Abstract
Resource constrained project scheduling problem with milestone payments (RCPSPDCF-MP) is an intractable combinatorial problem. This has prompted researchers to propose a variety of metaheuristic approaches to address the problem. Hybridizing different metaheuristics to produce synergetic effect is a complex endeavor. In this study, we propose a distributed adaptive metaheuristics selection (DAMS) that leverages distributed computing nodes of modern computing architecture to hybridize heterogeneous metaheuristics, evaluated based on Chernoff-Hoeffding upper confidence bounds (UCB1) to solve RCPSPDCF-MP. Our DAMS framework selects a tailored set of metaheuristics for each problem instance (project) from five different metaheuristics customized to solve RCPSPDCF-MP. The parameters of each metaheuristic were tuned off-line using classical particle swarm optimization (PSO). Our proposed framework is examined using project sets from the test library, Project Scheduling Library (PSPLIB). Experiments show that the hybrid of metaheuristics generated by UCB1 based DAMS framework outperform existing methods from the literature.