• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.142-145
• /
• 2003

Many control applications using Neural Network need a priori information about the objective system. But it is impossible to get exact information about the objective system in real world. To solve this problem, several control methods were proposed. Reinforcement learning control using neural network is one of them. Basically reinforcement learning control doesn't need a priori information of objective system. This method uses reinforcement signal from interaction of objective system and environment and observable states of objective system as input data. But many methods take too much time to apply to real-world. So we focus on faster learning to apply reinforcement learning control to real-world. Two data types are used for reinforcement learning. One is reinforcement signal data. It has only two fixed scalar values that are assigned for each success and fail state. The other is observable state data. There are infinitive states in real-world system. So the number of observable state data is also infinitive. This requires too much learning time for applying to real-world. So we try to reduce the number of observable states by classification of states with Self-Organizing Map. We also use neural dynamic programming for controller design. An inverted pendulum on the cart system is simulated. Failure signal is used for reinforcement signal. The failure signal occurs when the pendulum angle or cart position deviate from the defined control range. The control objective is to maintain the balanced pole and centered cart. And four states that is, position and velocity of cart, angle and angular velocity of pole are used for state signal. Learning controller is composed of serial connection of Self-Organizing Map and two Multi-layer Feed-Forward Neural Networks.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.39 no.3
• /
• pp.49-60
• /
• 2002

A systematic design approach for Power distribution network (PDN) is presented aiming at applications to DRAM module designs. Three main stages are comprised in this design approach: modeling and simulation of a PDN based on a two-dimensional transmission line structure employing a partial element equivalent circuit (PEEC); verification of the simulation results through comparison to measured values; and design space scanning with PDN parameters. Impedance characteristics for do-coupling capacitors are analyzed to devise an effective way to stabilize power and ground plane Performance within a target level of disturbances. Self-impedance and transfer-impedance are studied in terms of distance between circuit features and the size of do-coupling capacitors. A simple equation has been derived to find the do-coupling capacitance values yielding impedance lower than design target, and thereby reducing the overall computation time. The effectiveness of the design methodology has been demonstrated using a DRAM module with discrete do-coupling capacitors and a strip structure.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.39 no.9
• /
• pp.12-19
• /
• 2002

Guaranteed Frame Rate(GFR) service is intended to efficiently support TCP/IP traffic in ATM networks. The GFR service not only guarantees a minimum service rate at the frame level, but also supports a fair share of available bandwidth. The original GFR proposal outlined two switch implementation scheme : FIFO Queuing and perVC-Queuing. In general, it has been shown that FIFO Queuing is not sufficient to provide rate guarantees and perVC-Queuing with scheduling is needed. In perVC-Queuing implementation, scheduling algorithm plays key rule to provide rate guarantees and to improve fairness. We proposed a new scheduling algorithm for the GFR service. Proposed algorithm can provide minimum service rate guarantee and fair sharing to GFR VCs. Computer simulation results show that proposed scheduling scheme provide a much better performance in TCP Goodput and fairness than previous scheme.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.53 no.8
• /
• pp.19-27
• /
• 2016

Nowadays the adaptive buffering techniques for burst stream packet configuration and its operation algorithm to save energy in IP core network have been studied. This paper explains the selection method of packet buffering interval for energy saving when configuring burst stream packet at the ingress router in IP core network. Especially the adaptive buffering interval and its implementation scheme are required to improve the energy saving efficiency at the input part of the ingress router. In this paper, we propose the best adaptive buffering scheme that a current buffering interval is adaptively buffering scheme based on the input traffic of the past buffering interval, and analyze its characteristics of energy saving and end-to-end delay by computer simulation. We show the improvement of energy saving effect and reduction of mean delay variation when using an appropriate inverse-function selecting the buffering interval for the configuration of burst stream packet in this paper. We confirm this method have superior properties compared to other method. The proposed method shows that it is less sensitive to the various input traffic type of ingress router and a practical method.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.46 no.3
• /
• pp.116-122
• /
• 2009

Mobile Wireless Sensor Network is an organized collection of sensor nodes and mobile sink nodes, in which the sensor node transmits the signal to the sink node. In real environment, there are many cases in which sinks have mobility caused by the people, the vehicle and etc. Since all nodes in the sensor networks have limited energy, many researches have been done in order to prolong the lifetime of the entire network. In this paper we propose Dynamic Local Update-based Routing Protocol(D-LURP) that prolong the lifetime of the entire network to efficiently maintain frequent location update of mobile sink static sensor nodes in Mobile WSNs. When the sink node moves out of the local broadcasting area the proposed D-LURP configures dynamically the local update area consisted of the new local broadcasting area and the previous dissemination node(DN) and find the path between the DN and the sink node, instead of processing a new discovering path like LURP. In this way the processing of broadcasting sink node's location information in the entire network will be omitted. and thus less energy will be consumpted. We compare the performances of the proposed scheme and existing Protocols.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.6
• /
• pp.154-161
• /
• 2010

