• 한국정보통신학회논문지
• /
• 제25권12호
• /
• pp.1960-1963
• /
• 2021

In this paper, we propose a multiplay Thompson sampling algorithm in multipath communication system. Multipath communication system has advantages on communication capacity, robustness, survivability, and so on. It is important to select appropriate network path according to the status of individual path. However, it is hard to obtain the information of path quality simultaneously. To solve this issue, we propose Thompson sampling which is popular in machine learning area. We find some issues when the algorithm is applied directly in the proposal system and suggested some modifications. Through simulation, we verified the proposed algorithm can utilize the entire network paths. In summary, our proposed algorithm can be applied as a path allocation in multipath-based communications system.

• 한국지능정보시스템학회:학술대회논문집
• /
• 한국지능정보시스템학회 2001년도 The Pacific Aisan Confrence On Intelligent Systems 2001
• /
• pp.65-70
• /
• 2001

We consider a hyper-redundant system that consists of many uniform units. The hyper-redundant system has many degrees of freedom and it can accomplish various tasks. Applysing the reinforcement learning to the hyper-redundant system is very attractive because it is possible to acquire various behaviors for various tasks automatically. In this paper we present a new reinforcement learning algorithm "Q-learning with propagation of motion". The algorithm is designed for the multi-agent systems that have strong connections. The proposed algorithm needs only one small Q-table even for a large scale system. So using the proposed algorithm, it is possible for the hyper-redundant system to learn the effective behavior. In this algorithm, only one leader agent learns the own behavior using its local information and the motion of the leader is propagated to another agents with time delay. The reward of the leader agent is given by using the whole system information. And the effective behavior of the leader is learned and the effective behavior of the system is acquired. We apply the proposed algorithm to a snake-like hyper-redundant robot. The necessary condition of the system to be Markov decision process is discussed. And the computer simulation of learning the locomotion is demonstrated. From the simulation results we find that the task of the locomotion of the robot to the desired point is learned and the winding motion is acquired. We can conclude that our proposed system and our analysis of the condition, that the system is Markov decision process, is valid.

• 한국정밀공학회지
• /
• 제22권8호
• /
• pp.34-41
• /
• 2005

Generally, it is not easy to get a suitable controller for multi variable systems. If the modeling equation of the system can be found, it is possible to get LQR control as an optimal solution. This paper suggests an LQR learning method to design LQR controller without the modeling equation. The proposed algorithm uses the same cost function with error and input energy as LQR is used, and the LQR controller is trained to reduce the function. In this training process, the Jacobian matrix that informs the converging direction of the controller Is used. Jacobian means the relationship of output variations for input variations and can be approximately found by the simple experiments. In the simulations of a hydrofoil catamaran with multi variables, it can be confirmed that the training of LQR controller is possible by using the approximate Jacobian matrix instead of the modeling equation and this controller is not worse than the traditional LQR controller.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권8호
• /
• pp.4255-4269
• /
• 2019

This paper proposes an image generation method using a Multi Discriminator Generative Adversarial Net (MDGAN) as a next generation 2D game sprite creation technique. The proposed GAN is an Autoencoder-based model that receives three areas of information-color, shape, and animation, and combines them into new images. This model consists of two encoders that extract color and shape from each image, and a decoder that takes all the values of each encoder and generates an animated image. We also suggest an image processing technique during the learning process to remove the noise of the generated images. The resulting images show that 2D sprites in games can be generated by independently learning the three image attributes of shape, color, and animation. The proposed system can increase the productivity of massive 2D image modification work during the game development process. The experimental results demonstrate that our MDGAN can be used for 2D image sprite generation and modification work with little manual cost.

• 인지과학
• /
• 제10권1호
• /
• pp.1-11
• /
• 1999

최근 에이전트 기반 시스템 기술은 소프트웨어 시스템의 개념화, 설계, 구현을 위한 새로운 패러다임을 제공하며 많은 기대를 받아왔다. 특히 멀티 에이전트 시스템은 분산적이고 개방적인 인터넷 환경에 잘 부합되는 특징을 가지고 있어서 많은 연구가 진행되고 있다. 멀티 에이전트 시스템에서는 각 에이전트들이 자신의 목적을 위해 행동하기 때문에 에이전트간 충돌이 발생하는 경우에 조정을 통해 협력할 수 있어야 한다. 그러나 기존의 멀티 에이전트 시스템에서의 에이전트 간 협력 방법에 관한 연구 방법들은 동적 환경에서 서로 다른 목적을 갖는 에이전트간의 협동 문제를 올바로 해결할 수 없다는 문제가 있었다. 본 논문에서는 강화학습을 이용한 자동 역할 조정 방법을 통하여 에이전트가 처한 동적 환경에서 서로 다른 목적을 갖는 에이전트간의 협력 문제를 해결한다. 이를 위하여 멀티 에이전트 시스템 분야의 전통적인 문제인 추적 문제에 동적 환경과 서로 다른 목표를 갖는 에이전트들을 모델 링 하여, 두 가지 수정된 추적 문제를 제안하고 이 문제의 해결을 통하여 제안한 방법이 타당함을 보였다.

