Dual nature of mass multi-agent systems

Dual nature of mass multi-agent systems (mMAS) emerging as an internal discord of two spheres - micro (virtual) consisting of agents and their internal phenomena, and macro arising at the interface to the real world $stems the necessity of a new approach to analysis, design and utilisation of such systems. Based on the concept of VR decomposition, the problem of management of such systems is discussed. As a sub-type that makes mMAS closer to the application sphere, an evolutionary multi-agent system (EMAS) is proposed. EMAS combines features of mMAS with advantages of an evolutionary model of computation. As an illustration of this consideration two particular EMAS are presented, which allow us to obtain promising results in the fields of multiobjective optimisation and time-series prediction, and thus justify the approach.