Polymer-based Artificial Skin for the Robotic Application

리후안 경상대학교 대학원 2018 국내박사

Skin, the largest organ of human body, consisting mainly the two layers i.e., outermost epidermis over underlying dermis, protects the interior organs and transduces various tactile stimuli (e.g., mechanical force, vibration, temperature, etc.) from external environment. During social interaction humans extract important informations of these stimuli for their recognition and discrimination purposes. Mimicking of human skin via integration of electronics within the flexible substrate having skin like physical properties (e.g., rheology, tribology, moisture absorptivity, mechanical property, etc.) is a topic of innovative research because of its promise for broad applications in robotics, artificial intelligence, and human–machine interfaces, all of which promote the development of artificial skin (alias, electronic skin (e-skin)). Keeping an eye on future artificial skin for different biomedical applications, here we have introduced the flexible, biocompatible thermoplastic polyurethane (TPU) having human skin simulative hydration effect, which doesn’t require any hazardous solvent for processing purposes but can be thermally processed. Instead of incorporating any external sensing material (e.g., traditionally used piezoelectric component), we have modified the TPU using different techniques such as compounding, supercritical fluid foaming and 3D printing to enabling TPU-based human skin like flexible and positive triboelectric all-in-one e-skin material. Tactile sensations of the e-skin materials based single electrode triboelectric nanogenerator (S-TENG) like devices (which work on the basis of contact electrification and electrostatic induction method) towards the objects of different triboelectric properties viz., skin of bare human finger, polymeric materials (i.e., rubber gloves, cotton gloves, polyimide, polytetrafluoroethylene (PTFE), etc.), ceramic material (i.e., glass), metal (i.e., Al-foil), etc. have been optimized using Oscilloscope and the discrimination (i.e., object manipulation) aptitude of the S-TENGs have been visualized by employing principle component analysis (PCA) on the respective triboelectic responses. S-TENGs were fabricated by attaching one induction electrode (i.e., a copper wire using silver paste) on a single surface of the modified TPU substrates. Firstly, considering the keratin-elastin composition of human epidermis we have developed an e-skin (i.e., TPK) by melt-mixing of bio-waste human hair keratin with TPU, which not only shows skin like mechanical property, tribological property (i.e., friction property), rheological property (i.e., viscoelasticity), morphology (i.e., porosity), hydration effect (i.e., water contact angle), skin like positive triboelectric property, etc., but also the TPK-based S-TENG well discriminates among the various tactile sensations provided by different materials. Secondly, inspired by epidermis-dermis composition of human-skin, we have developed a light-weight, robust, flexible and conformable S-TENG based prototype of bi-layer artificial-skin, by attaching one induction electrode with unfoamed skin layer of microcellular TPU foam, which shows high performance object manupulation towards the different objects. Comparative foaming behavior of eco-friendly supercritical fluids viz., CO2 over N2 under variable temperatures (e.g., 130 °C and 150 °C) and constant pressure (15 Mpa) have been examined to pursue the soft and flexible triboelectric TPU foam. The foam derived by CO2-foaming at 150 °C has been prioritized for development of effective S-TENG. Moreover, TPK-based 3D-printed artificial skin was prepared using fused deposition modelling (FDM) technology. Inspired by the potential performances of both the TPK-film and TPU-foam, 3D-printing technique has been attempted here to fabricate a multilayer skin structure of those TPU-materials towards the development of low-cost, time resolved, robust and embedded e-skin for futuristic robotic applications.

Becoming artificial: H. G. Wells and the scientific discourses of modernism

Quamen, Harvey Noel The Pennsylvania State University 2001 해외박사(DDOD)

My dissertation, “Becoming Artificial: H. G. Wells and the Scientific Discourses of Modernism,” examines how the constructed categories of “artificial” and “natural” function in three seemingly disparate discourses: evolutionary theory, early artificial intelligence and the literature of H. G. Wells. Surprisingly, each discourse depends upon a definition of “artificial” that is distinctly <italic>not</italic> the traditional one of being a “substitute for, or an imitation of, nature.” Indeed, the artificial often exists prior to the natural, illuminating newly emergent dynamics and sometimes even creating patterns and objects that we subsequently misidentify as objects of nature. The artificial, as Michel Foucault has defined it, is that which allows nature to speak for itself. Charles Darwin, for example, patterned his concept of natural selection after the artificial selection of pigeon and cattle breeders; artificial intelligence serves as a magnifying lens that delineates the dynamics of intelligence along a continuum from humans to insects; and H. G. Wells proclaimed that human evolution itself was an “artificial process,” arguing that humans could avoid extinction only by insulating themselves from the relentless forces of Darwinian natural selection. H. G. Wells functions as the central touchstone of the dissertation not only because he chronologically bridges the other two discourses but because his literature stands as an important primary source revealing the rhetorical, ideological, political, technoscientific and historical underpinnings of the concept of artificiality. Wells's notion of artificiality depends fully upon creating an insulating buffer between the forces of nature that exert power upon the body and the artificial forces that engineer technology, culture and even the human mind. Rereading Wells in this manner shows him to be far less a Darwinist than many critics have claimed and it also serves to illuminate a century of Modernism that, from Wells's vantage point, spans backward to Darwin's <italic>Origin of Species</italic> in 1859 as well as forward to the famous Dartmouth Conference on Artificial Intelligence in 1956. Thus, the dissertation contributes not only to Wellsiana and the growing discipline of science studies but also to the intellectual and literary history of that period we have come to call Modern.

