Dzinko A. Multi-agent approach to material flow control processes automation in logistic systems

The dissertation is devoted to development, creation and study of material flow control processes automation in logistic systems based on multi-agent approach. Throughout the dissertation, an analysis of existing methods and material flows control systems was conducted. Also existing methods and approaches to dynamics of material flows movement were analyzed, control efficiency criteria were identified: general system optimality criteria, and criteria based on intelligent agents goal functions. During the work on the dissertation, a material flow dynamics model was developed based on universal properties of material flows which exist on all hierarchical levels of logistic systems. As a base for the model, a supply chain model by Lisovskyi-Ponomarenko was taken giving a static representation of planned logistic system. Based on positive and negative factors of existing criteria, methods, control systems features and material flow movement models, a multi-agent model of material flow movement environment was created, a multi-level structure and function of the dispatching agent were presented. Analytical model of multi-agent environment was created, two groups of effectiveness criteria were presented. Based on presented criteria, a discrete-stochastic material flow movement presentation model was proposed along with material flows dispatch function based on discrete-stochastic dynamic programming and decision making Markov chains. The function for searching of quasi-optimal solution for material flow control tasks according to discrete-stochastic dynamic programming principles was developed. All decisions are made based on numeric assessments of the next system's step efficiency. In order to check the efficiency of proposed control method, an imitational modeling of multi-agent environment and proposed methods was conducted.