진동수 기반 내하력 평가모델을 실교량에 접목시켜 실시간으로 그 결과를 추정하기 위해 모니터링 및 평가 시스템 프레임 웍을 제시하고자한다. 실시간 내하력 평가 시스템은 무선 IOT가속도계를 부착하여 유효 상시진동데이터를 원격으로 획득하고 이를 통해 진동수 및 점성비를 분석하여 모델에서 요구하는 성능계수를 결정하고 이 결과들을 종합하여 이전 내하력 결과 대비 현재의 내하력을 추정하는 방법이다.

Collecting a rich but meaningful training data plays a key role in machine learning and deep learning researches for a self-driving vehicle. This paper introduces a detailed overview of existing open-source simulators which could be used for training self-driving vehicles. After reviewing the simulators, we propose a new effective approach to make a synthetic autonomous vehicle simulation platform suitable for learning and training artificial intelligence algorithms. Specially, we develop a synthetic simulator with various realistic situations and weather conditions which make the autonomous shuttle to learn more realistic situations and handle some unexpected events. The virtual environment is the mimics of the activity of a genuine shuttle vehicle on a physical world. Instead of doing the whole experiment of training in the real physical world, scenarios in 3D virtual worlds are made to calculate the parameters and training the model. From the simulator, the user can obtain data for the various situation and utilize it for the training purpose. Flexible options are available to choose sensors, monitor the output and implement any autonomous driving algorithm. Finally, we verify the effectiveness of the developed simulator by implementing an end-to-end CNN algorithm for training a self-driving shuttle.

하천수사용 관리를 적정하게 관리하기 위해서는 객관적인 하천수사용량 자료가 수집되어야 한다. 그러나, 하천수 사용자의 취수여건을 고려하여 유량계 외에 수문조작 등의 간접적인 계측방법을 인정하고 있고, 사용자의 자발적인 보고에 의존함에 따라 자료의 객관화가 어려운 실정이다. 이에 한강홍수통제소에서는 농업용수 사용 비중이 커 하천유량 파악이 어려운 만경강의 고산~봉동 수위관측소 구간에 위치한 어우보 취수로에 V-ADCP를 이용한 계측시설을 설치 및 운영하고 있다. 본 연구에서는 V-ADCP로 측정된 유속을 이용하여 실시간 하천수사용량 산정을 위한 유속분포법의 적 용성을 평가하였다. 이를 위해 Chiu의 2차원 유속분포식의 매개변수 민감도를 분석하고, 실측유량 자료에 기초한 최적 매개변수를 산정하였다. 또한 수위-유량관계법, 지표유속법과 비교 평가하여 유속분포법의 특성을 분석하였다.

In the midst of disaster, such as an earthquake or a nuclear radiation exposure area, there are huge risks to send human crews. Many robotic researchers have studied to send UGVs in order to replace human crews at dangerous environments. So far, two-dimensional camera information has been widely used for teleoperation of UGVs. Recently, three-dimensional information based teleoperations are attempted to compensate the limitations of camera information based teleoperation. In this paper, the 3D map information of indoor and outdoor environments reconstructed in real-time is utilized in the UGV teleoperation. Further, we apply the LTE communication technology to endure the stability of the teleoperation even under the deteriorate environment. The proposed teleoperation system is performed at explosive disposal missions and their feasibilities could be verified through completion of that missions using the UGV with the Explosive Ordnance Disposal (EOD) team of Busan Port Security Corporation.

현재 많은 증강현실 콘텐츠가 존재하지만 두 개 이상의 단말기가 실시간으로 통신하며 동시에 이용하는 증강현실 콘텐츠는 거의 상용화되지 않았다. 본 논문에서는 블루투스 통신을 이용하여 네트워크 증강현실 콘텐츠의 가능성에 대해 연구하려 한다. 이 연구를 통해 네트워크 증강현실 개발과 콘텐츠의 종류를 한층 더 넓힐 수 있을 것이다.

In this paper, real-time damage assessment technology was developed for detection the damage of bridges in real time and the performance of the developed technology was verified by vibration test. Real-time damage assessment technology was developed by combining statistical pattern recognition technology and simulation technology. In order to verify the developed technology, the earthquake response acquisition experiment was conducted according to the cable damage of the model cable-stayed bridge. As a result, it was confirmed that the developed real-time damage assessment technology can provide information on the location of damaged cable.