2015 개정 <인공지능 수학> 교과서의 '벡터'와 '행렬'의 서술 내용에 대한 분석 및 고찰

이영미 연세대학교 교육대학원 2022 국내석사

본 연구에서는 2015 개정 수학과 교육과정에서 5종으로 출판된 <인공지능 수학> 교과서의 ‘자료의 표현’, ‘자료의 분류’, ‘자료의 처리’ 단원에 대하여 인공지능을 이해하는데 필수적인 수학 개념이자 관련 학습 요소인 ‘벡터’와 ‘행렬’에 대한 정의와 관련 하위 개념들이 어떻게 제시되고 다루어지고 있는지 <고급 수학I>과 <기하>와 비교분석하여 각 출판사의 <인공지능 수학> 교과서의 특징을 파악하고 시사점을 논의하였다. 연구 결과, 벡터의 정의와 벡터의 하부 개념과 관련된 내용제시가 교과서별로 상이하였고, 벡터의 활용을 전개하는데 있어 벡터의 크기, 두 벡터 사이의 거리나 벡터의 내적에 대한 맥락의 수준 및 수학적인 해석에 차이가 있었다. 이를 통해 교과서 별로 벡터와 관련된 개념을 수학 교과의 연계성에 치중하여 설명한 교과서와 수학적 개념과 원리보다는 인공지능과 관련한 지식을 학습하는 것에 초점을 맞춘 교과서가 있다는 것을 확인할 수 있었다. 행렬의 경우, 기본 개념 제시에는 큰 차이가 없었으나 교과서별로 이미지 자료를 처리하는데 있어 활용한 행렬의 하부 개념 유형이나 이용한 행렬의 연산에는 다소 차이가 있음을 확인하였다. 이 연구를 통하여 내실 있는 <인공지능 수학> 교과 운영을 위해 제언하고자 하는 내용을 기술하면 다음과 같다. 첫째, 수학적 용어와 인공지능 관련 용어에 대한 제시와 개념 설명에 있어 통일성과 일관성을 고려할 필요가 있다. 둘째, <인공지능 수학>과 관련한 수학적 개념과 원리를 충분히 연습할 수 있는 내용정리 또는 적용문제를 제시할 필요가 있다. 셋째, <인공지능 수학>을 지도하는데 필요한 인공지능 관련 지식에 대한 연수와 자료를 표현하고 처리하는 원리를 효과적으로 학습하기 위해 공학적 도구를 사용할 수 있는 가이드나 방안이 필요하다. 넷째, <인공지능 수학>은 수학을 인공지능에서의 활용적 측면에 중점을 둠에 따라 교과서별로 수학적 내용을 축소하거나 생략한 경우가 있는데 이는 인공지능 활용에 필수적인 수학적 개념을 놓치거나 학생들에게 오개념을 야기할 수 있으므로 수학적인 개념과 원리에도 초점을 맞출 필요가 있다. This study compares and analyzes the definitions and subconcepts of vector and matrix, which are essential mathematical concepts and related learning elements for understanding artificial intelligence in the 2015 revised mathematics and curriculum. As a result of the study, the definition of vectors and the presentation of contents related to the subconcept of vectors were different by textbook, and there were differences in the size of vectors, the level of context of the distance between the two vectors, and mathematical interpretation. Through this, it was confirmed that there are textbooks that focus on learning knowledge related to artificial intelligence rather than mathematical concepts and principles that focus on the connection of vector-related concepts for each textbook. In the case of matrices, there was no significant difference in the presentation of basic concepts, but it was confirmed that there were some differences in the sub-concept type of matrix used or the operation of the matrix used in processing image data for each textbook. Through this study, the contents to be suggested for substantial <Artificial Intelligence Mathematics> curriculum management are as follows. First, the content related to the operation of the vector needs to be supplemented or further explained. Second, it is necessary to appropriately present content organization or application problems that can sufficiently practice mathematical concepts and principles. Third, there is a need for training on artificial intelligence-related knowledge necessary to guide <Artificial Intelligence Mathematics> and a guide or plan to use engineering tools to effectively learn the principles of expressing and processing text or image data. Finally, <Artificial Intelligence Mathematics> sometimes reduces or omits mathematical content for each textbook as it focuses on the utilization aspect of artificial intelligence, which can lead to missing mathematical concepts essential for using artificial intelligence needs to focus on mathematical concepts and principles.

