PENGEMBANGAN ALGORITMA ANT COLONY OPTIMIZATION UNTUK PERANCANGAN LINTASAN PERAKITAN KOLABORASI MANUSIA DAN ROBOT DENGAN MEMPERTIMBANGKAN WAKTU SETUP

• Main Author: Setiyadin Putri, Firda
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77871
• Institution: Institut Teknologi Bandung

This study aims to develop an Ant Colony Optimization (ACO) metaheuristic algorithm as an alternative to solve assembly line balancing problems with the characteristics of the Assembly Line Balancing Problem Human Robot Collaboration (ALBP-HRC) considering setup time. The objective function was to minimize cycle time. The data used in this study are secondary data obtained from four previous studies with various data sizes and precedences. The development of the ACO algorithm was divided into three stages. The first stage was modifying the standard ACO procedure to be specific for ALBP-HRC and adding development to make the search for solution more efficient. After that, a simulation was carried out with small data to get a more detailed picture. It was used to ensure that procedures was logical and produced feasible and convergent solutions efficiently. Furthermore, the last stage was the program code development according to the flowchart validated in the previous stage. Computational results using secondary data showed that the developed algorithm could reach the optimal solution on tasks ?25 and close to optimal on data with tasks >25 with a much faster time than the analytical method. In addition, there is an upper limit parameter for cycle time (?) developed in this study, which empirically contributes to the algorithm's efficiency by 52% faster. This study contributed to developing research on the ALBP-HRC problem using the ACO algorithm by considering setup time and adding parameters to increase the algorithm's effectiveness. In terms of performance, the ACO algorithm resulted in an average difference with the optimal solution of 0.67% and an average decrease in computation time of 61.75%. With the resulting performance, the developed algorithm was expected to contribute to policy-making inputs across assembly lines to increase effectiveness and efficiency.