폐기물 가스화 시스템은 폐기물을 원료로 하여 CO, H2 및 CH4이 주성분인 합성가스를 생산한다. 생산한 합성가스는 정제, 개질 공정을 통하여 발전, 연료, 화학원료 등으로 사용할 수 있다. 폐기물 가스화 시스템은 원료로 폐기물을 사용하고 복합 공정(전처리, 가스화, 정제, 개질, 합성가스 이용)으로 구성되어 있어 안정적으로 합성가스를 생산하기 위해서는 다양한 환경 및 운전변수를 고려한 고급 제어 기술이 필요하다. 일반적으로 폐기물 가스화 발전 시스템 제어는 PLC(Programmable Logic Controller)/DCS(Distributed Control System)와 연결된 HMI(Human Machine Interface)를 통하여 이루어진다. 오퍼레이터는 HMI를 통하여 폐기물 가스화 발전 시스템의 Equipment와 Instrument와 1:1로 매핑 된 정보를 확인하고 제어를 수행한다. 오퍼레이터가 이상 상태 발생 및 운전 조건 변경 상황에서 다양한 운전 변수들의 상관관계를 고려하여 제어하는 것은 불가능하다. 본 연구에서는 폐기물 가스화 발전 시스템의 운전 데이터를 실시간으로 모니터링하고 운전 변수들의 상관관계를 확인할 수 있는 시스템을 개발하였다. 폐기물 가스화 발전 운전 데이터 실시간 상관관계 모니터링 시스템은 운전 데이터를 데이터베이스에 저장하는 기능, 데이터베이스에 저장된 데이터를 조회하는 기능, 데이터베이스에 저장된 운전 데이터의 상관관계를 확인할 수 있는 기능, 조회된 데이터를 저장하는 기능 등으로 구성되어 있다. 운전 데이터를 데이터베이스에 저장하는 기능은 폐기물 가스화 발전 시스템을 제어하는 HMI에 스크립트로 처리하였으며, 다른 기능들은 시스템의 접근성 및 활용도를 고려하여 웹시스템으로 개발하였다.

The river levees along the state owned rivers or local rivers are a major infrastructure of the nation, and it is necessary to be able to quickly and accurately evaluate the safety of each structure in case of a disaster such as flood or earthquake, so as to enable preemptive disaster response. In this study, a “real-time safety assessment method of river levees” was developed to understand the safety of river levee in real time and cope with disaster situations preemptively and effectively, and its applicability was examined. The real-time safety evaluation method of the river embankment developed in this study is based on two systems; ① real-time monitoring system by instruments and ② safety evaluation DB system of river embankment based on fragility curves using numerical analysis results. In order to validate its applicability, the safety of Munsan levee located in the upstream area of Kangjeong-Goryeong Weir was evaluated by applying the proposed method.

This paper proposes real-time image-based damage detection method for concrete structures using deep learning. The proposed method is composed of three steps: (1) collection of a large volume of images containing damage information from internet, (2) development of a deep learning model (i.e., convolutional neural network (CNN)) using collected images, and (3) automatic selection of damage images using the trained deep learning model. The whole procedure of the proposed method has been applied to some figures taken in a real structure. This method is expected to facilitate the regular inspection and speed up the assessment of detailed damage distribution the without losing accuracy.

As drones gain more popularity these days, drone detection becomes more important part of the drone systems for safety, privacy, crime prevention and etc. However, existing drone detection systems are expensive and heavy so that they are only suitable for industrial or military purpose. This paper proposes a novel approach for training Convolutional Neural Networks to detect drones from images that can be used in embedded systems. Unlike previous works that consider the class probability of the image areas where the class object exists, the proposed approach takes account of all areas in the image for robust classification and object detection. Moreover, a novel loss function is proposed for the CNN to learn more effectively from limited amount of training data. The experimental results with various drone images show that the proposed approach performs efficiently in real drone detection scenarios.

One of the most frequently performed tasks in human-robot interaction (HRI), intelligent vehicles, and security systems is face related applications such as face recognition, facial expression recognition, driver state monitoring, and gaze estimation. In these applications, accurate head pose estimation is an important issue. However, conventional methods have been lacking in accuracy, robustness or processing speed in practical use. In this paper, we propose a novel method for estimating head pose with a monocular camera. The proposed algorithm is based on a deep neural network for multi-task learning using a small grayscale image. This network jointly detects multi-view faces and estimates head pose in hard environmental conditions such as illumination change and large pose change. The proposed framework quantitatively and qualitatively outperforms the state-of-the-art method with an average head pose mean error of less than 4.5° in real-time.