The properties of sensor node having a restricted energy in WSN have a difficult in various application fields to apply. Our paper proposed the cluster head selection which is an effective energy in order to manage in wireless sensor network. The proposed algorithm improves an energy efficient and is applied to various network environment considering energy capacity between cluster head and nodes and distance between cluster head and base station(sink node). By using the ns-2 simulator, we evaluate the performance of the proposed scheme in comparison with the original LEACH-C. Experimental results validate our scheme, showing a better performance than original LEACH-C in terms of the number of outliving nodes and the quantity of energy consumption as time evolves.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.6
• /
• pp.146-153
• /
• 2010

For the indoor location, wireless fingerprinting is most favorable because fingerprinting is most accurate among the technique for wireless network based indoor location which does not require any special equipments dedicated for positioning. As fingerprinting method,k-nearest neighbor(KNN) has been widely applied for indoor location in wireless location area networks(WLAN), but its performance is sensitive to number of neighborsk and positions of reference points(RPs). So possibilistic fuzzy c-means(PFCM) clustering algorithm is applied to improve KNN, which is the KNN/PFCM hybrid algorithm presented in this paper. In the proposed algorithm, through KNN,k RPs are firstly chosen as the data samples of PFCM based on signal to noise ratio(SNR). Then, thek RPs are classified into different clusters through PFCM based on SNR. Experimental results indicate that the proposed KNN/PFCM hybrid algorithm generally outperforms KNN and KNN/FCM algorithm when the locations error is less than 2m.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.3
• /
• pp.1-7
• /
• 2007

In order to recognize a vehicle license plate, the region of the license plate should be extracted from a vehicle image. Then, character region should be separated from the background image and characters are recognized using some neural networks with selected feature vectors. Of course, choice of feature vectors which serve as the basis of the character recognition has an important effect on recognition result as well as reduction of data amount. In this paper, we propose a novel feature extraction method in which number images are decomposed into linear combination of eigennumbers and show the validity of this method by applying to the recognition of numeric characters in license plates. The experimental results show the recognition rate of 95.3% for about 500 vehicle images with multi-layer perceptron neural network in the eigennumber space. Compared with the conventional mesh feature, it shows a better recognition rate by 5%.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.20 no.1
• /
• pp.193-198
• /
• 2020

The various uses of the Convolutional Neural Network technology are accelerating the evolution of the computing area, but the opposite is causing serious hardware performance shortages. Neural network accelerators, next-generation memory device technologies, and high-bandwidth memory architectures were proposed as countermeasures, but they are difficult to actively introduce due to the problems of versatility, technological maturity, and high cost, respectively. This study proposes DRAM-based main memory technology that enables read operations to be completed without waiting until the end of the refresh operation using pre-stored XOR bit values, even when the refresh operation is performed in the main memory. The results showed that the proposed technique improved performance by 5.8%, saved energy by 1.2%, and improved EDP by 10.6%.

• Journal of Electrical Engineering and Technology
• /
• v.12 no.2
• /
• pp.689-700
• /
• 2017

This paper presents an effective approach for wind turbine (WT) condition assessment based on the data collected from wind farm supervisory control and data acquisition (SCADA) system. Three types of assessment indices are determined based on the monitoring parameters obtained from the SCADA system. Neural Networks (NNs) are used to establish prediction models for the assessment indices that are dependent on environmental conditions such as ambient temperature and wind speed. An abnormal level index (ALI) is defined to quantify the abnormal level of the proposed indices. Prediction errors of the prediction models follow a normal distribution. Thus, the ALIs can be calculated based on the probability density function of normal distribution. For other assessment indices, the ALIs are calculated by the nonparametric estimation based cumulative probability density function. A Back-Propagation NN (BPNN) algorithm is used for the overall WT condition assessment. The inputs to the BPNN are the ALIs of the proposed indices. The network structure and the number of nodes in the hidden layer are carefully chosen when the BPNN model is being trained. The condition assessment method has been used for real 1.5 MW WTs with doubly fed induction generators. Results show that the proposed assessment method could effectively predict the change of operating conditions prior to fault occurrences and provide early alarming of the developing faults of WTs.