Triaxial compression test methods for artificial frozen soils

김종찬 Graduate School, Korea University 2015 국내석사

An artificial ground freezing system has been widely studied to obtain high quality undisturbed frozen samples from saturated sand deposits. Triaxial compression tests are conducted to evaluate the strength parameters of specimens using undisturbed frozen samples. In this paper, the results of the triaxial compression test on the undisturbed frozen specimens are discussed. Moreover, the best thawing and restoration method for the undisturbed frozen specimens is explained. The triaxial compression test was conducted using frozen-thawed specimens and unfrozen specimens. This chapter investigates the influence of the freeze-thaw process on the specimens and discusses the observation of the characteristics of shear strength and elastic wave velocity of frozen-thawed specimens and unfrozen specimens. The frozen specimens and unfrozen specimens were prepared with relative densities of 60% and 80%, respectively, using Jumunjin sand. The water pluviation method was used to simulate the fully saturated condition under the groundwater table. During the triaxial compression test, compression and shear waves were measured continuously. The compression wave velocity increased with an increase of the B-value during the saturation phase. During the consolidation and shear phases, the compression wave velocity tended to remain constant. The shear wave velocity decreased with increasing B-value in the process of saturation, but changed in accordance with the change in effective stress in the consolidation and shear phases. The compression wave velocities showed similar values, regardless of the freeze-thaw process, but the shear wave velocities were slightly lower in the frozen-thawed specimens than in the unfrozen specimens. The results showed that the freeze-thaw process had a minor effect on the peak deviatoric stress and shear strength values, and that the process did not affect the internal friction angle. This study demonstrates the characteristics of shear strength and elastic wave velocity of frozen-thawed and unfrozen specimens. The frozen specimen, which is obtained using the artificial ground freezing system, should be thawed and in-situ stress should be applied prior to the triaxial test. The objective of this study is to find the optimum thawing and restoration method that minimizes disturbance during the triaxial test. The tests were conducted according to the drain condition at the thawing phase. Moreover, the restoration of the in-situ condition was divided into thawing and consolidation phases. To simulate the undisturbed frozen specimen, a freezing mold was used. The frozen specimens were manufactured with Jumunjin sand with a relative density of 60%. Elastic wave velocity and axial deformation, which were measured during the triaxial compression test, were used as indicators of the disturbance ratio of the specimen. The result showed that the drained thawing method has less effect on changes of specimen height and elastic wave velocity. Furthermore, the specimen, in which all of the in-situ stress was restored at the consolidation phase, showed the most similar elastic wave velocity to that of before freezing. This study demonstrates that the drain condition at the thawing phase and all stress restoration at the consolidation phase generate less disturbance of the frozen specimen.

THE DIGITIZATION OF NETWORK ROUTING SYSTEM, ARTIFICIAL NEURAL NETWORK AND REINFORCEMENT LEARNING FOR LOGISTICS AUTOMATION

