• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.7 no.2
• /
• pp.267-277
• /
• 2017

One of fundamental problems in developing robust distributed services is how to achieve distributed consensus agreeing some data values that should be shared among participants in a distributed service. As one of algorithms for distributed consensus, Raft is known as a simple and understandable algorithm by decomposing the distributed consensus problem into three subproblems(leader election, log replication and safety). But, the algorithm dose not mention any types of dynamic configuration of participant peers such as adding new peers to a consensus group or deleting peers from the group. In this paper, we present a new consensus algorithm named Raft-D, which supports the dynamic configuration of participant peers by extending the Raft algorithm. For this, Raft-D manages the additional information maintained by participant nodes, and provides a technique to check the connection status of the nodes belonging to the consensus group. Based on the technique, Raft-D defines conditions and states to deal with adding new peers to the consensus group or deleting peers from the group. Based on those conditions and states, Raft-D performs the dynamic configuration process for a consensus group through the log update mechanism of the Raft algorithm.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.18 no.5
• /
• pp.1080-1086
• /
• 2014

In this paper, we analyze the tradeoff performance of a consensus algorithm when it is applied to the CSMA/CA-based distributed wireless network. The consensus algorithm has a faster convergence speed as the number of cooperating neighbors increases, but the transmission delay on the wireless network increases due to access collisions as the number of cooperating neighbors increases. Therefore, there exists a tradeoff relationship between these two performances and so there exists an optimal number of cooperating neighbors that minimizes the consensus time. The result for the optimal number of neighbors according to the number of nodes that participate in the consensus shows that it is optimal for all nodes to cooperate together in the small-scale network but it is optimal to limit the number of neighbors to a fixed value in the large-scale network with nodes greater than a certain value.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.39B no.5
• /
• pp.323-325
• /
• 2014

The consensus algorithm has a faster convergence speed as the number of cooperating neighbors increases, but the information sharing delay in the wireless network increases due to access collisions as the number of cooperating neighbors increases. Therefore, there exists a tradeoff between these two performances according to node's coverage. In this paper, we present a method of coverage control that minimizes consensus time according to network scale.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2014.10a
• /
• pp.409-411
• /
• 2014

There have been various advanced research on how changes of analysts' profit forecasts affect stock prices. Also, consensus, which is usually drawn by the arithmetic mean of profit forecasts, has been widely harnessed among investors in stock market. Recently, it is emphasized to reflect the internal factors of individual forecasts to raise the accuracy of consensus. Based on national and international research, this study proposes a new methodology in consensus by applying statistically meaningful factors in computation.

• Journal of Advanced Navigation Technology
• /
• v.26 no.4
• /
• pp.226-234
• /
• 2022

With advances in autonomous vehicles and networked control, there is a growing interest in the consensus control of a multi-agents system to control multi-agents with distributed control beyond the control of a single agent. Since consensus control is a distributed control, it is bound to have delay in a practical system. In addition, it is often difficult to have a very accurate mathematical model for a system. Even though a reinforcement learning (RL) method was developed to deal with these issues, it often experiences slow convergence in the presence of large uncertainties. Thus, we propose a slide RL which combines the sliding mode control with RL to be robust to the uncertainties. The structure of a sliding mode control is introduced to the action in RL while an auxiliary sliding variable is included in the state information. Numerical simulation results show that the slide RL provides comparable performance to the model-based consensus control in the presence of unknown time-varying delay and disturbance while outperforming existing state-of-the-art RL-based consensus algorithms.

• Korea Petroleum Association Journal
• /
• s.305
• /
• pp.8-11
• /
• 2017

• Korea Petroleum Association Journal
• /
• s.305
• /
• pp.12-15
• /
• 2017

• TTA Journal
• /
• s.134
• /
• pp.128-129
• /
• 2011

한국인은 다양한 의견을 수렴하여 합의에 도달하는 것에 서투른 것 같다. 의견이 다른 것이야 어찌보면 당연하고 오히려 바람직한 것인데, 입장 차이를 줄이고 원만히 합의하는 것은 말처럼 쉬운 일이 아니다. 좀처럼 합의에 이르지 못하는 사람들을 보면 때로는 그들에게는 '합의'라는 DNA가 없는 것이 아닌가 하는 생각마저 든다.

• The Monthly Diabetes
• /
• s.288
• /
• pp.40-45
• /
• 2013

당뇨병 환자의 암 발병율이 높다는 연구결과가 주목을 받고 있으며, 당뇨병과 암의 관계는 이미 학계에서 어느 정도 컨센서스(consensus)가 이뤄져 있다. 미국 암학회(ACS)와 미국 당뇨병학회(ADA)는 당뇨병과 암발병 사이의 연관성을 인정하였고, 당뇨병과 암에 관한 공동성명(Diabetes Care 2010)에서, 당뇨병이 일부 암의 위힘증가와 관련성이 있음을 발표하였다.

• Journal of the Korea Society of Computer and Information
• /
• v.8 no.4
• /
• pp.35-40
• /
• 2003

In this paper, I raise an issue regarding the relationships between the Election problem and the Consensus problem in asynchronous systems with unreliable failure detectors. First, I describe our system model, and then define Leader Election. After then, I show that the Election problem is harder to resolve than the Consensus problem. Each correct process eventually gets into the state in which it considers only one process to be a leader. Therefore a Perfect Failure Detector is the weakest failure detector which is sufficient to solve the Election. In order to show that the Election problem is harder to resolve than the Consensus problem, I utilize the Reduction protocol in this paper.