• Journal of the Korean Society for Precision Engineering
• /
• v.14 no.12
• /
• pp.86-94
• /
• 1997

This paper proposes the robot inverse calibration method using a neural networks. A high-order networks called Pi-Sigma networks has been used. The Pi-Sigma networks uses linear summing units in the hidden layer and product unit in output layer. The inverse calibration model which compensates the diff- erence of joint variables only between measuring value and analytic value about the desired pose(position, orientation) of a robot is proposed. The compensated values are determined by using the weights obtained from the learning process of the neural networks previously. To prove the reasonableness, the SCARA type direct drive robot(4-DOF) and anthropomorphic robot(6-DOF) are simulated. It shows that the proposed calibration method can reduce the errors of the joint variables from .+-. 5 .deg. to .+-. 0.1 .deg. .

• Journal of Navigation and Port Research
• /
• v.32 no.8
• /
• pp.589-596
• /
• 2008

In this paper, Artificial Neural Network(ANN) is applied to automatic berthing control for a ship. ANN is suitable for a maneuvering such as ship's berthing, because it can describe non-linearity of the system. Multi-layer perceptron which has more than one hidden layer between input layer and output layer is applied to ANN. Using a back-propagation algorithm with teaching data, we trained ANN to get a minimal error between output value and desired one. For the automatic berthing control of a containership, we introduced low speed maneuvering mathematical models. The berthing control with the structure of 8 input layer units in ANN is compared to 6 input layer units. From the simulation results, the berthing conditions are satisfied, even though the berthing paths are different.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.60 no.12
• /
• pp.2352-2360
• /
• 2011

In this paper, we introduce a comprehensive design methodology of Radial Basis Function Neural Networks (RBFNN) that is based on mechanism of clustering and optimization algorithm. We can divide some clusters based on similarity of input dataset by using clustering algorithm. As a result, the number of clusters is equal to the number of nodes in the hidden layer. Moreover, the centers of each cluster are used into the centers of each receptive field in the hidden layer. In this study, we have applied Fuzzy-C Means(FCM) and K-Means(KM) clustering algorithm, respectively and compared between them. The weight connections of model are expanded into the type of polynomial functions such as linear and quadratic. In this reason, the output of model consists of relation between input and output. In order to get the optimal structure and better performance, Particle Swarm Optimization(PSO) is used. We can obtain optimized parameters such as both the number of clusters and the polynomial order of weights connection through structural optimization as well as the widths of receptive fields through parametric optimization. To evaluate the performance of proposed model, NXT equipment offered by National Instrument(NI) is exploited. The situation awareness system-related intelligent model was built up by the experimental dataset of distance information measured between object and diverse sensor such as sound sensor, light sensor, and ultrasonic sensor of NXT equipment.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.16 no.4
• /
• pp.350-357
• /
• 1991

A speech recognition system based on a neural network, which can be used for telephon number services was tested. Because in Korea two different cardinal number systems, a koreanic one and a sinokoreanic one, are in use, it is necessary that the used systems is able to recognize 22 discret words. The structure of the neural network used had two layers, also a structure with 3 layers, one hidden layreformed of each 11, 22 and 44 hidden units was tested. During the learning phase of the system the so called BP-algorithm (back propagation) was applied. The process of learning can e influenced by using a different learning factor and also by the method of learning(for instance random or cycle). The optimal rate of speaker independent recognition by using a 2 layer neural network was 96%. A drop of recognition was observed by overtraining. This phenomen appeared more clearly if a 3 layer neural network was used. These phenomens are described in this paper in more detail. Especially the influence of the construction of the neural network and the several states during the learning phase are examined.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.2
• /
• pp.189-194
• /
• 2020

This paper analyzes the gradient descent method, which is the one most used for learning neural networks. Learning means updating a parameter so the loss function is at its minimum. The loss function quantifies the difference between actual and predicted values. The gradient descent method uses the slope of the loss function to update the parameter to minimize error, and is currently used in libraries that provide the best deep learning algorithms. However, these algorithms are provided in the form of a black box, making it difficult to identify the advantages and disadvantages of various gradient descent methods. This paper analyzes the characteristics of the stochastic gradient descent method, the momentum method, the AdaGrad method, and the Adadelta method, which are currently used gradient descent methods. The experimental data used a modified National Institute of Standards and Technology (MNIST) data set that is widely used to verify neural networks. The hidden layer consists of two layers: the first with 500 neurons, and the second with 300. The activation function of the output layer is the softmax function, and the rectified linear unit function is used for the remaining input and hidden layers. The loss function uses cross-entropy error.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.38 no.3
• /
• pp.228-235
• /
• 2010