엔당 서울대학교 대학원 2023 국내박사

물류 시스템은 경제성장과 지속가능성을 위해 필수적으로 뿐만 아니라 물류 분야에서 운송 및 유통은 경제성장에 영향을 미칠 수 있다. 그러므로 배전시스템의 효율성은 차량의 수용 능력과 차량수의 이용을 최적화하기 위해 거리와 시간을 최소화하게 배전 경로를 결정함으로써 이루어질 수 있다. 심층 학습은 인간 두뇌의 작동 방식을 모방하고 경험을 통해 학습하는 일련의 신경망인 데다가 인간 두뇌의 작용이 영감을 준다. 인공 지능은 기계가 효율적으로 작동하고 문제를 해결할 수 있도록 한다. 물류 시스템은 상품의 이동 과정으로 설명할 수 있으므로 지속 가능한 경제 발전을 위해서는 효과적이고 효율적인 물류 시스템이 필수적이다. 강화 학습은 적절한 조치를 결정하고 최대 결과를 얻기 위해 상황별로 지도 제작하기로 결정될뿐더러 인간의 학습 능력을 모방하여 환경과의 상호 작용에서 장기적인 이점을 극대화하는 행동을 선택할 수 있다. 이 연구는 운송 네트워크를 구축하고 불확실성을 줄이며 효율성과 지속 가능성을 높이기 위해 전략적으로 변경하는 것을 목표로 합니다. 따라서 이 박사학위논문을 통해서 세 가지 다른 연구를 제시한다. 첫 번째 연구는 물류에서 거리, 시간 및 비용을 최소화하기 위해 유통 경로, 용량을 설정하고 운송의 불확실성을 줄이는 것을 목표로 합니다. 거리와 시간을 줄이고 가장 큰 절감 가치를 기반으로 경로를 생성하기 위해 얼마나 많은 조처를 할 수 있는지를 측정하여 절감을 결정하기 위해 수정된 절감 알고리즘을 정의하였다. 또한 본 연구에서는 트럭, 드론, 전기자동차를 이용한 배송에 대한 모델 시뮬레이션을 이용하여 거리, 시간, 비용에 따른 성능을 비교한다. 물론 이 모델의 시뮬레이션을 통해 운영 최적화를 허용하면서 구현 전에 결정의 영향을 확인할 수 있다. 결과는 운영 전략 영역에서 결정을 내릴 수 있으므로 효율성을 높일 수 있다. 두 번째 연구에서는 인공신경망을 구축하여 화물 예측 및 효율성 향상을 목표로 한다. 본 연구에서는 적재, 거리, 비용, 시간 등의 관련 요인을 고려하여 경로와 효율 예측을 결합한 계획에 따라 선적 수량을 예측하기 위한 인공신경망을 구축한다. 이 결과를 통해서 전략적 공식화 및 의사 결정을 계획하고 개발할 기회를 제공할 수 있다. 마지막으로 세 번째 연구는 운송 경로, 주문, 용량을 설정하고 불확실성을 줄이고 활용도와 효율성을 극대화하기 위해 허브의 어떤 함대가 어느 스포크에 배달해야 하는지를 목표로 합니다. 강화 학습은 역동적인 환경을 이해하고 대응하며, 예상치 못한 초과 주문에 능동적으로 대비하고, 예상되는 상황 변화에 즉시 대응하여 더 나은 결과를 수립할 수 있다. 이 연구에서 심층 강화 학습은 시스템에 대한 관찰을 기반으로 행동을 결정하므로 명령을 수행할 위치를 결정할 수 있고 다른 지점에서 명령을 수행할 수 있다. 따라서 효율성, 전략 및 의사 결정을 향상할 수 있다. 전반적으로 이 박사학위논문의 연구는 인공 지능 시뮬레이션, 인공 신경망 및 강화 학습 구축에 대한 학계 및 관리를 제공한다. 따라서 더 지시되고 집중되고 연결되는 운송 모드의 할당을 설정할 수 있으며 본 연구는 운영전략의 영역에서 의사결정을 수립함으로써 효율성을 높일 수 있을 것이다. 또한, 상품의 흐름을 효과적이고 효율적으로 합리화하면 경제 발전, 사회적 향상 및 지속 가능성을 향상할 수 있다. The logistics system is crucial for economic growth and sustainability. In the field of logistics, transportation and distribution may affect economic growth. Hence, the logistics system can be conducted through establish transportation networks, attenuate uncertainty and strategically alter them to enhance efficiency and sustainability. Deep learning is a set of neural networks that seek to mimic the workings of a human brain and learn from its experiences. The workings of the human brain inspire it, and artificial intelligence may allow machines to work efficiently and solve problems. The logistics system can be described as the process of the movement of goods. Hence, effective and efficient logistics systems are essential for sustainable economic development. Reinforcement learning is determined to decide iv on appropriate action and to map from situation to action to get the maximum results. Reinforcement learning can mimic human learning abilities to choose actions that maximize long-term benefits in their interactions with the environment. The research aims to establish transportation networks, reduce uncertainty and strategically alter them to enhance efficiency and sustainability. Hence, the dissertation establishes three different studies. The first study of this dissertation aims to establish distribution route, capacity and reduce uncertainty of shipment to minimize distance, time and cost in logistics. This research established the modified saving algorithm to determine savings as measured by how much can be done to reduce the distance and time used through linking existing nodes and making a route based on the largest saving value here the distance and time between the source and the destination node. This research establish model simulations for delivery established by truck, drone, and electric vehicles and then compares the performance based on distance, time, and cost. Hence, the model simulations may enable to see the impact of decisions before implementation while allowing for operations optimization. The results may establish decisions in operational strategies, thereby increasing efficiencies. The second study of this dissertation aims to establish artificial neural network to make forecast load of shipment and efficiency enhancement. This research established an artificial neural network for forecasting the number of shipments based on combined route and efficiency forecasting with related factors such as load, distance, cost, v and time. The results may enhance efficiency, provide opportunities for planning, strategic formulation and decision making. The third study of this dissertation aims to establish transportation routes, order, capacity, reduce uncertainty, and which fleet on hub should deliver to which spoke to maximize utilization and efficiency. Reinforcement learning may understand and respond to dynamic environments, proactively prepare for unexpected excess orders and establish better results by responding immediately to expected changes in circumstances. The deep reinforcement learning in this research determined actions based on the observations it has made on the system. Hence, it may determine where to carry the order and may carry the orders from different points. Therefore, these may enhance efficiencies, strategies, and decisions making. Overall, the study provides academia and management on establishing artificial intelligent simulations, artificial neural networks, and reinforcement learning. Hence, it may establish the allocation of transportation modes that become more directed, focused, and connected. Hence, this study may establish decisions in operational strategies, thereby enhancing efficiencies. Furthermore, streamlining the flow of goods effectively and efficiently may enhance economic development, social enhancement, and sustainability.

