• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.252-255
• /
• 2020

Unsupervised deep learning methods have shown impressive results for the challenging monocular depth estimation task, a field of study that has gained attention in recent years. A common approach for this task is to train a deep convolutional neural network (DCNN) via an image synthesis sub-task, where additional views are utilized during training to minimize a photometric reconstruction error. Previous unsupervised depth estimation networks are trained within a fixed depth estimation range, irrespective of its possible range for a given image, leading to suboptimal estimates. To overcome this suboptimal limitation, we first propose an unsupervised adaptive depth estimation method guided by minimum and maximum (min-max) depth priors for a given input image. The incorporation of min-max depth priors can drastically reduce the depth estimation complexity and produce depth estimates with higher accuracy. Moreover, we propose a novel network architecture for adaptive depth estimation, called the AdaMM-DepthNet, which adopts the min-max depth estimation in its front side. Intensive experimental results demonstrate that the adaptive depth estimation can significantly boost up the accuracy with a fewer number of parameters over the conventional approaches with a fixed minimum and maximum depth range.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.11
• /
• pp.2594-2603
• /
• 1997

A novel method of adaptive beam forming is presented in this paper. The proposed technique provides for a suboptimal beam pattern that increases the Signal to Noise/Interference Ratio (SNR/SIR), thus, eventually increases the capacity of the communication channel, under an assumption that the desired signal is dominant compared to each component of interferences at the receiver, which is precoditionally achieved in Code Division Multiple Access (CDMA) mobile communications by the chip correlator. The main advantages of the new technique are:(1)The procedure requires neither reference signals nor training period, (2)The signal interchoerency does not affect the performance or complexity of the entire procedure, (3)The number of antennas does not have to be greater than that of the signals of distinct arrival angles, (4)The entire procedure is iterative such that a new suboptimal beam pattern be generated upon the arrival of each new data of which the arrival angle keeps changing due tot he mobility of the signal source, (5)The total amount of computation is tremendously reduced compared to that of most conventional beam forming techniques such that the suboptimal beam pattern be produced at vevery snapshot on a real-time basis. The total computational load for generating a new set of weitht including the update of an N-by-N(N is the number of antenna elements) autocovariance matrix is $0(3N^2 + 12N)$. It can further be reduced down to O(11N) by approximating the matrix with the instantaneous signal vector.

• Journal of computational fluids engineering
• /
• v.13 no.4
• /
• pp.96-106
• /
• 2008

We review a series of studies on turbulent skin friction drag reduction in wall-turbulence recently conducted in Japan. First, an identity equation relating the skin friction drag and the Reynolds shearstress (the FIK identity) is introduced. Based on the implication of the FIK identity, a new analytical suboptimal feedback control law requiring the streamwise wall-shear stress only is introduced and direct numerical simulation (DNS) results of turbulent pipe flow with that control is reported. We also introduce DNS of an anisotropic compliant surface and parameter optimization using an evolutionary optimization technique.

• Proceedings of the IEEK Conference
• /
• 2001.09a
• /
• pp.609-612
• /
• 2001

An adaptive sub-optimal method using a preset threshold for combining partial transmit sequence (PTS) of an orthogonal frequency division multiplexing (OFDM) signal is presented. The results show that adaptive sub-optimal method reduces the 0.1% PAP by 3.4dB, while iterative flipping method reduces it by 3dB and PTS reduce it by 4.1dB. The complexity of adaptive sub-optimal method is 2.4% of that of PTS method for M=8.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.6 no.2
• /
• pp.191-195
• /
• 1982

연속계의 동적 시스템을 제어하고저 할 때에는 구동장치와 측정장치를 어느 위치에 두느냐에 따라 시스템의 제어 성능이 상당히 달라지기 때문에 위치선정 문제가 매우 중요한 설계변수의 하나이다. 본 연구에서는 일반 연속계의 제어시스템을 설계하고저 할 때를 고려하여, 그들의 최적 위치를 결정할 수 있는 조건식을 최적 제어 이론을 바탕으로 유도하였다.

• Proceedings of the KIEE Conference
• /
• 1991.07a
• /
• pp.651-656
• /
• 1991

A mixed $H_2/H{\infty}$ design methodology is proposed for a single-input-single-output system. The reciprocal stability margin(RSM) function is defined for the perturbed plant with additive uncertainties. A suboptimal $H_2$ controller is sought to minimize the RSM function.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.13 no.2
• /
• pp.25-33
• /
• 1988

An efficient heuristic solution procedure is developed for the minimum location problems. The gradient direction method and modified gradient approach are developed due to the differentiability of the objective functions. Suboptimal step size is obtained analytically. A Modified Gradient Procedure (NGP) is presented and compared with the hyperboloid approximation procedure (HAP) which is one of the best known methods.

• 제어로봇시스템학회:학술대회논문집
• /
• 1993.10a
• /
• pp.486-490
• /
• 1993

This paper presents an improved algorithm which enables to find a suboptimal $H^{\infty}$ controller. In the $H^{\infty}$ control problem with output multiplicative uncertainty, the Glover-Doyle algorithm has sorne constraints for the standard plant. The proposed algorithm removes them by reformulating the standard plant. We show the validity of this algorithm by investigating the variation of norm-bound.d.

• 제어로봇시스템학회:학술대회논문집
• /
• 1993.10a
• /
• pp.540-545
• /
• 1993

When modeling an underwater vehicle uncertainty arises in the presence of unsteady flow. It is difficult to include the uncertainty in the model and is therefore desirable to investigate robust controller design methods for the underwater vehicle. In the paper two robust control methods are applied for the underwater system. One is standard H$_{\infty}$ control and the other is time-varying sliding mode control with modified saturation function. Suboptimal design parameters for H$_{\infty}$ control and design parameters for time-varying switching surfaces are provided. Simulations and comparison are carried out.t.

• 제어로봇시스템학회:학술대회논문집
• /
• 1996.10b
• /
• pp.32-35
• /
• 1996

In this paper, it is shown that the dominant errors of baro-altimeters can be characterized by bias and scale factor errors. Also an optimal filter for estimating both bias and scale factor is derived based on the concept of model transition. The optimal filter is, however, not realizable because the model transition hypotheses increase exponentially. Therefore a realizable suboptimal filter using the interacting multiple model(IMM) technique is proposed. Computer simulation results show that the estimation errors of the proposed filter are smaller than those of the conventional least squares algorithm with a forgetting factor when both the bias and the scale factor are varying.