This paper deals with an autonomous flight controller design problem for a tilt-rotor aircraft in rotary-wing mode. The inner-loop algorithm is designed using the output-based approximate feedback linearization. The model error originated from the feedback linearization is cancelled within allowable tolerance by using single-hidden-layer neural network. According to Lyapunov direct stability theory, the adaptive update law is derived to run the neural network on-line, which is based on the linear observer dynamics. Moreover, the outer-loop algorithm is designed to track the trajectory generated from way-point guidance. Especially, heading and flight-path angle line-of-sight guidance are applied to the outer-loop to improve accuracy of the landing tracking performance. The 6-DOF nonlinear simulation shows that the overall performance of the flight control algorithm is satisfactory even though the collective input response shows instantaneous actuator saturation for a short time due to the lack of the neural network and the saturation protection logic in that loop.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.28 no.5C
• /
• pp.452-460
• /
• 2003

In this paper, an equalizer(RNE) using nonlinear multilayer combining techniques in Bayesian equalizer with a structure of radial basis function network is proposed in order to simplify the structure and enhance the performance of the equalizer(RE) using a radial basis function network. The conventional RE Produces its output using linear combining the outputs of the basis functions in the hidden layer while the proposed RNE produces its output using nonlinear combining the outputs of the basis function in the first hidden layer. The nonlinear combiner is implemented by multilayer perceptrons(MLPs). In addition, as an infinite impulse response structure, the RNE with decision feedback equalizer (RNDFE) is proposed. The proposed equalizer has simpler structure and shows better performance than the conventional RE in terms of bit error probability and mean square error.

• Journal of the Korea Society of Computer and Information
• /
• v.13 no.7
• /
• pp.189-197
• /
• 2008

ENS-2 simulator is the most widely-used simulator in the research of wired/wireless communication. Since an IEEE 802.11 module has been implemented in NS-2. much research using this module has been done for various MAC layer functions. The most important problem in terms of performance is the hidden terminal problem. To solve this problem, the solutions using a busy-tone such as BTMA, DBTMA, etc. have been proposed. In order to be able to use a busy-tone, channels using the frequency bandwidth except the frequency bandwidth of the data channel is required for the collision prevention, and a MAC layer has to be able to menage such multiple channels. In this research, we implement a module that can use a busy-tone in NS-2, implement a modified RI-BTMA by applying the implemented module, and verify its performance.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.6
• /
• pp.735-740
• /
• 2012

In this paper, we introduce design methodologies of polynomial radial basis function neural network classifier with the aid of Principal Component Analysis(PCA) and Linear Discriminant Analysis(LDA). By minimizing the information loss of given data, Feature data is obtained through preprocessing of PCA and LDA and then this data is used as input data of RBFNNs. The hidden layer of RBFNNs is built up by Fuzzy C-Mean(FCM) clustering algorithm instead of receptive fields and linear polynomial function is used as connection weights between hidden and output layer. In order to design optimized classifier, the structural and parametric values such as the number of eigenvectors of PCA and LDA, and fuzzification coefficient of FCM algorithm are optimized by Artificial Bee Colony(ABC) optimization algorithm. The proposed classifier is applied to some machine learning datasets and its result is compared with some other classifiers.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.6
• /
• pp.842-849
• /
• 2022

With the development of deep learning technology, there are many cases of using DNNs in embedded systems such as unmanned vehicles, drones, and robotics. Typically, in the case of an autonomous driving system, it is crucial to run several DNNs which have high accuracy results and large computation amount at the same time. However, running multiple DNNs simultaneously in an embedded system with relatively low performance increases the time required for the inference. This phenomenon may cause a problem of performing an abnormal function because the operation according to the inference result is not performed in time. To solve this problem, the solution proposed in this paper first reduces the computation by applying the Tucker decomposition to DNN models with big computation amount, and then, make DNN models run in parallel as much as possible in the unit of hidden layer inside the GPU. The experimental result shows that the DNN inference time decreases by up to 75.6% compared to the case before applying the proposed technique.