공간환경 변화에 유연하게 적용할 수 있는 모듈러유닛 기반 MC(Modular Coordination) 설계방안

최정애 국민대학교 테크노디자인전문대학원 2023 국내박사

본 연구의 목적은 모듈러공간의 분석을 통해 핵심 특성을 도출하고, 이를 통해 공간환경의 변화에 유연하고 탄력적으로 적용할 수 있는 설계방법을 제안하는 것이며, 인공지능 기술을 기반으로 모듈러공간의 새로운 설계 체계와 공간환경의 구축 방안을 제시하는 것이다. 이를 위해 문헌분석을 통하여 도시공간의 변화, 공간환경의 변화, 기후환경의 변화, 건축환경의 변화를 분석하였고, 이를 통해 인공지능 및 첨단 기술과 모듈러 시스템 기반의 도입이 중요함을 도출하였다. 그리고, 모듈러공간의 개념 및 체계, 그리고 문제점을 파악하여 앞으로의 모듈러공간의 발전 방향을 제시하였다. 또한, 모듈러공간의 선행연구와 사례분석을 통해 주요 특성 도출하였고, 모듈러공간의 설계를 위해서는 고정성과 변동성이 동시에 가능하도록 설계되어야 함을 파악하였다. 이를 토대로 모듈러유닛의 고정성에는 MC(Modular Coordingaion) 치수를, 변형성에는 형상문법을 적용하여 모듈러유닛을 개발하였고, 실제로 모듈러공간에서 본 연구를 통해 개발된 모듈러유닛을 적용할 수 있는지에 대한 실효성의 파악을 위해 연구모형 프로토타입 디자인을 제시하였다. 그리고, 연구의 확장적 측면에서 설계자와 인공지능의 상호보완적 협업에 대한 방안으로 생성형 인공지능 플랫폼을 활용하여 모듈러공간의 새로운 설계방법과 모듈러공간의 다양한 공간생성의 시각과 접근 방식에 대한 가능성을 제시하였다. 또한, 공간환경뿐 아니라 도시공간까지도 모듈러유닛의 활용 가능성을 파악하기 위해 생성적적대네트워크(GAN)를 활용한 방법을 제안하여, 이를 기반으로 한 미래 공간 활용 측면에서도 모듈러유닛이 유용하게 사용될 공간적 활용 모델로서의 가치와 가능성을 확인하였다. 본 연구에 대한 결론은 다음과 같다. 첫째, 미래의 공간환경은 더욱 다양하게 진화하고, 그에 따라 공간환경 변화에 유연성을 가지고 탄력적으로 적용할 수 있는 설계방안 필요성을 확인하였다. 둘째, 4차 산업혁명과 더불어 건축환경 패러다임의 전환은 더욱 가속화되고 있으며, 그에 따른 모듈러건축 기술의 도입과 인공지능을 활용한 설계방법이 중요하게 작용함을 확인하였다. 셋째, 모듈러공간의 건축적 측면에서 여러 장점에도 불구하고, 모듈러 건축산업의 성장을 저해하는 요소가 있음을 파악하였고, 스마트 건축기술 인력 부족으로 인한 기술적 측면, 유해물질을 배출한다는 인식적 측면, 임시 시설이나 간이 시설로만 사용하는 용도적 측면, 제작사 위주의 설계방법으로 인한 사용자 측면, 단조로운 사각형 형태만을 고수하는 설계적 측면의 문제점이 있었고, 이에 따른 설계방안의 고려가 필요하였다. 넷째, 모듈러공간의 중요 특성을 추출하기 위해 시행된 분석에서, 선행연구와 사례분석을 통해 분석된 결과 유닛, 유연성, 변형성, 다양성, 결합성, 분해성, 이동성, 가변성, 적층성, 확장성이 중요한 요소임을 도출하였다. 이에 따라, 모듈러 구조를 결합, 분해, 이동, 확장, 적층하기 위해서는 접합부 중심의 구조 해석을 통한 고정성이 필요하며, 모듈러 유닛을 이용하여 공간을 유연, 다양, 변형, 가변적으로 사용하기 위해서는 모듈러유닛의 변형성이 동시에 요구되므로 이에 따른 새로운 설계방안을 제시하였다. 다섯째, 모듈러공간의 설계 시 고정성의 MC(Modular Coordination)는 유닛의 일정한 규칙을 통한 표준화의 목적으로 사용되어야 하므로, 이에 따라 기본모듈은 1M(100mm), 증대모듈은 3M(300mm), 수직모듈은 1M(100mm), 보조모듈은 M/2(50mm) 또는 M/4(25mm)로 기준 치수를 준용해야 함을 알 수 있었다. 또한, 모듈러공간의 설계 시 변형성으로 사용된 형상문법 방법론을 통해, 유닛의 규칙을 통한 조합으로 새로운 설계 시스템과 형태를 도출할 수 있었다. 여섯째, 모듈러공간 특성에서 도출된 중요 요소인 고성성과 변형성을 활용하여 새로운 설계방법을 개발하였고, 모듈러공간의 설계를 체계화하여 새로운 공간환경의 설계방안을 구축하였다. 일곱째, 모듈러공간에서 도출된 특성과 개발된 모듈러유닛, 그리고 모듈러유닛 설계방법을 통해 실제로 모듈러공간에 적용할 수 있는지 그 실효성을 파악하기 위해 연구모형 프로토타입의 디자인을 시행하였고, 개발된 모듈러유닛은 대지, 건축의 형태, 설치 위치 등 조건에 구애받지 않고 유연하게 적용할 수 있음을 확인하였다. 마지막으로, 어플리케이션 측면으로 이미지 생성형 인공지능 플랫폼을 통해 모듈러공간의 이미지를 다양하고 효율적으로 생성하여 인공지능을 활용한 모듈러 공간설계 분야의 개발 가능성을 확인하였고, 이를 통해 설계자와 인공지능이 상호보완적으로 협업할 수 있는 모듈러공간의 새로운 설계방법으로의 가능성을 파악하였다. 또한, 생성적적대네트워크(GAN)를 통해, 모듈러공간 단일유닛의 한계를 넘어, 공간환경이 확장되어 도시환경까지 설계될 가능성을 제시함으로써, 미래 공간의 활용 측면에서 유용하게 사용될 공간적 활용 모델로의 가치와 가능성을 볼 수 있었다. This study aims to extract key characteristics through the analysis of modular spaces and propose a design methodology that can be flexibly and resiliently applied to changes in the spatial environment. Additionally, it seeks to present a new design system for modular spaces through artificial intelligence technology and suggest strategies for constructing spatial environments. To achieve this, a literature review was conducted to analyze changes in urban spaces, spatial environments, climate environments, and architectural environments. Through this analysis, the importance of incorporating artificial intelligence and advanced technologies in a modular system was emphasized. Furthermore, a conceptual understanding of modular spaces, their systems, and identified challenges were explored, providing insights into the crucial role of introducing artificial intelligence and advanced technology. Additionally, by examining previous research and case studies related to modular spaces, key features were identified. It was recognized that for the design of modular spaces, a balance between stability and adaptability is essential, allowing for both fixed and flexible elements. In conclusion, this study not only contributes to the understanding of the evolution of modular spaces but also highlights the significance of integrating artificial intelligence and advanced technologies for their future development and effective design. Based on this, we developed modular units by applying Modular Coordination (MC) dimensions for stability and shape grammar for adaptability. To assess the practicality of applying these developed modular units in modular spaces, we presented a research model prototype design. Furthermore, from an expansive perspective of the research, we proposed the use of a generative artificial intelligence platform as a collaborative approach between designers