Search

Search Results

• EVOLUTIONARY DESIGN AS A TOOL FOR DEVELOPING MULTI-AGENT SYSTEMS

  L.A. Gladkov , N.V. Gladkova

  2020-11-22

  Abstract ▼

  The article is devoted to the discussion of the problems of constructing evolving multi-agent systems
  based on the use of the principles of evolutionary design and hybrid models. The concept of an
  agent is considered. A set of basic properties of the agent is presented. The analogies between multiagent
  and evolutionary systems are considered. The principles of construction and organization of multi-
  agent systems are considered. The similarities between the main definitions of the theory of agents
  and the theory of evolution are noted. It that the main evolution models and evolutionary algorithms can
  be successfully used in the design of multi-agent systems is noted. The analysis of existing methods andmethodologies for designing agents and multi-agent systems is carried out. The existing differences in
  approaches to the design of multi-agent systems are noted. The main types of models are described and
  their most important characteristics are given. A model of agent interaction, including a description of
  services (services), relationships and obligations existing between agents is presented. The model of
  relations (contacts), which defines communication links between agents is described. The importance
  and prospects of using the agent-based approach to the design of multi-agent systems are noted. The
  concept of designing agents and multi-agent systems, according to which the design process includes the
  basic components of self-organization, including the processes of interaction, crossing, adaptation to the
  environment, etc is proposed. Various approaches to the evolutionary design of artificial systems are
  considered. An evolutionary model of the formation of agents and agencies as the main component of
  evolutionary design is proposed. Modified evolutionary crossing-over operators to implement the agent
  design process are proposed.

• TOP-DOWN VS BOTTOM-UP METHODOLOGIES FOR ADAS SYSTEM DESIGN

  D.E. Chickrin, А. А. Egorchev

  2021-07-18

  Abstract ▼

  Selection of the principal design methodology has a significant impact on final product
  quality, including its evolvability and scalability. The article discusses the features of traditional
  bottom-up and top-down design methodologies in the context of ADAS (driver assistance and automated
  driving systems). Necessity of the combined design methodology is shown due to unacceptability
  of “pure” methodologies for design of this kind of systems. For this purpose, the features
  and limitations of the top-down approach are considered: commitment to maximum compliance
  of the developed system with its requirements; methodological rigor of the approach; difficulty
  of system testing in the process of the development; sensitivity to changes in requirements.
  The features and limitations of the bottom-up approach are considered: possibility of iterative
  development with obtaining intermediate results; possibility of using standard components; scalability
  and flexibility of the developed system; possibility of discrepancy of functions of subsystems
  to requirements, which may appear only at later stages of development; possible inconsistency in
  development of separate subsystems and elements. The features and factors of ADAS system development
  are considered: increased requirements for reliability and safety of the system; heterogeneity
  of used components. Two stages of ADAS-systems development are distinguished: the
  stage of intensive development and the stage of extensive evolution. The applicability of one or
  another methodology to various aspects of ADAS system development and evolution (such as:
  requirements definition; compositional morphism; scalability and extensibility; stability and sustainability;
  cost and development time; development capability) is considered. A comparison of themethodologies concludes that there are aspects of technical system design and development in
  which there is a significant advantage of one or the other of the methodologies. Only the bottomup
  approach can ensure the proper evolution of the system. However, for complex systems, it is
  critical to define the initial requirements for the system, which can only be achieved using the topdown
  methodology.

• EVOLUTIONARY DESIGN AS A TOOL FOR DEVELOPING MULTI-AGENT SYSTEMS

  L. A. Gladkov, N. V. Gladkova

  2021-11-14

  Abstract ▼

  The article is devoted to the discussion of the problems of constructing evolving multi -
  agent systems. Possible methodologies for designing multi-agent systems are considered. The
  relevance of developing new principles for constructing multi -agent systems based on evolutionary
  design methods is noted. The correspondences between the terms of the theory of
  agents and the theory of evolution are highlighted. The prospects of using hybrid approaches
  to the design of multi-agent systems are noted. The principles of construction and the poss ibility
  of using fuzzy genetic algorithms in the design of multi -agent systems are considered.
  It is suggested that the models and methods of the theory of evolutionary modeling can be
  successfully applied in the design of multi-agent systems. An evolving multi-agent system is
  proposed. The procedure for the formation of new agents in the process of evolution is described.
  The set of parameters for assessing the state of each agent in the population has
  been determined. The resource parameters are proposed to be used to assess the current state
  of the agent and the possibilities of its interaction with other agents. The definitions of an
  agency and a family, the minimum elements of an evolving multi -agent system are given. An
  evolutionary strategy for constructing a model of an evolving multi -agent system is proposed.
  The procedures for the execution of the original evolutionary operators for processing the
  population of agents are described. Based on the proposed methodology, a software system
  for supporting the evolutionary design of agents and multi-agent systems was developed. Atpresent, computational experiments are being carried out to study the proposed design model
  for multi-agent systems, as well as to evaluate the effectiveness of various operators and
  schemes for the formation of descendant agents, the necessary conditions for survival.