Task allocation in distributed artificial intelligence using collective behavior models

The paper presents a modified search architecture, based on the paradigm of multi-agent approach to solving complex problems. This approach allows us to parallelize the process of finding solutions and managing the problem pre-convergence algorithms. In this paper we propose a decomposition mechanism of the original problem based on the bee algorithm to determine the most perspective solutions and neighborhoods and further search process delegation to the agents that implement various optimization techniques. Conducted series of experiments have shown the efficiency of the designed search engine, compared with the genetic, evolutionary and swarms optimization algorithms. Parallel computing application for solving optimization problems improves the quality of the obtained solutions up to 8 percent.