and artificial intelligence. This approach aims to explore new design methods for modular spaces and possibilities for diverse spatial creation in modular spaces. Additionally, to examine the potential utilization of modular units not only in spatial environments but also in urban spaces, we suggested a method using Generative Adversarial Networks (GAN). This method aims to identify the applicability of modular units, confirming their value and potential as a spatial utilization model in future spatial contexts. The conclusion of this study is as follows. Firstly, we have affirmed the necessity for design approaches that can flexibly and adaptively respond to the increasingly diverse evolution of future spatial environments. Secondly, in conjunction with the Fourth Industrial Revolution, the transition in the architectural environment paradigm is accelerating. We have confirmed the importance of introducing modular construction technology and design methods utilizing artificial intelligence in response to this shift. Thirdly, despite the various advantages in the architectural aspects of modular spaces, we have recognized factors hindering the growth of the modular construction industry. Issues include a shortage of personnel with expertise in smart building technologies from a technical perspective, a perceptual aspect related to the emission of harmful substances, a purposive aspect limited to temporary or makeshift facilities, a user-centric aspect resulting from manufacturer-centric design methods, and a design-related aspect of adhering only to monotonous square shapes. Accordingly, consideration of design approaches is needed to address these challenges. Fourthly, through the analysis conducted to extract the crucial characteristics of modular spaces, the results revealed that unit, flexibility, adaptability, diversity, connectivity, decomposability, mobility, variability, layering, and expandability are vital elements. Consequently, we have proposed design approaches considering the need for structural stability through joint-centric structural analysis for combining, decomposing, moving, expanding, and stacking modular structures. Simultaneously, we emphasize the importance of adaptability in modular units to utilize space flexibly, diversely, and variably, providing design recommendations accordingly. Fifthly, in the design of modular spaces, the stability of MC (Modular Coordination) should be utilized for the purpose of standardization through consistent rules for units. Accordingly, it was evident that the base module should adhere to a standard dimension of 1M (100mm), the enlarged module to 3M (300mm), the vertical module to 1M (100mm), and the auxiliary module to M/2 (50mm) or M/4 (25mm). Furthermore, in the design of modular spaces using the methodology of shape grammar for adaptability, new systems and forms could be derived through combinations based on the rules of the units. Sixthly, by utilizing the crucial elements of stability and adaptability derived from the characteristics of modular spaces, a new design methodology was developed. The design of modular spaces was systematized to construct a new design approach for creating a novel spatial environment. Seventhly, to assess the practicality of applying the characteristics of modular spaces, the developed modular units, and the design methodology of modular units, a research model prototype design was implemented. It was confirmed that the developed modular units can be flexibly applied without constraints related to conditions such as site, architectural form, or installation location. Lastly, from an application perspective, we confirmed the potential development in the field of AI-assisted modular space design by efficiently generating diverse images of modular spaces through an image-generating artificial intelligence platform. Through this, we identified the possibility of a new design approach for modular spaces, where designers and artificial intelligence can collaboratively complement each other. Additionally, by utilizing Generative Adversarial Networks (GAN), we presented the potential to go beyond the limits of individual units in modular spaces, extending the design capabilities to urban environments. This highlights the value and potential of serving as a useful spatial utilization model in the future utilization of spatial aspects.

