SOLVING THE OPTIMAL RESOURCE ALLOCATION IN MULTIMODAL STOCHASTIC ACTIVITY NETWORKS USING AN OPTIMAL COMPUTING BUDGET ALLOCATION TECHNIQUE

Abstract

The problem of optimal Resource Allocation in Multimodal Stochastic Activity Networks (RAMSAN) has been studied for more than a decade. Many researchers proposed solution approaches, including dynamic programming and meta-heuristics, for solving the RAMSAN, and they commonly applied Monte Carlo Simulation (MCS) for evaluating the objective function values for the candidate solutions. However, since the computational load of MCS is very demanding, the solution approaches using MCS become impractical, even for only medium-size problems. In this study, we propose a Genetic Algorithm (GA) with an Optimal Computing Budget Allocation (OCBA) approach for solving the RAMSAN. We utilize the technique of OCBA to optimally allocate the computational budget among the candidate solutions in the evolutionary process of GA for evaluating their objective functions. Based on the benchmark instances in the literature, we demonstrate the proposed GA with OCBA is more effective than the other solution approaches in the literature and a GA without using OCBA.