scheduling policy is also studied.

Bi-Objective Multi-Mode Project Scheduling Under Risk Aversion

“…Tereso, Araujo, Moutinho, and Elmaghraby (2008) investigate dynamic programming as well as diverse metaheuristic algorithms on a special stochastic resource allocation problem in project scheduling. Ballestin and Leus (2009) present a comprehensive investigation of the RCPSP under stochastic activity durations, considering diverse objective functions such as the expected makespan, the makespan standard deviation and the probability of meeting a due date. As the computational solution technique, a greedy randomized adaptive search procedure (GRASP) is applied.…”

Project portfolio selection under uncertainty with outsourcing opportunities

“…3-4 restrict the decision variables x irjt and y i to nonnegative and binary values, respectively. Two possible generalizations of the basic model (1)(2)(3)(4), which can also be combined, are the following:…”

“…Later, when the search has been guided to more promising areas of the search space, the accuracy of objective function approximation has to be increased in order to identify the true optima. 3 In the context of our problem, we modify the S-VNS algorithm by replacing sampling with numerical optimization: Similarly as in the original S-VNS, where the objective function value of a solution y is only estimated (by its sample average), we only estimate the optimal solution value min x g(y, x) of the subproblem assigned to portfolio y, but now we use the (deterministic) Frank-Wolfe procedure for obtaining a numerical estimate. Obviously, the larger the number of iterations given to the Frank-Wolfe procedure is chosen, the better does the procedure approximate the true value of min x g(y, x).…”

“…In order to obtain good initial solutions for Frank-Wolfe, we solve the LPs given by the deterministic boundary case (see Sect. 3…”

“…[26]). Within this framework (which differs from the model presented here), stochastic work times have been investigated in a large number of articles, see, e.g., [3] for a recent example. For treating the staffing aspect, it is often necessary to take account of heterogeneous competencies or skills of employees.…”