인공지능기반 산업용 협동 로봇의 안전성 확보에 관한 안전기준 및 법제 개선방안 연구 : 국내 산업현장 적용을 중심으로

양은지 연세대학교 공학대학원 2022 국내석사

현재, 4차 산업혁명 시대로 진입함에 따라 산업 현장의 패러다임은 자동화·스마트화로 바뀌고 있으며 그 중심에는 산업용 로봇과 산업용 협동 로봇이 있다. 이러한 흐름 속에서 제조혁신 고도화를 위해 산업 현장에서는 향후 로봇에 인공지능 기술을 융합하여 적용할 가능성이 크다. 산업용 협동 로봇은 산업용 로봇과 달리 산업안전보건기준에 관한 규칙의 제223조(운전 중 위험방지 조건)에 따른 안전펜스 등의 방호장치를 통해 작업자와 물리적 경계를 형성하지 않아도 작업자와 협동작업을 할 수 있어 안전성이 무엇보다 중요하다. 하지만, 인공지능기반 산업용 협동 로봇은 아직 시장 진입 초기 단계에 해당하여 구체적인 안전기준이나 법·제도적 기준이 마련되지 않아 산업용 협동 로봇과 같은 기준을 적용하고 있다. 인공지능기반 산업용 협동 로봇에 전통적인 산업용 협동 로봇의 안전기준을 적용할 때 기존에 고려하지 않았던 위험 요인들이 잔존할 것으로 예측된다. 본 연구에서는 기존의 산업용 협동 로봇에 적용되고 있는 안전기준·법제의 한계점을 파악하고, 개선하기 위한 방안을 고안하고자 자율주행자동차에 적용된 안전기준을 활용하였다. 최종적으로 인공지능기반 산업용 협동 로봇의 현장 작동성을 강화할 수 있는 안전기준과 법·제도적 개선안을 제시함으로써 국내 산업현장의 재해예방에 기여하고자 한다. Currently, as we enter the era of the 4th industrial revolution, the paradigm of the industrial field is changing to automation and smartization, and industrial robots and industrial collaborative robots are at the center. In this trend, there is a high possibility that artificial intelligence technology will be fused and applied to robots in the future in industrial fields to advance manufacturing innovation. Unlike industrial robots, industrial collaborative robots can work cooperatively with workers without forming a physical boundary with them through protective devices such as safety fences in accordance with Article 223 (conditions to prevent danger during operation) of the rules on occupational safety and health standards. Therefore, safety is of paramount importance. However, artificial intelligence-based industrial collaborative robots are still in the early stages of entering the market, and specific safety standards or legal and institutional standards have not been established, so standards such as industrial collaborative robots are applied. When applying the safety standards of traditional industrial collaborative robots to artificial intelligence-based industrial collaborative robots, risk factors that were not previously considered are expected to remain. In this study, the safety standards applied to autonomous vehicles were used to identify the limitations of the existing safety standards and laws applied to collaborative robots and to devise measures to improve them. Finally, it is intended to contribute to the prevention of disasters in domestic industrial sites by presenting safety standards and legal and institutional improvement plans that can strengthen the field operation of artificial intelligence-based industrial collaborative robots.

챗봇 상담이 소비자가 지각하는 판매자에 대한 사회적 실재감과 신뢰에 미치는 영향

진굉위 韓國外國語大學校 大學院 2021 국내석사

Facing opportunities and challenges in the e-commerce environment, online retailers are offering real-time consulting services such as “Live Chat” to customers in order to win in the fierce market competition and low conversion rates. Through this live chat, customers can initiate a chat anytime and anywhere to express their doubts or ask for more information, and the online retailer will give an immediate reply to reassure the customer. Now more and more consumers are willing to use live chat to communicate with online store. Therefore, this study argues that consumers use live chat because they trust it. In addition, with the development of artificial intelligence technologies such as natural language processing and deep learning, many businesses have begun to use AI-based chatbots to provide real-time consulting services to consumers. While live chat is generally well received by consumers, for chatbots, regardless of the extensive use of chatbots by several business leaders in recent years, consumers’ acceptance and its continuance usage remain relatively low. In this study, live chat was divided into two types, namely, ‘chatbots’ and 'human live chat’, aimed explore the differences in social presence and trust in chatbots and human live chat are perceived by consumers in e-commerce. Specifically, there are three research problems. The first is 'Do customers care whether consulting service provider is human or artificial'. It's a question of whether the rapid growth of AI enabled people to break the stereotype that "Chatbots are very foolish", and start to face up to the usefulness of chatbots. The second is ‘How do consumers differently form a level of trust about chatbots and human live chat, and how do these impact their behavior’. Many studies have shown that human live chat can be used to solve the lack of trust, however, we explored whether chatbots can win consumer’ trust. The third is ‘How do consumers differently form a sense of social presence about chatbots and human live chat, and how do these impact their behavior’. Considering that social presence generally has a positive effect on the consumer behavior, we explored the differences between chatbots and human live chat. The questionnaire was distributed online to gather the data required to test the research hypotheses. The sample frame consisted of consumers from the China between 20s and 40s who general use Taobao, JD.com online shopping platforms. The respondents were shown two live chat scenarios(shown in pictures and videos) in turn with the Xiaomi 5G smartphone as the stimulus, and were asked the same questions about social presence, trust and usage intention. The two scenarios highlight that chatbots mainly focus on keyword-oriented explanations compared with human live chat, and, it is difficult to form consensus such as additional questions and emoticons. After insincere responses were excluded, 232 were used for the final analysis. The analysis was performed using the SPSS and AMOS. T-test analysis confirmed that the live chat type variable was successfully manipulated. Frequency analysis was conducted to identify demographic characteristics. Exploratory factors analysis and reliability analysis were performed to verify the validity and reliability of variables. Pearson correlation coefficient was conducted to verify the possibility of multi-collinearity. Finally, structural equation modeling was performed to verify the hypothesis. Through analysis, the results are as follows. First, except the attentional allocation, mutual understanding and emotional connectedness have a significant and positive impact on consumer trust formation and live chat usage intentions. Second, the mediating effect of trust has been demonstrated. Third, the moderating effect of chatbots was not statistically significant, which seems to confirm that people have changed their stereotype about chatbots from thinking they are stupid to embrace their benefits. The implications of this study are as follows. First, this study explored the differences in social presence and trust in chatbots and human live chat. It can contribute much to the implementation of AI-based chatbots which has recently received much attention. Second, when e-commerce operators provide human chatting services to consumers, the chatting provider' understanding ability and language expression ability, etc., to obtain consumer trust and improve live chat usage intentions is very important, which in turn may increase the purchase rate of consumers and bring benefits to the businesses. Third, today's artificial intelligence technology has advanced a lot, but the success or failure of the chatbots service lies in the understanding capacity of the chatbots, so it is necessary to educate the chatbots well. Fourth, businesses that are using chatbots will be able to win consumers' hearts more easily by spreading chatbots counseling services among millennials a breakthrough. Fifth, to maximize profits, managers will be able to combine chatbots with human live chat to provide better service to consumers.

