Neighborhood Based Optimization Algorithm

Evolutionary Computation (EC) algorithms are proposed as stochastic methods to solve complex optimization problems. Recently, the design of EC methods incorporates the design of complex operators to simulate the used metaphor. On the other hand, in a multi-agent system, the complex interactions among agents are coordinated, by the synergy of simple local behaviors. In this paper, a novel EC algorithm called Neighborhood-based Consensus for Continuous Optimization (NCCO) is presented. NCCO combines the simplicity of local consensus formulations, and reactive responses, based on neighborhood movement decisions to conduct the search strategy. NCCO considers a double pair of evolutionary operators for exploration-exploitation stages. The first pair; separation-alignment, conducts the search into wider zones, while the second pair; cohesion-seek, locally explores and exploits search spaces concentrating all agents into groups. These features guarantee leaderless movement decisions, since traditional EC approaches follow false-positive solutions, increasing the possibility of being trapped in local minima. Under such assumptions, the proposed algorithm exhibits a higher performance, against several state-of-art EC approaches evaluating common benchmark functions, and engineering design problems.