• 로봇학회논문지
• /
• 제17권1호
• /
• pp.40-47
• /
• 2022

A reward function suitable for a task is required to manipulate objects through reinforcement learning. However, it is difficult to design the reward function if the ample information of the objects cannot be obtained. In this study, a demonstration-based object manipulation algorithm called stochastic exploration guided by demonstration (SEGD) is proposed to solve the design problem of the reward function. SEGD is a reinforcement learning algorithm in which a sparse reward explorer (SRE) and an interpolated policy using demonstration (IPD) are added to soft actor-critic (SAC). SRE ensures the training of the critic of SAC by collecting prior data and IPD limits the exploration space by making SEGD's action similar to the expert's action. Through these two algorithms, the SEGD can learn only with the sparse reward of the task without designing the reward function. In order to verify the SEGD, experiments were conducted for three tasks. SEGD showed its effectiveness by showing success rates of more than 96.5% in these experiments.

• 한국정밀공학회지
• /
• 제13권10호
• /
• pp.130-138
• /
• 1996

In general, the scattler is observed in fatigue test data due to the nonhomogeneity of a material. Consequently. It is necessary to use the statistical method to describe the fatigue crack growth process precisely. Bogdanoff and Kozin suggested and developed the B-model which is the probabilistic models of cumulative damage using the Markov process in order to describe the damage process. But the B-model uses only constant probability ratior(r), so it is not consistent with the actual damage process. In this study, the r-decreasing model using a monotonic decreasing function is introduced to improve the B-model. To verify the model, thest data of fatigue crack growth of A12024-T351 and A17075-T651 are used. Compared with the empirical distribution of test data, the distribution from the r-decreasing model is satisfactory and damage process is well described from the probabilistic and physical viewpoint.

• Communications for Statistical Applications and Methods
• /
• 제29권6호
• /
• pp.679-694
• /
• 2022

Mortality risk is a significant threat to individual life, and quantifying the risk is necessary for making a national population plan and is a traditionally fundamental task in the insurance and annuity businesses. Like other advanced countries, the sustainability of life pensions and the management of longevity risks are becoming important in Asian countries entering the era of aging society. In this study, mortality and pension value sustainability trends are compared and analyzed based on national population and mortality data, focusing on four Asian countries from 1990 to 2017. The result of analyzing the robustness and accuracy of generalized linear/nonlinear models reveals that the Cairns-Blake-Dowd model, the nonparametric Renshaw-Haberman model, and the Plat model show low stability. The Currie, CBD M5, M7, and M8 models have high stability against data periods. The M7 and M8 models demonstrate high accuracy. The longevity risk is found to be high in the order of Taiwan, Hong Kong, Korea, and Japan, which is in general inversely related to the population size.

• 한국정밀공학회지
• /
• 제21권4호
• /
• pp.120-131
• /
• 2004

In this paper, we propose the method of a motion planning generation in which the movement of the 3-link leg subsystem is constrained to a slider-link and a singular posture can be easily avoided. The proposed method is the jumping control moving in vertical direction which mimics a cat's behavior. That is, it is jumping toward wall and kicking it to get a higher-place. Considering the movement from the point of constraint mechanical system, the robotic system which realizes the motion changes its configuration according to the position and it has several phases such as; ⅰ) an one-leg phase, ⅱ) in an air-phase. In other words, the system is under nonholonomic constraint due to the reservation of its momentum. Especially, in an air-phase, we will use a control method using state transformation and linearization in order to control the landing posture. Also, an iterative learning control algorithm is applied in order to improve the robustness of the control. The simulation results for jumping control will illustrate the effectiveness of the proposed control method.

• 전기학회논문지
• /
• 제62권8호
• /
• pp.1132-1137
• /
• 2013

This paper considers the $H_{\infty}$ control problem for networked control systems(NCSs). In order to solve the problem which comes from discontinuous control signal in NCSs, an approach that discontinuous control signals treat time-varying delayed continuous signals is applied to achieve $H_{\infty}$ stability of NCSs. In addition, randomly occurring packet losses and disturbances are considered by introducing stochastic variables with Bernoulli distribution. Based on Lyapunov stability theory, a new stability condition is obtained via linear matrix inequality formulation to find the $H_{\infty}$ controller which achieves the mean square stability of NCSs. Finally, the proposed method is applied to a numerical example in order to show the effectiveness of our results.