Arguing artificially: Understanding the debates that have shaped cognitive science (Allen Newell, Herbert Simon, Rodney Brooks, Roger Penrose)

Gibson, Keith E The Pennsylvania State University 2003 해외박사(DDOD)

Rhetoricians of science are constantly confronted with certain recurring questions about their field: To what extent and in what ways is there a rhetorical component to scientific knowledge? Can particular scientific fields be identified with characteristic lines of argument and patterns of persuasion? Are there specific kinds of proofs that are persuasive for certain scientific communities? In <italic>Arguing Artificially</italic>, I seek answers to these questions by offering a rhetorical analysis of three key moments of debate in the history of artificial intelligence research that are epitomized in several key documents: (1) Allen Newell and Herbert Simon's “GPS, a Program that Simulates Human Thought” that helped establish the dominance of programmed machines, such as expert systems or Deep Blue; (2) Rodney Brooks's “Intelligence Without Representation” that led to a broader acceptance of learning machines, including neural nets and robotics; and (3) Roger Penrose's <italic> Shadows of the Mind</italic>, which inspired a lively online debate over quantum consciousness. I investigate the way these scientists employed epistemological tools considered by scientists to be non-rhetorical in distinctly rhetorical ways. Newell and Simon, for instance, relied on two inartistic proofs (a psychological protocol and the performance of their program) to craft three distinctly artistic appeals in making their case for the future usefulness of their work. Rodney Brooks used multiple inductive analogies (such as a parable about the quest for AF-Artificial Flight) to build a single deductive enthymeme on which his argument rests. And Roger Penrose and his colleagues were able to deal with multiple audiences (among them quantum physicists, AI researchers, and the general public) as they conducted an online debate on a technical topic. Through this analysis, I expect my findings to be helpful to rhetoricians of science seeking answers to the above questions and to teachers of technical writing struggling with these issues, but I also provide an unusual and, perhaps, valuable perspective to those researchers intimately involved in the quest for artificial intelligence.

Elemental analysis of caries-affected root dentin and artificially demineralized dentin

성영혜 Seoul National University 2016 국내박사

Elemental analysis of caries-affected root dentin and artificially demineralized dentin Younghye Sung Program in Conservative Dentistry Department of Dental Science Graduate School, Seoul National University (Directed by Professor Ho-Hyun Son, D.D.S., M.S.D., Ph.D.) Objectives This study aimed to analyze the mineral composition of naturally- and artificially-produced caries-affected root dentin and to determine the elemental incorporation of resin-modified glass ionomer (RMGI) into the demineralized dentin. Materials and methods Box-formed cavities were prepared on buccal and lingual root surfaces of extracted sound human premolars (n = 15). One cavity was exposed to a microbial caries model using a strain of Streptococcus mutans. The other cavity was subjected to a chemical model under pH cycling. Premolars and molars with root surface caries were used as a natural caries model (n = 15). Outer caries lesion was removed using a carbide bur and a hand excavator under a dying technique and restored with RMGI (Fuji Ⅱ LC, GC Corp.). The weight percentages of calcium (Ca), phosphate (P), and strontium (Sr) and the widths of demineralized dentin were determined by electron probe microanalysis and statistically analyzed using ANOVA and Tukey’s post hoc test (α = 0.05). Results Demineralized surface was observed under SEM in all samples. Artificial models induced greater losses of Ca and P and larger widths of demineralized dentin than did a natural caries model (p < 0.05). Sr was diffused into the demineralized dentin layer from RMGI. Conclusions Both the microbial and the chemical caries models produced similar patterns of mineral composition on the caries-affected dentin. However, within the limitation of this study, the artificial lesions had a relatively larger extent of demineralization than the natural lesions. Sr from RMGI was incorporated into the superficial layer of the caries-affected dentin.