• Journal of Information Processing Systems
• /
• v.8 no.3
• /
• pp.409-420
• /
• 2012

The decision-making by agents in games is commonly based on reinforcement learning. To improve the quality of agents, it is necessary to solve the problems of the time and state space that are required for learning. Such problems can be solved by Macro-Actions, which are defined and executed by a sequence of primitive actions. In this line of research, the learning time is reduced by cutting down the number of policy decisions by agents. Macro-Actions were originally defined as combinations of the same primitive actions. Based on studies that showed the generation of Macro-Actions by learning, Macro-Actions are now thought to consist of diverse kinds of primitive actions. However an enormous amount of learning time and state space are required to generate Macro-Actions. To resolve these issues, we can apply insights from studies on the learning of tasks through Programming by Demonstration (PbD) to generate Macro-Actions that reduce the learning time and state space. In this paper, we propose a method to define and execute Macro-Actions. Macro-Actions are learned from a human subject via PbD and a policy is learned by reinforcement learning. In an experiment, the proposed method was applied to a car simulation to verify the scalability of the proposed method. Data was collected from the driving control of a human subject, and then the Macro-Actions that are required for running a car were generated. Furthermore, the policy that is necessary for driving on a track was learned. The acquisition of Macro-Actions by PbD reduced the driving time by about 16% compared to the case in which Macro-Actions were directly defined by a human subject. In addition, the learning time was also reduced by a faster convergence of the optimum policies.

• Journal of Korea Game Society
• /
• v.11 no.2
• /
• pp.141-149
• /
• 2011

The substantial effort is required to make a virtual character smoothly move like a human being in the virtual environment. Since a virtual character acts on the basis of the actions, it is the most critical to define actions for smooth flow of action. It has been actively studied the programming by demonstration which defines series of actions to be performed by a virtual character based on the actions operated by a person. However, such approaches can't easily draw many sequential actions because they create sequential actions in the same length all the time or restrict the actions used to create such actions. For smooth flow of action, it is required to derive sequential actions as much as possible from the actions of a virtual character and to select the representative set of actions. Accordingly, it is necessary to study how to create sequential actions as reducing diverse limits. This study suggests the approach to select sequential actions suitable for a virtual character by collecting the actions of a character manipulated by a person and deriving a set of actions to be performed by a virtual character. The experiment describes the process to create the actions by applying the approach suggested in this study to the driving game. In accordance with the analysis results, it was found that a set of actions performed by a person was generated without being restricted by a length or a part to divide. Finally, we confirmed that the suggested method selects the best sequential actions, appropriate to virtual character, among more generated actions.

• Journal of Korea Multimedia Society
• /
• v.11 no.3
• /
• pp.353-363
• /
• 2008

A multi-agent system has a longer learning period and larger state-spaces than a sin91e agent system. In this paper, we suggest a new method to reduce the learning time of Nash Q-learning in a multi-agent environment. We apply Macro-actions to Nash Q-learning to improve the teaming speed. In the Nash Q-teaming scheme, when agents select actions, rewards are accumulated like Macro-actions. In the experiments, we compare Nash Q-learning using Macro-actions with general Nash Q-learning. First, we observed how many times the agents achieve their goals. The results of this experiment show that agents using Nash Q-learning and 4 Macro-actions have 9.46% better performance than Nash Q-learning using only 4 primitive actions. Second, when agents use Macro-actions, Q-values are accumulated 2.6 times more. Finally, agents using Macro-actions select less actions about 44%. As a result, agents select fewer actions and Macro-actions improve the Q-value's update. It the agents' learning speeds improve.

• Proceedings of the Korea Contents Association Conference
• /
• 2004.11a
• /
• pp.558-563
• /
• 2004

We develop a method to simulate diverse actions for a virtual world consisting of numerous interrelated objects. We focus on how to efficiently model those countless actions rather realistic representation of each action. To this end, those actions are structured into hierarchies according to their characteristic parameters so actions can be efficiently added based on inheritance. The first key issue in this model is how to extract commonality among actions and identify their idiosyncrasy. The second issue is how to reify or substantiate inheritance of the actions. We apply the developed model to the walk actions of quadruped, and their simulated results are shown in articulated figures.

