TY - CONF
T1 - Modified multi-objective evolutionary programming algorithm for solving project scheduling problems
AU - Mohamed El Azouni, Ashraf
AU - Abido, M.
PY - 2021/6/11
Y1 - 2021/6/11
N2 - Though the Genetic Algorithm (GA) has received considerable attention recently in solving multi-objective optimization problems, inefficiency regarding performance has been reported in applications related to project scheduling. The degradation in efficiency was magnificent in applications of highly epistatic objective functions, including scheduling problems wherein the parameters being optimized are highly correlated. Furthermore, the crossover, being the dominant operator in GA, added significantly to the observed inefficiency for causing violations in the dependency between activities. Unlike GA, the Evolutionary Programming (EP) algorithm employs only a mutation operator which makes it less vulnerable to the dependency violation issue. This study proposes a modified Multi-Objective Evolutionary Programming (MOEP) algorithm to model and solve scheduling problems of multi-mode activities, including time–cost trade-off and finance-based scheduling with resource levelling. The modification involves the implementation of a new mutation operator to accommodate the scheduling problems in hand. Furthermore, the modified MOEP algorithm is benchmarked against the two multi-objective algorithms of SPEA-II and NSGA-II which have been used extensively in the literature to solve project scheduling problems. The results indicated that the modified MOEP algorithm outperformed SPEA-II and NSGA-II in terms of the diversity and quality of the Pareto optimal set.
AB - Though the Genetic Algorithm (GA) has received considerable attention recently in solving multi-objective optimization problems, inefficiency regarding performance has been reported in applications related to project scheduling. The degradation in efficiency was magnificent in applications of highly epistatic objective functions, including scheduling problems wherein the parameters being optimized are highly correlated. Furthermore, the crossover, being the dominant operator in GA, added significantly to the observed inefficiency for causing violations in the dependency between activities. Unlike GA, the Evolutionary Programming (EP) algorithm employs only a mutation operator which makes it less vulnerable to the dependency violation issue. This study proposes a modified Multi-Objective Evolutionary Programming (MOEP) algorithm to model and solve scheduling problems of multi-mode activities, including time–cost trade-off and finance-based scheduling with resource levelling. The modification involves the implementation of a new mutation operator to accommodate the scheduling problems in hand. Furthermore, the modified MOEP algorithm is benchmarked against the two multi-objective algorithms of SPEA-II and NSGA-II which have been used extensively in the literature to solve project scheduling problems. The results indicated that the modified MOEP algorithm outperformed SPEA-II and NSGA-II in terms of the diversity and quality of the Pareto optimal set.
M3 - Paper
ER -