• 한국지능정보시스템학회:학술대회논문집
• /
• 한국지능정보시스템학회 2003년도 춘계학술대회
• /
• pp.255-263
• /
• 2003

A multi-agent system designed to represent newly deregulated electricity markets in the USA is aimed at testing the capability of the multi-agent model to replicate the observed price behavior in the wholesale market and developing a smart business intelligence which quickly searches the optimum offer strategy responding to the change in market environments. Simulation results show that the optimum offer strategy is to withhold expensive generating units and submit relatively low offers when demand is low, regardless of firm size; the optimum offer strategy during a period of high demand is either to withhold capacity or speculate for a large firm, while it is to be a price taker a small firm: all in all, the offer pattern observed in the market is close to the optimum strategy. From the firm's perspective, the demand-side participation as well as the intense competition dramatically reduces the chance of high excess profit.

• 한국멀티미디어학회논문지
• /
• 제22권11호
• /
• pp.1251-1258
• /
• 2019

Intelligent tutoring system enables users to effectively learn by utilizing various artificial intelligence techniques. For instance, it can recommend a proper curriculum or learning method to individual users based on their learning history. To do this effectively, user's characteristics need to be analyzed and classified based on various aspects such as interest, learning ability, and personality. Even though data labeled by the characteristics are required for more accurate classification, it is not easy to acquire enough amount of labeled data due to the labeling cost. On the other hand, unlabeled data should not need labeling process to make a large number of unlabeled data be collected and utilized. In this paper, we propose a semi-supervised learning method based on feedback variational auto-encoder(FVAE), which uses both labeled data and unlabeled data. FVAE is a variation of variational auto-encoder(VAE), where a multi-layer perceptron is added for giving feedback. Using unlabeled data, we train FVAE and fetch the encoder of FVAE. And then, we extract features from labeled data by using the encoder and train classifiers with the extracted features. In the experiments, we proved that FVAE-based semi-supervised learning was superior to VAE-based method in terms with accuracy and F1 score.

• 한국소음진동공학회:학술대회논문집
• /
• 한국소음진동공학회 1998년도 춘계학술대회논문집; 용평리조트 타워콘도, 21-22 May 1998
• /
• pp.273-279
• /
• 1998

An immune system has powerful abilities such as memory, recognition and learning to respond to invading antigens, and is applied to many engineering algorithm recently. In this paper, the combined optimization algorithm is proposed for multi-optimization problem by introducing the capability of the immune system that controls the proliferation of clones to the genetic algorithm. The optimizing ability of the proposed optimization algorithm is identified by using two multi-peak functions which have many local optimums and optimization of the unbalance response function for rotor model.

• 한국공작기계학회:학술대회논문집
• /
• 한국공작기계학회 1998년도 추계학술대회 논문집
• /
• pp.54-59
• /
• 1998

In order to minimize straightness error of deflected shafts, a automatically bending process control system is designed, fabricated, and studied. The multi-step straightening process and the three-point bending process are developed for the geometric adaptive straightness control. Load-deflection relationship, on-line identification of variations of material properties, on-line springback prediction, and studied for the three-point bending processes. Selection of a loading point supporting condition are derved form fuzzy inference and fuzzy self-learning method in the multi-step straighternign process. Automatic straightening machine is fabricated by using the develped ideas. Validity of the proposed system si verified through experiments.

• 산업기술연구
• /
• 제16권
• /
• pp.197-205
• /
• 1996

In this paper, we propose a new neural architecture. We synthesize the architecture from a combination of structures known as MRCCN (Multi-resolution Radial-basis Competitive and Cooperative Network) and BPN (Backpropagation Network). The proposed neural network is able to improve the learning speed of MRCCN and the mapping capability of BPN. The ability and effectiveness of identifying a ninlinear dynamic system using the proposed architecture will be demonstrated by computer simulation.