Interactive preferences in multiobjective ant colony optimisation for assembly line balancing

In this contribution, we propose an interactive multicriteria optimisation framework for the time and space assembly line balancing problem. The framework allows decision maker interaction by means of reference points to obtain the most interesting non-dominated solutions. The principal components of the framework are the g-dominance preference scheme and a state-of-the-art memetic multiobjective ant colony optimisation approach. In addition, the framework includes a novel adaptive multi-colony mechanism to be able to handle the preferences in an interactive way. Results show how the multiobjective framework can interactively obtain the most useful solutions with higher convergence than previous a priori methods. The experimentation alsomakes use of original data of theNissan Pathfinder engine and practical bounds to define industrially feasible solutions in a set of scenarios. By solving the problem in these scenarios, we show the search guidance advantages of Communicated by V. Loia. This work has been supported by Ministerio de Economía y Competitividad under project SOCOVIFI2 (TIN2012-38525-C02-01 and TIN2012-38525-C02-02), and under PROTHIUS-III: DPI2010-16759, both including EDRF funding. M. Chica (B) · Ó. Cordón · S. Damas European Centre for Soft Computing, Mieres, Spain e-mail: [email protected] S. Damas e-mail: [email protected] Ó. Cordón DECSAI and CITIC-UGR, University of Granada, Granada, Spain e-mail: [email protected] J. Bautista Universitat Politècnica de Catalunya, Barcelona, Spain e-mail: [email protected] using an interactive multiobjective ant colony optimisation method.