• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.4
• /
• pp.143-152
• /
• 2021

It is very important to learn behavioral policies that allow multiple agents to work together organically for common goals in various real-world applications. In this multi-agent reinforcement learning (MARL) environment, most existing studies have adopted centralized training with decentralized execution (CTDE) methods as in effect standard frameworks. However, this multi-agent reinforcement learning method is difficult to effectively cope with in a dynamic environment in which new environmental changes that are not experienced during training time may constantly occur in real life situations. In order to effectively cope with this dynamic environment, this paper proposes a novel multi-agent reinforcement learning system, C-COMA. C-COMA is a continual learning model that assumes actual situations from the beginning and continuously learns the cooperative behavior policies of agents without dividing the training time and execution time of the agents separately. In this paper, we demonstrate the effectiveness and excellence of the proposed model C-COMA by implementing a dynamic mini-game based on Starcraft II, a representative real-time strategy game, and conducting various experiments using this environment.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.26 no.5
• /
• pp.329-335
• /
• 2006

It is necessary to study on the stress concentration experimentally, which is the main reason to avoid mechanical dilapidation and failure, when designing a mechanical structure. Stress concentration factor of a specimen of cantilever beam with a circular hole in the center was measured using both strain gage and photoelastic methods in this paper. In strain-gage measurement, three strain gages along the line near a hole of the specimen were installed and maximum strain was extrapolated from three measurements. In photoelastic measurement, two methods were employed. First, the Babinet-Soleil compensation method was used to measure the maximum strain. Secondly, photoelastic 4-step phase shilling method was applied to observe the strain distribution around the hole. Measurements obtained by different experiments were comparable within the range of experimental error.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.7 no.2
• /
• pp.185-191
• /
• 2003

This paper presents the thermal problems of steel-concrete composite structure and the basic references in placing concrete into the structure. Based on the analytical investigation, application of the composite structure has the effect of highly decreasing the temperature difference between the center section and the surface section of the structure, though its application has little influence on the decrement of the maximum temperature under the same using material and placing temperature. Meanwhile, its application causes the section decrement of a concrete structure, and in the section which the decrement ratio is relatively large, the restraint action by a old concrete structure takes place highly tensile stresses over a tensile strength. And, the stress is concentrated on sharp edges of the steel-concrete boundary. Both restraint action and stress concentration considerably decrease the reduction effect of the temperature difference. Therefore, the prediction of thermal response and the reasonable steps are required through the simulation considering the factors and the sections related to those problems.

• Korean Journal of Construction Engineering and Management
• /
• v.16 no.2
• /
• pp.38-45
• /
• 2015

The risk analysis phase of construction risk management process is subdivided into the qualitative risk analysis that plays a major role, and the quantitative risk analysis acting as a supportive role. The traditional calculation method for quantifying a risk value that has been applied so far is an equation to multiply a probability by an impact simply, but its result shows the low risk value biased distribution. Although another equation that shows the high risk biased distribution as an alternative of traditional method was proposed, both of the low or high risk biased equations do not match with the statistical general knowledge that most natural phenomenons are close to the normal distribution. This study proposes a new risk value calculation method that is concentrated to the moderate risk value. Because the risk value distribution by a new method shows a normal shape similar to natural phenomenon, it helps to choose a middle level not biased to the low or high levels when choosing the level of risk response. Furthermore, it could contribute to improve the flexibility and rationality of risk analysis method by providing additional options for the risk value calculation.

• Journal of the Korean Society for information Management
• /
• v.14 no.1
• /
• pp.183-206
• /
• 1997

The information Retrieval Protocol, Z39.50 is applied to the shared cataloging gateway. The history and functions of the Z39.50 are reviewed and the characteristics and the database update service of Z39.50 are taken into account in the notion of the shared cataloging. And for the test, one of the domestic on-line shared cataloging systems is adopted and newly remodelled using Z39.50 protocol. Two kinds of Z39.50 shared cataloging gateway models (th centralized shared cataloging gateway model and the distributed shared cataloging gateway model) are operated in Web server.

