• Proceedings of the Korean Society of Computer Information Conference
• /
• 2018.07a
• /
• pp.205-206
• /
• 2018

본 논문에서는 IoT 환경의 무선 센서 네트워크 시스템 상의 효율적인 패킷 전달을 위해 큐러닝(Q-learning)에 기반한 다중 대기열 동적 스케쥴링 기법을 제안한다. 이 정책은 다중 대기열(Multiple queue)의 각 큐가 요구하는 딜레이 조건에 맞춰 최대한 패킷 처리를 미룸으로써 효율적으로 CPU자원을 분배한다. 또한 각 노드들의 상태를 큐러닝(Q-learning)을 통해 지속적으로 상태를 파악하여 기아상태(Starvation)를 방지한다. 제안하는 기법은 무선 센서 네트워크 상의 가변적이고 예측 불가능한 환경에 대한 사전지식이 없이도 요구하는 서비스의 질(Quality of service)를 만족할 수 있도록 한다. 본 논문에서는 모의실험을 통해 기존의 학습 기반 패킷 스케쥴링 알고리즘과 비교하여 제안하는 스케쥴링 기법이 복잡한 요구조건에 따라 유연하고 공정한 서비스를 제공함에 있어 우수함을 증명하였다.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.23 no.4
• /
• pp.11-20
• /
• 1998

This paper deals with a mathematical model and an algorithm for the problem of determining k most vital arcs in the shortest path problem. First, we propose a 0-1 integer programming model for finding k most vital arcs in shortest path problem given the ordered set of paths with cardinality q. Next, we also propose an algorithm for finding k most vital arcs ln the shortest path problem which uses the 0-1 Integer programming model and shortest path algorithm and maximum flow algorithms repeatedly Malik et al. proposed a non-polynomial algorithm to solve the problem, but their algorithm was contradicted by Bar-Noy et al. with a counter example to the algorithm in 1995. But using our algorithm. the exact solution can be found differently from the algorithm of Malik et al.

• The Journal of the Acoustical Society of Korea
• /
• v.32 no.3
• /
• pp.279-285
• /
• 2013

In this paper, we propose a novel speech enhancement method to improve the performance of the conventional global soft decision which is based on the spectral gradient method applied to the ratio of a priori speech absence and presence probability value (q). Conventional global soft decision scheme used a fixed value of q in accordance with the hypothesis assumed, but the proposed algorithm is a technique for improving the speech absence probability which is applied adaptively variable value of q according to the speech presence or absence in the previous two frames and the conditions of the spectral gradient value. Experimental results show that the proposed improved global soft decision method based on the spectral gradient method yields better results compared to the conventional global soft decision technique based on the performance criteria of the ITU-T P. 862 PESQ (Perceptual Evaluation of Speech Quality).

• The KIPS Transactions:PartB
• /
• v.11B no.2
• /
• pp.207-212
• /
• 2004

This paper presents a reinforcement learning framework for stock trading systems. Trading system parameters are optimized by Q-learning algorithm and neural networks are adopted for value approximation. In this framework, cooperative multiple agents are used to efficiently integrate global trend prediction and local trading strategy for obtaining better trading performance. Agents Communicate With Others Sharing training episodes and learned policies, while keeping the overall scheme of conventional Q-learning. Experimental results on KOSPI 200 show that a trading system based on the proposed framework outperforms the market average and makes appreciable profits. Furthermore, in view of risk management, the system is superior to a system trained by supervised learning.

• IEIE Transactions on Smart Processing and Computing
• /
• v.1 no.3
• /
• pp.152-160
• /
• 2012

This paper proposes a frame-level rate control algorithm for low delay video applications to reduce the fluctuations in the bitrate. The proposed algorithm minimizes the bitrate fluctuations in two ways with minimal coding loss. First, the proposed rate control applies R-Q model to all frames including the first frame of every group of pictures (GOP) except for the first one of a sequence. Conventional rate control algorithms do not use any R-Q models for the first frame of each GOP and do not estimate the generated-bit. An unexpected output rate result from the first frame affects the remainder of the pictures in the rate control. Second, a rate-distortion (R-D) cost is calculated regardless of the hierarchical coding structure for low bitrate fluctuations because the hierarchical coding structure controls the output bitrate in rate distortion optimization (RDO) process. The experimental results show that the average variance of per-frame bits with the proposed algorithm can reduce by approximately 33.8% with a delta peak signal-to-noise ratio (PSNR) degradation of 1.4dB for a "low-delay B" coding structure and by approximately 35.7% with a delta-PSNR degradation of 1.3dB for a "low-delay P" coding structure, compared to HM 8.0 rate control.

