Metaheurística inspirada en Inteligencia Colectiva humana para problemas de optimización numérica continua

Abstract

Drawing inspiration from online question-and-answer (Q&A) online communities often regarded as embodiments of Collective Intelligence (CI), this study investigates the dynamics of reputation-driven and distributed network interactions in multi-agent systems as a model for problem-solving in engineering optimisation. We explore the interplay among diverse participants, including Solvers motivated by reputation and Users seeking net benefits, recognising its critical role in fostering success within these communities. Our study translates the principles of collective intelligence inherent in these interactions into a novel agent-based search algorithm for unconstrained numerical optimisation of continuous-valued cost functions. Empirical testing across a suite of established benchmark problems allows a comparative analysis of its performance against alternative agent-based methodologies. These findings underscore the algorithm’s advantages across diverse optimisation landscapes, highlighting the potential of the CI framework as a promising avenue in metaheuristic research. They illustrate how the interaction between individual actors and the collective, favours the emergence of global solutions in engineering contexts, mirroring similar emergent phenomena observed in social organisations.