• The Journal of the Korea Contents Association
• /
• v.7 no.8
• /
• pp.66-75
• /
• 2007

The traditional centralized network management method using the SNMP has problems, such as, increase of communication traffic and severe bottleneck phenomena at the specific manager-node caused by a frequent polling mechanism. Several mobile agent based network management methods have been studied to solve and overcome these problem. However, in these mobile agent based methods, as the number of nodes to be managed is increased, the mobile data are increased accumulatively and the effectiveness of the network management is decreased. In this paper, we propose the split management method by using mobile agents to handle these problems. In our method, the nodes are divided into groups to be handled by separate mobile agents. Each mobile agent visits its own nodes and gets management information from them. We demonstrate the effectiveness of our proposed method using experimental results.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2016.05a
• /
• pp.545-545
• /
• 2016

전 세계적으로 여름철 국지성 집중호우 및 태풍으로 인한 피해가 빈번하게 발생하고 있으며, 산지가 많은 국내에서는 산지지역 뿐만 아니라 도심지에서의 토석류 피해 또한 급증하고 있다. 또한 2011년 7월 우면산 토석류, 춘천시 펜션 토석류 등 규모면에서 이전 보다 큰 토석류가 발생되고 있다. 이러한 토석류로 인한 피해를 줄이기 위해서는 토석류의 수치적 해석이 필요하며 현재 다양한 토석류 수치해석 모델을 이용하여 토석류를 분석하고 있다. 또한 토석류는 주로 강우로 인해 발생되기 때문에 토석류 해석을 위해선 강우-유출해석이 선행되어야 한다. 본 연구에서는 청송 풍력단지를 조성하는 면봉산 일대를 대상지역으로 선정하였다. 토석류를 유발시키는 강우를 해석하기 위해 분포형 강우-유출모형인 S-RAT모형을 사용하였으며, 토석류 해석을 위해 토석류 2차원 수치해석모형인 RAMMS모형을 사용하였다. 두 모형의 커플링을 통해 토석류 확산범위 및 유동심, 속도, 충격력을 산정하였다. 이를 통해 유역내 토석류 발생시 동일규모의 사방시설을 토석류 흐름구간에 설치하여 위치에 따른 저감효과를 분석하여 사방시설의 적정위치에 대한 분석을 실시하였다.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.10 no.4
• /
• pp.350-357
• /
• 2017

This study was conducted as a post - study on the development of a centralized controller and a hydraulic lift system including structural analysis and remote control for the development of a vertically elevated car. The safety review was carried out through the structural modification of the elevator lift which was developed during the previous research. 3D modeling was performed with Solidworks, and a model of finite element was created through Hypermesh S / W. In addition, the loading environment of the work vehicle for the evaluation is a condition in which the loading amount is 250 kg per position (total, upper, upper, lower, and lower) on the work table, ), The structural analysis was carried out under the condition that the load was 600 kg, and safety was examined in various aspects. As a result, when the allowable load of 250 kg and the excess load of 600 kg are excluded (except Case-11), the stress level is below the yield strength. In the case of Case-11, there is a region exceeding the yield strength at the center support portion of the safety bar at the upper end even after excluding the component which generates the maximum stress, but it does not affect the safety aspect of the whole structure Respectively. Looking at the deflection results, it can be seen that in all cases the maximum deflection occurs in the same table, and the tendency of sagging in both 250 kg and 600 kg is the same.

• Journal of Korean Society of Transportation
• /
• v.19 no.6
• /
• pp.183-194
• /
• 2001

The centralized traffic information system is to gather and analyze real-time traffic information, to receive traffic information request from user, and to send user processed traffic information such as a path finding. Position information, result of destination search, and other information. In the centralized traffic information system, a server received path-finding requests from many clients and must process clients requests in time. The algorithm of multiple path-finding is needed for a server to process clients request, effectively in time. For this reason, this paper presents a heuristic algorithm that decreases time to compute path-finding requests. This heuristic algorithm uses results of the neighbor nodes shortest path-finding that are computed periodically. Path-finding results of this multiple path finding algorithm to use results of neighbor nodes shortest path-finding are the same as a real optimal path in many cases, and are a little different from results of a real optimal path in non-optimal path. This algorithm is efficiently applied to the general topology and the hierarchical topology such as traffic network. The computation time of a path-finding request that uses results of a neighbor nodes shortest path-finding is 50 times faster than other algorithms such as one-to-one label-setting and label-correcting algorithms. Especially in non-optimal path, the average error rate is under 0.1 percent.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.4
• /
• pp.165-172
• /
• 2020

In this paper we propose an uses on-device-based edge computing technology and big data analysis methods through the use of on-device-based edge computing technology and analysis of big data, which are distributed computing paradigms that introduce computations and storage devices where necessary to solve problems such as transmission delays that occur when data is transmitted to central centers and processed in current general smart factories. However, even if edge computing-based technology is applied in practice, the increase in devices on the network edge will result in large amounts of data being transferred to the data center, resulting in the network band reaching its limits, which, despite the improvement of network technology, does not guarantee acceptable transfer speeds and response times, which are critical requirements for many applications. It provides the basis for developing into an AI-based facility prediction conservation analysis tool that can apply deep learning suitable for big data in the future by supporting intelligent facility management that can support productivity growth through research that can be applied to the field of facility preservation and smart factory industry with integrated hardware technology that can accommodate these requirements and factory management and control technology.