Flock-based Evolutionary Multi-Agent System in Solving Noisy Multi-Objective Optimization Problems

It has been already proofed and discussed in our previous works that thanks to its computational as well as implemental simplicity Evolutionary Multi-Agent System (EMAS) approach can be easily adjusted and widely used for solving almost any type of multi-objective optimization task. Unfortunately, results obtained by distributed and decentralized agent-based evolutionary heuristic approach applied for solving multi-objective optimization problems (MOOPs) can be limited by such factors, as stagnation of agents' evolution or problems with distributing agents evenly over the whole approximation of the Pareto frontier. In the course of this paper three main shortcomings of EMAS-based approach are diagnosed and ideas of possible solutions of identified problems with the use of "flock-based" mechanisms are presented. Next, one of many possible realization of flock-based ideas i.e. so-called DensityPlus realization of floEMAS paradigm is discussed. Finally, experimental results obtained by floEMAS system in solving noisy multi-objective optimization tasks are presented. The goal of this paper is: to discuss shortcomings identified in applying EMAS for solving MOOPs, to present flock-based mechanisms allowing for overcoming such shortcomings, to discuss one of possible realization of floEMAS approach and to present the behavior of proposed floEMAS realization in noisy environment-since in the case of EMAS-based approaches there is no need to introduce any additional mechanisms to solve efficiently noisy problems.