Архитектура системы управления интеллектуальным агентом на основе семиотической сетистатья из журнала
База данных: Каталог библиотеки СФУ (Р 587)
Библиографическое описание: Ровбо, Максим Александрович. Архитектура системы управления интеллектуальным агентом на основе семиотической сети = Control system architecture of an intelligent agent based on a semiotic network / М. А. Ровбо, П. С. Сорокоумов. - (Проблемы информатизации экономики и управления). - Текст : непосредственный // Открытое образование. - 2018. - № 5. - С. 84-93. - Библиогр.: с. 92-93 (24 назв.). - ISSN 1818-4243.
Аннотация: В данном исследовании предлагается формализовать влияние на поведение коллектива как задачу оптимизации управления сложной системой. Для этого поведение отдельного члена коллектива (агента) моделируется с использованием целевой функции, участвующей в выборе одного из возможных действий в соответствии с параметрами — относительными приоритетами допустимых действий, и методов оптимизации. Считается, что эти параметры поддаются внешнему контролю. Знания каждого агента об окружающем мире описываются в виде семиотической сети, пригодной для анализа текущего состояния агента и планирования его деятельности. Поведение управляемого разработанной системой одиночного агента исследуется на примере расширенной задачи фуражировки с проведением вычислительных экспериментов на компьютерной модели. Оптимизация приоритетов выполнения различных целей определяет успешность работы агента. Помимо приоритетов, свобода действий агента ограничивается необходимостью выживания. В этих условиях агент должен адаптироваться к внешней среде и внешним требованиями, при этом он будет способен как поддерживать своё функционирование, так и добиваться целей в соответствии с приоритетами. Моделирование работы двух типов агентов показало применимость подхода с сохранением адаптивных свойств агента. При этом целевое поведение меняется в широких пределах. Результаты, полученные для одиночного агента, в дальнейшем планируется дополнительно проверить для групп социально взаимодействующих агентов.
The aim of this research is to develop a novel method of affecting actions of an intelligent agent that allows changing the group behavior of such agents. The topic is relevant because group control is a complex and important task with considerable practical value. Proper management of a group of workers, schoolchildren or students has a beneficial effect for the participants, increases practical results achieved by them and uniting them. Therefore, the developed method can improve the efficiency of education. From the technical point of view, the relevance of the work is in the contribution to the development of an approach to controlling groups of robots or software agents with elements of social structures. Many management methods for groups of people have been developed in pedagogy, management, psychology, and other humanities. The achieved results are significant; however, many developed methods have important drawbacks. Some of the created approaches are non-formalizable, and their use is more an art than a science. In other cases, known methods may be unsuccessful because of a non-strict formulation of the problem and the multitude of adverse factors and applicability conditions. It is reasonable to develop a more robust method to influence team behavior. Some methods in artificial intelligence describe how to build a control system for distributed groups of agents: teams, packs or swarms. If these methods can be reformulated to be useful for specific practical tasks, as education or management, then rigor and reliable control of social groups will be possible. It is possible to formalize control of a team’s behavior as an optimization task. The behavior of an individual team member (agent) is modeled using an objective function, which is considered in the selection of one of the possible actions. Relative priorities of allowed actions are factors of this choice as parameters of the optimization process. An external controller can set these parameters. The world model of the agent is described as a semiotic network that is used to analyze the current state of the agent and plan its activities. The behavior of a single agent with the proposed method is investigated in a foraging task setting using a computer simulation. Different goals’ priorities optimization determines the performance of the agent. Agent’s freedom of action is limited by the priorities and the need for survival. The agent adapts to the conditions prevailing in the environment with these limiting factors. At the same time it is capable of both maintaining its functioning and achieving goals in accordance with its priorities. Simulation of two different types of agents showed the applicability of the approach and its preservation of the adaptive properties of the agent. The results acquired for a sole agent will be investigated for groups of socially interacting agents in future works.
The aim of this research is to develop a novel method of affecting actions of an intelligent agent that allows changing the group behavior of such agents. The topic is relevant because group control is a complex and important task with considerable practical value. Proper management of a group of workers, schoolchildren or students has a beneficial effect for the participants, increases practical results achieved by them and uniting them. Therefore, the developed method can improve the efficiency of education. From the technical point of view, the relevance of the work is in the contribution to the development of an approach to controlling groups of robots or software agents with elements of social structures. Many management methods for groups of people have been developed in pedagogy, management, psychology, and other humanities. The achieved results are significant; however, many developed methods have important drawbacks. Some of the created approaches are non-formalizable, and their use is more an art than a science. In other cases, known methods may be unsuccessful because of a non-strict formulation of the problem and the multitude of adverse factors and applicability conditions. It is reasonable to develop a more robust method to influence team behavior. Some methods in artificial intelligence describe how to build a control system for distributed groups of agents: teams, packs or swarms. If these methods can be reformulated to be useful for specific practical tasks, as education or management, then rigor and reliable control of social groups will be possible. It is possible to formalize control of a team’s behavior as an optimization task. The behavior of an individual team member (agent) is modeled using an objective function, which is considered in the selection of one of the possible actions. Relative priorities of allowed actions are factors of this choice as parameters of the optimization process. An external controller can set these parameters. The world model of the agent is described as a semiotic network that is used to analyze the current state of the agent and plan its activities. The behavior of a single agent with the proposed method is investigated in a foraging task setting using a computer simulation. Different goals’ priorities optimization determines the performance of the agent. Agent’s freedom of action is limited by the priorities and the need for survival. The agent adapts to the conditions prevailing in the environment with these limiting factors. At the same time it is capable of both maintaining its functioning and achieving goals in accordance with its priorities. Simulation of two different types of agents showed the applicability of the approach and its preservation of the adaptive properties of the agent. The results acquired for a sole agent will be investigated for groups of socially interacting agents in future works.
Год издания: 2018
Источник: Открытое образование
Выпуск: № 5
Номера страниц: 84-93
Количество экземпляров:
- Книгохранилище научной литературы (пр. Свободный, 79, 3 этаж): свободно 1 из 1 экземпляров
Ключевые слова: групповое управление, искусственные агенты, прикладная семиотика, робот, семиотическая сеть, фуражировка
Рубрики: Образование. Педагогика,
Высшее профессиональное образование,
Радиоэлектроника,
Искусственный интеллект. Экспертные системы
Высшее профессиональное образование,
Радиоэлектроника,
Искусственный интеллект. Экспертные системы
ISSN: 1818-4243
Идентификаторы: полочный индекс Р 587, шифр otob/2018/5-499520726