Adaptation of fuzzy cognitive maps by migration algorithms

Conventional rule‐based systems are insufficient for description of complex dynamic systems requiring nontrivial decision procedures. Fuzzy cognitive maps seem to be convenient to overcome these limitations. However, they lack ability of self‐learning and therefore some adaptation approaches are needed. The purpose of this paper is both to show the use of fuzzy cognitive maps for such systems and to present migration algorithms as convenient adaptation means.

Some problems of a complex dynamic system description by knowledge‐based means are discussed. Fuzzy cognitive maps are presented as a possible way to solve these problems followed by description of migration algorithms as their adaptation means. Their use is clarified on an example of the so‐called parking problem based on path planning using a graph search algorithm and a traffic simulation system.

After series of simulations the reality of the proposed system and selected methods with their modifications was proved. It has shown the robustness of the presented solution under circumstances of uncertainty, too.

The paper points to stability investigation of the proposed approach introducing uncertainties into the traffic simulation system to take into account, e.g. unexpected events. Further, a possibility of developing a linguistic information retrieval system is mentioned.

The proposed approach can find various implementations not only in planning tasks but also in robotic navigation and multi‐agent applications in general. In addition, it suggests possibilities of knowledge‐based systems, directly using human‐like approaches, to areas of decision making under uncertainties and contradictories.

An new modification of migration algorithms for adaptation of parameters for fuzzy cognitive maps is introduced and compared to other known self‐learning methods. Further, the concept of a traffic simulation system for path planning is presented.