As the development of autonomous vehicles becomes realistic, many automobile manufacturers and components producers aim to develop ‘completely autonomous driving’. ADAS (Advanced Driver Assistance Systems) which has been applied in automobile recently, supports the driver in controlling lane maintenance, speed and direction in a single lane based on limited road environment. Although technologies of obstacles avoidance on the obstacle environment have been developed, they concentrates on simple obstacle avoidances, not considering the control of the actual vehicle in the real situation which makes drivers feel unsafe from the sudden change of the wheel and the speed of the vehicle. In order to develop the ‘completely autonomous driving’ automobile which perceives the surrounding environment by itself and operates, ability of the vehicle should be enhanced in a way human driver does. In this sense, this paper intends to establish a strategy with which autonomous vehicles behave human-friendly based on vehicle dynamics through the reinforcement learning that is based on Q-learning, a type of machine learning. The obstacle avoidance reinforcement learning proceeded in 5 simulations. The reward rule has been set in the experiment so that the car can learn by itself with recurring events, allowing the experiment to have the similar environment to the one when humans drive. Driving Simulator has been used to verify results of the reinforcement learning. The ultimate goal of this study is to enable autonomous vehicles avoid obstacles in a human-friendly way when obstacles appear in their sight, using controlling methods that have previously been learned in various conditions through the reinforcement learning.

Sunyoung Lee. 2017. Online Processing of Dependency between an NPI and Negation in Korean. Studies in Modern Grammar 95, 19-36. The present study investigated online processing of syntactic dependency between a negative polarity item (NPI) and its negative licensor in Korean. A total of 32 adult native speakers of Korean participated in the study. The study employed a region-by-region, self-paced reading paradigm to measure online reading times. The results showed that as soon as a parser encountered an NPI in the sentence, it expected a negative element to come up at the earliest potential position. Slow reading times occurred at the first candidate site if such an expectation was not satisfied immediately. The findings of the study support the Filled Gap Effect in English (Fodor 1978; Crain and Fodor 1985; Stowe 1986; Frazier and Clifton 1989) and the Typing Mismatch Effect in Japanese (Miyamoto and Takahashi 2001, 2002); Aoshima et al. 2004; Ueno and Kluender 2010). The results suggest that the human parsers, despite the different structural properties of each language, use the similar sentence processing mechanism.

Solar radiation forecasts are important for predicting the amount of ice on road and the potential solar energy. In an attempt to improve solar radiation predictability in Jeju, we conducted machine learning with various data mining techniques such as tree models, conditional inference tree, random forest, support vector machines and logistic regression. To validate machine learning models, the results from the simulation was compared with the solar radiation data observed over Jeju observation site. According to the model assesment, it can be seen that the solar radiation prediction using random forest is the most effective method. The error rate proposed by random forest data mining is 17%.

This paper aims to develop a wireless unified-maintenance system(WUMS) to obtain real-time structural response of civil structures. The WUMS integrates the GPS module, the strain module and the acceleration module based on the standalone controller, and includes the RF module for remote measurement and the battery for long term operation. In order to evaluate the performance of the WUMS, a response test using a modal shaker and a modal test using a model bridge were performed. The test results using USMS were compared with those measured by wired system. Finally, the developed WUMS was appropriate to obtain the structural response of the civil structure in real-time, unified, and wirelessly.

The Expanded Guide Circle (EGC) method has been originally proposed as the guidance navigation method for improving the efficiency of the remote operation using the sensory information. The previous algorithm is, however, concerned only for the omni-directional mobile robot, so it needs to suggest a suitable one for a mobile robot with non-holonomic constraints. The ego-kinematic transform is a method to map points of R2 into the ego-kinematic space which implicitly represents non-holonomic constraints for admissible paths. Thus, robots with non-holonomic constraints in the ego-kinematic space can be considered as “free-flying object”. In this paper, we propose an effective obstacle avoidance method for mobile robots with non-holonomic constraints by applying EGC method in the ego-kinematic space using the ego-kinematic transformation. This proposed method shows that it works better for non-holonomic mobile robots such as differential-drive robot than the original one. The simulation results show its effectiveness of performance.

