Project Management { Multiple Resources Allocation

1. Abstract Given a project network under stochastic conditions, the goal is to determine the optimal resource allocation to the activities in order to minimize the total project cost. This cost includes the resource cost and the tardiness cost. In this work we consider the multiple resources case, which is an extension of the models previously developed by the first author and other researchers, considering a single resource. We assume that all the resources are independent and abundant. The work consists mainly of two parts: formalization of the new models, and their implementation in Java. In order to formalize the models, it was necessary to establish an allocation strategy for the multiple resources. This is required to ensure the desired equality of expected durations yielded by each resource in the same activity. We study four different allocation strategies: two of them are derived from the stochastic nature of the work content by equalizing the expected durations, thus determining the allocation vectors; and the other two go down to the level of all possible values to devise an allocation method (among all the allocation vectors, selects those leading to equal expected durations). Then the probability distributions of the variables required for analysis and evaluation were determined. Although the research has covered four strategies, one proved to be inferior compared to the others, and another was too complex to be easily implemented. The remaining two are strong rivals with neither dominating the other, and one of them was arbitrarily chosen for implementation. The implementation covers three algorithms: Dynamic Programming Algorithm, Electromagnetic Algorithm and Evolutionary Algorithm. Concurrent programming was exploited to enhance performance. We report on the performance of our application over a representative set of project networks.