• Earthquakes and Structures
• /
• 제11권3호
• /
• pp.505-516
• /
• 2016

This paper evaluates the scaling of ground motions recorded from nine intermediate-depth earthquakes produced in the Vrancea seismic zone in Romania. The considered ground motion database consists of 363 horizontal recordings obtained on soil classes B and C (according to Eurocode 8). An analysis of the inter- and intra-event spectral accelerations is performed in order to gain information regarding the magnitude and distance scaling of the Vrancea ground motions. The analyses reveal a significant influence of the earthquake magnitude and focal depth on the distance scaling and different magnitude and distance scaling for the two soil classes. A linear magnitude and distance scaling is inferred from the results for the range of magnitudes $5.2{\leq}M_W{\leq}7.1$. The results obtained are checked through stochastic simulations and the influence of the stress drop and kappa values on the ground motion levels is assessed. In addition, five ground motion models which were tested in other studies using recordings from Vrancea earthquakes are analyzed in order to evaluate their corresponding host stress drop and kappa. The results show generally a direct connection between the host kappa values and the host stress drop values. Moreover, all the ground motion models depict magnitude dependent host kappa and stress drop levels.

• 물과 미래
• /
• 제14권1호
• /
• pp.51-58
• /
• 1981

본 연구는 하천일류량을 추계학적으로 분석하여 그 특성을 구명하고 이에 적합한 모의발생 모델을 설정하여 검토하는데 목적을 두고 있다. 한강, 낙동강 및 금강유역의 주요관측지점의 일류량으로부터 통계학적 특성인 Correlogram 및 Spectrum 분석에 의한 시계별 특성을 구하였으며 모의발생모델의 결정에 위해 Markov 모델의 Correlogram을 원자료의 그것과 비교한 결과에 따라 모의 발생 모델의 적합도를 예측할 수 있었으며, 또한 다차수모델의 결정계수 R$$에 대한 해석으로부터 다차수 모델의 차수를 결정했으며, 모의발생에 있어서는 수정 Markov 모델과 다차수 모델을 적용한 결과 대체적으로 2차 수정 Markov 모델과 2차 다차수모델이 최적의 결과를 보여주었다.

• 대한기계학회논문집A
• /
• 제24권3호
• /
• pp.661-666
• /
• 2000

Optimization of the elastic joint of train is performed according to the minimization of ten responses which represent driving safety and ride comfort of train and analyzed by using the each response se surface model from stochastic design of experiments. After the each response surface model is constructed, the main effect and sensitivity analyses are successfully performed by 2nd order approximated regression model as described in this paper. We can get the optimal solutions using by nonlinear programming method such as simplex or interval optimization algorithms. The response surface models and the optimization algorithms are used together to obtain the optimal design of the elastic joint of train. the ten 2nd order polynomial response surface models of the three translational stiffness of the elastic joint (design factors) are constructed by using CCD(Central Composite Design) and the multi-objective optimization is also performed by applying min-max and distance minimization techniques of relative target deviation.

• 전자공학회논문지S
• /
• 제36S권11호
• /
• pp.22-33
• /
• 1999

본 논문에서는 시스템의 차수가 고차이고 잡음과 시간지연이 있으며 파라미터가 변하는 비최소위상 시스템에 적응할 수 있는 다변수 일반화 자기동조 제어기의 설계 하중다항식 계수들을 온-라인으로 조정하는 방법을 제안한다. 다변수 일반화 최소분산 자기동조 제어기의 파라미터는 순환최소자승법으로 추정하고 설계 하중다항식 계수들의 값은 확률근사법인 Robbins-Monro알고리듬을 이용하여 자동 조절하였다. 제안한 다변수 자기동조 방법은 극제한방법보다 간단하고 효과적이다. 컴퓨터 시뮬레이션을 통해 제안한 방법이 시스템의 파라미터가 변하고 시스템의 영점이 단위원 밖에 있는 고차 다변수 시스템에 잘 적응함을 보였다.

• 대한토목학회논문집
• /
• 제6권2호
• /
• pp.57-65
• /
• 1986

본(本) 연구(硏究)는 우량관측소(雨量觀測所)가 미비(未備)된 소유성(小流城)에서 실시간(實時間) 유출예측(流出豫測)을 위해 Kalman filter를 이용했으며 이때의 시스템모형(模型)으로 AR(2)를 택하였다. 시간별(時間別) 유출자료는 영산강유역(榮山江流域)의 나주(羅州) 관측지점(觀測地點)에서 관측된 시간별 유량자료률 이용하였다. 여기서 예측된 모든 결과는 통계적(統計的) 방법으로 분석(分折)한 결과, Kalman filter에 의한 유출예측(流出豫測)을 좋은 결과(結果)를 얻을 수 있었으며 과정모형(過程模型)으로서 AR(2)가 적합한 것을 알 수 있었다. 또한 홍수예측에도 효과적임이 입증되었다.