A comparison of first-come-first-served and multidimensional heuristic approaches for asset allocation of floor cleaning machines

This paper presents a comparison of three binary heuristic algorithms - Particle Swarm Optimization (BPSO), Differential Evolution (BDE), and a Genetic Algorithm (GA) - under two asset allocation approaches to obtain optimal deployment combinations of assets for a cleaning company. When a company requires planning for the best asset allocation, the common practice is to rely on the experience of managers to estimate what combination of assets is appropriate to deploy to fulfill the different cleaning tasks that are carried out simultaneously in different locations. This work considers a given asset inventory of 168 machines composed of 29 different types of cleaners obtained online from three vendors. The comparison is made for three scenarios with 5, 10, and 15 tasks. Each task has a desired cleaning area and maximum cleaning time that are taken as constraints while minimizing the number of selected machines to be allocated for each task. The allocation approaches are inspired by the First-Come-First-Served (FCFS) approach and the Multidimensional Knapsack Problem (MKP). In total, 30 trials were computed for each scenario where the best, worst, average, and standard deviation were obtained. The results indicate that the combination of BDE and the FCFS approach provides the best solutions by having the lowest number of assets selected for each scenario. Additionally, it has the lowest standard deviation in cleaning area and time, while having a low execution time compared to the other approaches and heuristic algorithms.