Multi-agent System Based on Self-adaptive Differential Evolution for Solving Dynamic Optimization Problems
Aleš Čep, Iztok Fister · 2017
This article presents the multi-agent system based on a selfadaptive differential evolution algorithm for solving dynamic problems.Characteristics of dynamic problems are that objective function, with which the quality of solutions is evaluated, changes over time.These changes can occur either after some predefined number of generations or, more commonly, in each generation, where the online responses are desired by the evolutionary algorithms.In this study, the former kind of problems were taken into consideration and were solved using multi-agent systems.Each agent in the systems is implemented as a self-adapted differential evolution with constant population size.In our comparative study, multiagent systems consisting of various population sizes in combination with various number of agents were compared, by solving the benchmark functions provided for CEC'09 Competition on Dynamic Optimization with respect to the same number of the fitness function evaluation.The main obstacle when using smaller populations is to prevent the stagnation of the population.That was partially overcome by using additional mechanisms such as: diversity measurement and information sharing between agents.As a result, the most appropriate multi-agent system was searched for, and the results of the best found multi-agent system were compared with the state-of-the-art algorithms.