• Structural Engineering and Mechanics
• /
• v.63 no.6
• /
• pp.751-760
• /
• 2017

In recent years, multifractal-based analysis methods have been widely applied in engineering. Among these methods, multifractal detrended cross-correlation analysis (MFDXA), a branch of fractal analysis, has been successfully applied in the fields of finance and biomedicine. For its great potential in reflecting the subtle characteristic among signals, a structural health monitoring (SHM) system based on MFDXA is proposed. In this system, damage assessment is conducted by exploiting the concept of multifractal theory to quantify the complexity of the vibration signal measured from a structure. According to the proposed algorithm, the damage condition is first distinguished by multifractal detrended fluctuation analysis. Subsequently, the relationship between the q-order, q-order detrended covariance, and length of segment is further explored. The dissimilarity between damaged and undamaged cases is visualized on contour diagrams, and the damage location can thus be detected using signals measured from different floors. Moreover, a damage index is proposed to efficiently enhance the SHM process. A seven-story benchmark structure, located at the National Center for Research on Earthquake Engineering (NCREE), was employed for an experimental verification to demonstrate the performance of the proposed SHM algorithm. According to the results, the damage condition and orientation could be correctly identified using the MFDXA algorithm and the proposed damage index. Since only the ambient vibration signal is required along with a set of initial reference measurements, the proposed SHM system can provide a lower cost, efficient, and reliable monitoring process.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.12 no.4
• /
• pp.361-365
• /
• 2002

오랜 세월동안 위대한 이동수단을 만들어내고자 하는 인간의 꿈은 오늘날 눈부신 각종 운송기구를 만들어 내는 결실을 얻고 있다. 자동차 네비게이션 시스템도 그러한 결실중의 한 예라고 할 수 있을 것이다. 지능적으로 판단하고 정보를 처리할 수 있는 자동차 네비게이션 시스템을 부착함으로써 한 단계 발전한 운송수단으로 진화할 수 있을 것이다. 이러한 자동차 네비게이션 시스템의 단점이라면 한정된 리소스만으로 여러 가지 작업을 수행해야만 하는 어려움이다. 그래서 네비게이션 시스템의 주요 작업중의 하나인 경로를 추출하는 경로추출(Route Planning) 작업은 한정된 리소스에서도 최적의 경로를 찾을 수 있는 지능적인 방법이어야만 한다. 이러한 경로를 추출하는 작업을 하는데 기존에 일반적으로 쓰였던 두 가지 방법에는 Dijkstra s algorithm과 A*algorithm이 있다. 이 두 방법은 최적의 경로를 찾아낸다는 점은 있지만 경로를 찾기 위해서 알고리즘의 특성상 각각, 넓은 영역에 대하여 탐색작업을 해야 하고 또한 수행시간이 많이 걸린다는 단점과 또한 경로를 계산하기 위해서 Heuristic function을 추가적인 정보로 계산을 해야 한다는 단점이 있다. 본 논문에서는 적은 탐색 영역을 가지면서 또한 최적의 경로를 추출하는데 드는 수행시간은 작으며 나아가 동적인 교통환경에서도 최적의 경로를 추출할 수 있는 최적 경로 추출방법을 강화학습의 일종인 Q- Learning을 이용하여 구현해 보고자 한다.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.25 no.9
• /
• pp.14-25
• /
• 2011

The solar vehicle is in the spotlight in the eco-friendly aspect of photovoltaic system using unlimited solar energy. The solar vehicle uses energy of photovoltaic and battery. The solar vehicle uses stored energy in battery when photovoltaic power is lower than consumption power by solar vehicle and if photovoltaic power is higher than consumption power by solar vehicle then photovoltaic power is stored to battery. To improve use efficiency of photovoltaic, the researches about MPPT method to operate maximum power point and interior permanent magnet synchronous motor(IPMSM)drive system using photovoltaic is necessary. This paper proposes MPPT control algorithm for solar vehicle using new fuzzy control(NFC). In this paper, to reduce switching loss, the DC-DC converter is omitted. The NFC controller can be use instead of PO. The NFC controller is performed MPPT control using solar cell voltage and q -axis current of IPMSM. The output of NFC is command q -axis current of IPMSM and this current is operated IPMSM. The response characteristics of algorithm proposed in this paper is compared response characteristics of conventional PO method by PSIM program and validity of this paper prove using this result.

• Journal of Korea Multimedia Society
• /
• v.21 no.10
• /
• pp.1182-1194
• /
• 2018

The Dynamic Adaptive Streaming over HTTP(DASH) is envisioned to evolve to meet an increasing demand on providing seamless video streaming services in the near future. The DASH performance heavily depends on the client's adaptive quality selection algorithm that is not included in the standard. The existing conventional algorithms are basically based on a procedural algorithm that is not easy to capture and reflect all variations of dynamic network and traffic conditions in a variety of network environments. To solve this problem, this paper proposes a novel quality selection mechanism based on the Deep Q-Network(DQN) model, the DQN-based DASH Adaptive Bitrate(ABR) mechanism. The proposed mechanism adopts a new reward calculation method based on five major performance metrics to reflect the current conditions of networks and devices in real time. In addition, the size of the consecutive video segment to be downloaded is also considered as a major learning metric to reflect a variety of video encodings. Experimental results show that the proposed mechanism quickly selects a suitable video quality even in high error rate environments, significantly reducing frequency of quality changes compared to the existing algorithm and simultaneously improving average video quality during video playback.

• Journal of Advanced Navigation Technology
• /
• v.13 no.5
• /
• pp.773-780
• /
• 2009

This study proposed a face region detection algorithm using fuzzy inference of pixel hue and intensity. The proposed algorithm is composed of light compensate and face detection. The light compensation process performs calibration for the change of light. The face detection process evaluates similarity by generating membership functions using as feature parameters hue and intensity calculated from 20 skin color models. From the extracted face region candidate, the eyes were detected with element C of color model CMY, and the mouth was detected with element Q of color model YIQ, the face region was detected based on the knowledge of an ordinary face. The result of experiment are conducted with frontal face color images of face as input images, the method detected the face region regardless of the position and size of face images.