Modified Niched Pareto Multi-objective Genetic Algorithm for Construction Scheduling Optimization

This research proposes a Genetic Algorithm based decision support model that provides decision makers with a quantitative basis for multi-criteria decision making related to construction scheduling. In an attempt to overcome the drawbacks of similar efforts, the proposed multi-objective optimization model provides insight into construction scheduling problems. In order to generate optimal solutions in terms of the three important criteria which are project duration, cost, and variation in resource use, a new data structure is proposed to define a solution to the problem and a general Niched Pareto Genetic Algorithm (NPGA) is modified to facilitate optimization procedure.

The main features of the proposed Multi-Objective Genetic Algorithm (MOGA) are: A fitness sharing technique that maintains diversity of solutions.
A non-dominated sorting method that assigns ranks to each individual solution in the population is beneficial to the tournament selection process. An external archive to prevent loss of optimal or near optimal solutions due to the random effect of genetic operators. A space normalization method to avoid scaling deficiencies.

The developed optimization model was applied to two case studies. The results indicate that a wider range of solutions can be obtained by employing the new approach when compared to previous models. Greater area in the decision space is considered and tradeoffs between all the objectives are found. In addition, various resource use options are found and visualized. Most importantly, the creation of a simultaneous optimization model provides better insight into what is obtainable by each option.

A limitation of this research is that schedules are created under the assumption of unlimited resource availability. Schedules created with this assumption in real world situations are often infeasible given that resources are commonly constrained and not readily available. As such, a discussion is provided regarding future research as to what data structure has to be developed in order to perform such scheduling under resource constraints.