최근 실시간으로 모션블러(motion blur)를 위한 연구들은 픽셀당 여러개의 시간 색상을 계산 한 후 평균내는 방식으로 샘플의 수가 적을 경우 아티펙트(artifacts)나 노이즈(noise)가 발생하 는 문제를 가지고 있다. 본 논문은 이러한 문제를 개선하기 위해서 이동궤적 근사 다면체 (motion trail)를 이용한 실시간 모션블러 알고리즘을 제안한다. 본 논문의 알고리즘에서는 현재 프레임과 이전프레임의 삼각형으로 이동궤적 근사 다면체를 만들고 전후 관계(front-to-back) 정렬방법과 시공간차원의 비트연산(bitwise operation)을 적용하여 여러 물체가 겹치는 순간의 가시성 문제를 해결했다. 결과적으로 가려지지 않은 이동궤적 근사 다면체만을 그리기에 매끄 러운 블러 효과를 얻는다.

This paper aims to develop a wireless unified-maintenance system(WUMS) to obtain real-time structural response of civil structures. The WUMS integrates the GPS module, the strain module and the acceleration module based on the standalone controller, and includes the RF module for remote measurement and the battery for long term operation. In order to evaluate the performance of the WUMS, a response test using a modal shaker and a modal test using a model bridge were performed. The test results using USMS were compared with those measured by wired system. Finally, the developed WUMS was appropriate to obtain the structural response of the civil structure in real-time, unified, and wirelessly.

최근 인공지능에 대중의 관심으로 인해, 인공신경망을 사용한 의료영상 처리가 학계와 산업계에서 관심이 커져가고 있다. 딥러닝을 이용한 컨볼루션 신경망은 영상을 효과적으로 표현할 수 있는 것으로 증명되었다. 그러나 학습을 위해서는 고성능 H/W 플랫폼이 요구된다. 따라서 고차원의 많은 학습 샘플을 저사양 H/W 플랫폼에서 학습하는 것은 매우 도전적인 문제이다. 본 논문에서는 온라인 인공 신경망을 사용해 라즈베리파이에서 동작할 수 있는 실시간 신경망 알고리즘을 제안하고자 한다. 다양한 실험 결과를 통해 제안된 방법은 실시간 학습이 가능함을 보여주었다.

폐기물 가스화 시스템은 폐기물을 원료로 하여 CO, H2 및 CH4이 주성분인 합성가스를 생산하고, 생산한 합성가스는 발전, 가스연료, 수송용 연료 및 화학원료 등으로 사용할 수 있다. 폐기물 가스화 시스템의 공정은 폐기물을 원료로 하기 때문에 안정적으로 합성가스를 생산할 수 있도록 제어하는 기술이 필요하다. 특히 폐기물가스화 가스엔진 발전 연계 시스템은 합성가스 엔진으로 공급되는 합성가스의 조성과 발열량이 중요한 공정제어 인자이다. 일반적으로 폐기물 합성가스의 조성은 기기분석(NDIR 등)을 통하여 실시간으로 모니터링 할 수 있으나 주요 성분에 대한 분석만 가능하며 합성가스에 포함된 탄화수소계 연료에 대한 함량을 측정할 수 없음에 따라 합성가스 발열량 측정값 오차가 발생한다. 합성가스의 연료가스 성분을 보다 정확하게 측정하여 합성가스의 발열량을 확인하고 가스엔진 발전효율을 산출하기 위하여 GC 분석을 수행한다. GC 분석 데이터는 GC와 연계된 컴퓨터에서 확인할 수 있고 자체적으로 고유한 형식의 파일로 저장됨에 따라 분석된 데이터로부터 발열량을 확인하기 위해서는 추가적으로 계산을 수행해야 하므로 분석이 완료된 시점에서 실시간으로 발열량을 확인할 수 없다. 폐기물 가스화 발전 시스템의 보다 안정적인 운전을 위하여 GC분석 결과를 중앙제어실 운전자가 실시간 모니터링하여 제어하는 ICT 모니터링 제어 기술을 구축하는 것이 시스템의 제어 효율성을 높일 수 있다. 따라서, 본 연구에서는 중앙제어실에서 폐기물 가스화 발전시스템을 효율적으로 제어하면서 실시간으로 GC분석 데이터와 자동으로 계산된 발열량을 확인할 수 있도록 모니터링 시스템을 개발하였다. GC 분석 데이터 실시간 모니터링 시스템은 Client/Server 구조로 개발하였으며, 모니터링 된 데이터는 데이터베이스로 저장하여 추가적인 분석이 가능하도록 하였다.