CC BY
7ТЕХНИЧЕСКИЕ НАУКИ
ПРОЕКТИРОВАНИЕ ИНТЕЛЛЕКТУАЛЬНЫХ АГЕНТОВ ОСНОВАННЫХ НА
ЛОГИЧЕСКИХ СХЕМАХ С ПОМОЩЬЮ АЛГОРИТМА _ГЕНЕТИЧЕСКОГО ПРОГРАММИРОВАНИЯ
Казанцева Юлия Валерьевна
Магистр
Сибирский Государственный аэрокосмический университет имени академика М. Ф. Решетнева, г. Красноярск
Липинский Леонид Витальевич
кандидат тех. наук, доцент Сибирский Государственный аэрокосмический университет имени академика М. Ф. Решетнева, г. Красноярск
Карчава Ольгап Витальевна
старший преподаватель Сибирский Государственный аэрокосмический университет имени академика М. Ф. Решетнева, г. Красноярск
INTELLIGENT AGENTS DESIGN BASED ON THE LOGIC CIRCUIT BY MEANS OF THE GENETIC PROGRAMMING
Kazantseva Yyliya, Master's degree Reshetnev Siberian State Aerospace University, Krasnoyarsk
Lipinsky Leonid, Candidate of Engineering Sciences, assistant professor Reshetnev Siberian State Aerospace University.
Krasnoyarsk
Karchava Olga, Senior teacher Reshetnev Siberian State Aerospace University. Krasnoyarsk АННОТАЦИЯ
В данной работе предложен метод проектирования интеллектуальных агентов, основанных на логических схемах при помощи генетического программирования. ABSTRACT
In this paper method of designing intelligent agents based on logic circuits, using genetic programming is suggested. Ключевые слова: интеллектуальные агенты, интеллектуальные технологии. Keywords: intelligent agents, genetic programming, logic circuits.
Intelligent information technologies (SIT) are successfully applied for a wide range of practical tasks [3]. Intelligent agents are one of the recent breakthroughs in IIT(, intended to solve different problems without a human. Initially, all the possible sequences of actions have to be created by a developer and installed into the intelligent agents. In case such system faces conditions that haven't been considered by the creator, the system may crash or
even fail.
In the field of Artificial Intelligence this is called an "intellectual essence" (software) [4], acquiring information about the outer environment, the current processes and their results through the system of sensors, and having an access to the management through the actuator system. The behavior of the agents is targeted at a certain aim [1]. A scheme of such agent is presented in Figure 1.
Agent
Sensors -4-
Perception
1
Decision Module
T
Actuator
Actions
Environment
Figure 1. Interaction of the intellectual agent between the environment.
One of the agent development methods is forming an agent with the help of logical schemes (Fig.2). In this case the agent has a list of formal logic statements, describing the causal relationship in the environment and possible
changes in it that can be caused by the agent's actions, interpreted at a current state of environment as a true one, on the assumption that the agent has performed the corresponding action.
Events
А
Actions
■o
В
о
D
О
> 4
1
2
3
С
D
Figure 2. An example of a generalized logical scheme of the agent.
For the automatic formation of the agent through the genetic programming the agent should be rendered into a tree [2]. For this task it is convenient to render an individual into a set of trees (as it is depicted in Fig.3). Each tree is responsible for one particular decision, based on one action. Each individual includes the number
of trees equal to the number of actions one agent can perform. The functional open set of the agent includes such logical operators as AND, OR, NAND and negation. The terminal set consists of all the events that the agent can detect and register.
Figure 3. The conception of an individual
The distinctive feature of the algorithm is the fact that the main evolutional operators, such as initiation, breeding and mutation will work with the genotype, presented not by one tree, but by several ones. This causes not alteration in the flow of initiation and mutation processes. However, in the process of breeding genetic code exchange occurs only between those trees that comply with the same actions. Applicability assessment is conducted within the agent environment functioning model. Agent is supplied with the necessary resources on functioning and a logical scheme on the basis of the estimated individual is formed. As the agent finishes all operations, the efficiency of its performance is evaluated and the agent is declared applicable or inapplicable for the evolutional algorithm.
The application of genetic programming for intelligent agents' development and optimization is probable to optimize both the time of agent development and
the efficiency of their usage. The intellectual agent development method has a wide range of application spheres from computer games to unmanned vehicle control.
Reference list
1. Bugaytshenko D.Y., Solovyev I.P. Abstract intelligent agent architecture and its realization methods. 2005.
2. Lipinsky L.V. Semyonkin E.S. Automation of intelligent information technologies design through genetic programming method. Fizmatlit, 2006.
3. Lipinsky L.V. Semyonkin E.S. Genetic programming algorithms for intellectual information technologies formation. Krasnoyarsk 2006.
4. Rassel S. Norwig P. Artificial Intelligence: modern approach. Williams. 2006. S 282-331
