Generalized enhanced exchange heuristic based resource constrained scheduler via the integration of evolutionary scheduling.

摘要

The Exchange Heuristic (EH) is a tool that solves Resource Constrained Scheduling (RCS) problems with general assumptions. EH attempts to balance resource utilization throughout the scheduling period. EH does this by shifting some activities later in the schedule to make enough space to assign a promising activity earlier in the schedule. This reassignment frequently leads to an improvement in the schedule by maximizing resource utilization. The promising activity is termed "target activity". Selecting the most promising target, as well as the order of activities to be shifted, constitutes the success of EH. The EH, as it is currently practiced, is highly dependent on an expert's intuition in these operations. This research suggests improving current EH practices by human experts with the use of neural networks (NN), due to their outstanding capability to both learn and deal with fuzzy data.;Different measures of attributes are considered to express the configuration of a schedule. Training neural networks requires a set of examples, and each example consists of a calculation of the attribute values. From the trained NN, synaptic weights are obtained, and these weights are used for the NN implemented in Generalized Enhanced Exchange Heuristic (GEEH). Also, EH is formally described mathematically in this study.;In GEEH, fixed resource capacity and requirements for each activity are relaxed. In addition, expendable resources are introduced. Expendable resources, like money, are an important consideration in practical applications. Therefore, the cost evaluation of the different scenarios in a project becomes possible. Furthermore, GEEH generalizes the concept of the predecessor by including weak predecessors.;As an extension of the study, the comparison study of EH with Evolutionary Scheduling (ES) is conducted. The ES is introduced, and the semiglobal nature of the EH is discussed. A Hybrid Method (HM) of ES and EH is suggested, and compared with EH and ES. It is proved in this study that ES is empirically superior than ES, and HM improves EH by over 4%.