• Hwankyungkyoyuk
• /
• v.18 no.1 s.26
• /
• pp.97-109
• /
• 2005

This study was motivated to understand ways in which actors (students) perceived internal and external conditions of actions in determining responsibility of so-called environment friendly actions. To understand hidden reasons by which students conduct specific environmental actions more or less frequently may be more important and meaningful for environment teachers than to measure and know the frequency that their students do such actions. This study has been implemented by applying questionnaire survey to 752 ruddle school and high school students and 211 environment teachers. To test the hypothesis of this study that there would be significant differences in awareness of environment friendly actions and barriers to implementing those actions between students and environment teachers, frequency analysis and f-test have been conducted. Primary results of this study can be summarized as follows; It has been found that there were a number of evidences showing significant differences in awareness of environment friendly behavior between students and teachers. In term of action frequency, significant differences have been observed in 11 of the listed 20 actions. Similar resulted have been found in terms of perceived difficult, perceived effectiveness, and perceived guilty. These results seem to support an argument that the current list-making and checklist style of connecting students to so-called environment friendly actions need to be reviewed in a critical manner.

• Journal of the Chungcheong Mathematical Society
• /
• v.19 no.4
• /
• pp.437-443
• /
• 2006

We study free actions of finite groups on the 3-dimensional nilmanifold for Type 1 and classify all such group actions, up to topological conjugacy. This work supplies missing one in [1, Theorem 3.11.].

• Journal of the Chungcheong Mathematical Society
• /
• v.20 no.3
• /
• pp.223-230
• /
• 2007

We study free actions of finite groups on the 3-dimensional nilmanifold and classify all such group actions, up to topological conjugacy. This work generalize Theorem 3.10 of [1].

• Bulletin of the Korean Mathematical Society
• /
• v.58 no.1
• /
• pp.133-146
• /
• 2021

For pseudo-free and recurrent self-similar groups, we show that continuous orbit equivalence of inverse semigroup partial actions implies continuous orbit equivalence of group actions. Conversely, if group actions are continuous orbit equivalent, and the induced homeomorphism commutes with the shift maps on their groupoids, we obtain continuous orbit equivalence of inverse semigroup partial actions.

• Nuclear Engineering and Technology
• /
• v.47 no.2
• /
• pp.187-195
• /
• 2015

Human-induced initiating events, also called Category B actions in human reliability analysis, are operator actions that may lead directly to initiating events. Most conventional probabilistic safety analyses typically assume that the frequency of initiating events also includes the probability of human-induced initiating events. However, some regulatory documents require Category B actions to be specifically analyzed and quantified in probabilistic safety analysis. An explicit modeling of Category B actions could also potentially lead to important insights into human performance in terms of safety. However, there is no standard procedure to identify Category B actions. This paper describes a systematic procedure to identify Category B actions for low power and shutdown conditions. The procedure includes several steps to determine operator actions that may lead to initiating events in the low power and shutdown stages. These steps are the selection of initiating events, the selection of systems or components, the screening of unlikely operating actions, and the quantification of initiating events. The procedure also provides the detailed instruction for each step, such as operator's action, information required, screening rules, and the outputs. Finally, the applicability of the suggested approach is also investigated by application to a plant example.

• Korean Journal of Pharmacognosy
• /
• v.26 no.2
• /
• pp.139-147
• /
• 1995

Crinum asiaticum var. japonicum is a wild plant growing only in Jeju-island, Korea, and in Japan. The whole part of this plant has been known to have the pharmacological actions such as analgesic, anti-inflammatory, platelet-aggregation inhibitory, antitussive, and expectorant. With these assumed actions, the leaves (Crinum folium) of this plant has been used in the folk remedies for arthritis and arthralgia. There is, however, no scientific evidences for the pharmacological actions of Crinum asiaticum var. japonicum. In the present study, the analgesic, anti-inflammatory, and platelet-aggregation inhibitory actions of Crinium folium were evaluated using writhing test, tail-flick test, carrageenin antiedema test, in vitro thromboxane $B_2$ quantitation assay and in vitro platelet aggregation test. In order to obtain the partially purified fraction whose pharmacological action is excellent, the methanol extract of Crinium folium was fractionated consecutively into four biological fractions such as ether, ethyl acetate, butanol, and water fractions and their pharmacological actions of the fractions were investigated. Putting our results together, Crinium folium, especially ethyl acetate fraction was proven to have significant analgesic, anti-inflammatory and platelet-aggregation inhibitory actions by inhibition of prostanoids biosynthesis as one of its mechanism of action.