МЕТОД ФОРМУВАННЯ ПОСЛІДОВНОСТІ ДІЙ ІНТЕЛЕКТУАЛЬНИХ АГЕНТІВ

Abstract

The article proposes an approach where the implementation of the formation of sequences of actions of intelligent agents is carried out by analogy with the activities of biological organisms using the mechanism of emotions to dynamically tune the body to perform actions. Thus, the functions of the limbic system are simulated in the organization of movements based on motivational behavior. When planning, first of all, the general condition of the agent is determined. Using the resulting state, a sequence of actions is formed. This approach will make it possible to dynamically reconfigure the sequence and respond to a dangerous situation or to a change in the internal state of the agent. An intelligent agent receives from the sensors and receptors signs of an initial condition, the goal is determined by it, and a sequence of actions is formed. Elements of a sequence of actions are elementary actions. An elementary action is characterized by a set of input parameters for functioning. Signs of the premise correspond to the first action in the sequence, the last action in the sequence is tied to the sign of the goal. The sequence of actions of the agent is represented by a digraph, where the vertices determine the elementary actions, and the edges determine the degree of bond strength between them. The initial conditions correspond to the first action in the sequence, the implementation of the sequence of actions begins with it. Signs of the goal correspond to the last peak in the sequence of actions Link weights change when general state variables are set, which allows you to perform a sequence of actions in real time with dynamic reconfiguration and select sequences of actions that are characteristic of a particular state. The method forms a sequence of actions that is initiated by emotional states and translates it into a sequence of automatic actions based on the achievement of the goal and which in the future will be performed in a normal state. To test the functioning of the method, a agent-work simulator is implemented in the V-REP program environment. The results obtained can be used for intelligent planning based on reinforcements and can be used in the management of agents, work in manufacturing enterprises, military agents, urban traffic flows, logistics systems, and social phenomena.