AUV based Precise Seabed Mapping with a Wave Energy-harvesting Surface Vehicle

조한길 포항공과대학교 일반대학원 (창의IT융합공학과) 2019 국내박사

The AUVs have a wide range of applications and are being deployed for various purposes in an oceanographic survey, geoscience, military surveillance, and industrial areas. However, in most cases, the AUVs have preprogrammed a plan to follow a preset route of waypoints and there are few reasoning and adapting for changes against an unexpected situation even in case of the commercial AUVs. To reach a higher intelligence level for AUV technology, the AUVs must perceive the surroundings and infer their current states based on the perceived information. For underwater perception, vision-based sensors are widely used but have limits to use in water due to rapid wavelength-dependent attenuation of light by water. With consideration for the water turbidity, sonars are a generic solution for underwater sensing. Compared to vision-based sensors, the lack of information of sonar data is indisputable: the loss of elevation information, perceptual ambiguity, and a high proportion of outlier, which complicate sonar data processing and three-dimensional (3D) map building. Another issue on AUV exploration is about connectivity. The AUVs should be connected to a network for sharing the data obtained from onboard sensors and intervention for high-level work. Therefore the subsea data can be transmitted to the air only via a relay station on the surface such as relay buoys. To overcome the issues, we propose a sustainable connected AUV system that consists of an AUV and surface vehicle. The AUV is able to perceive the environment regardless of water turbidity. The surface vehicle has affordable electrical payload for long-range data communication and maneuvering for relocation. The two vehicles are interlinked via acoustic communication. For the perception, sonar-based mapping is proposed, and for the electrical payload of the surface vehicle, a novel wave energy harvesting device is developed. First, we present a three-dimensional (3D) mapping method in one-way rectilinear scanning with an autonomous underwater vehicle (AUV) equipped with a forward-looking sonar (FLS) and a profiling sonar (PS). Our approach is to use an additional sonar and fuse acoustic measurements provided by the two sonar sensors. The FLS has a high resolution in a horizontal scan but has an uncertainty in the vertical direction. On the other hand, the PS provides a reliable vertical profile but its beam width is extremely narrow. An initial map is generated by the FLS and refined by combining vertical scan data provided by the PS. Second, a novel surface vehicle was proposed to support a long-term survey of AUV by harvesting wave energy. We proposed a wave energy converter called the wave turbine system (WTS) and verified the feasibility of the proposed system. To verify the proposed mechanism and identify the system parameters, we developed a hydrodynamic model for the WTS and simulated its behavior and power generation capability. From the quantitative simulation, optimal system parameters were analyzed. To check the reliability of the simulation result, we carried out verification tests in a water tank, and the simulation result was verified. Finally, The hardware systems for an AUV named Cyclops and an energy-harvesting surface vehicle were developed. The proposed method is implemented in the developed system and to demonstrate the validity and effectiveness of the proposed method, we conducted a series of tests in a water tank and also at sea. The total system was integrated, and validity was demonstrated through the sea trial.

User Characteristics by News platform and Comment generation : Multiple methods approach for Interest and Interpretation of User

이민구 포항공과대학교 융합대학원 2023 국내석사

Depending on the feature of the media, the characteristics of the users are fixed. As a result, the information selection and interpretation may vary. Therefore, media study should focus on the process of user accepting and interpreting messages rather than simply looking at media as a tool for information delivery. In this study, Internet portal news and Online video platform news are compared from the audiences- centered perspective. Through comparison, we would like to explore whether user characteristics according to media platform appear in comments. For analysis, NAVER and YouTube, the representative of internet portal and Online video platform were analyzed. We collected the same news contents uploaded to each platform and built a dataset of 2,145,698 comments. As a result, it was found that comments of NAVER and YouTube not only concentrated on different news contents by reflecting the characteristics of users, but also there are differences in the trend of comments generation. Also, there are difference in the way how they interpreted the message of news contents. Through this, it was revealed that the characteristics of the user by media platform are actually impact on interest and interpretation of users.

An Analysis on the Consumption Structure of Contemporary Popular Culture with Network Science

이정우 포항공과대학교 융합대학원 2022 국내석사

Globalization and the development of information technology have enabled people in different societies to share their culture and consume cultural products through digital devices. This social change has made contemporary popular culture to transcend the borders between countries and penetrate the daily lives of consumers. Our thesis focused on investigating which social factors affect the consumption structure of contemporary popular culture. We constructed a consumption network of mobile games between countries to reflect the characteristics of contemporary popular culture. Using Hofstede's cultural dimensions theory and Facebook's social connectedness index, we revealed cultural distance and social ties between countries play important roles in shaping the consumption structure.

Influence of News Coverage about Focusing Events by Topic on Legislation and its Factor

박찬웅 포항공과대학교 융합대학원 2022 국내석사

The media and focusing events play a critical role in legislation, and many researchers have studied the impact of media and events on legislation. Most studies have often been limited to specific cases, and only their indirect influence on the bill’s formation has been measured. In this study, I measure the influence of the media on bill formation by topic using news data and legal proposal data. First, I classify the news and bills by topic using Word2vec. Next, I measure the media’s impact on legislation in each topic using the Vector AutoRegression (VAR) model. Last, I construct a regression model that explains the topics’ impacts using the news’ and events’ temporal time-series patterns. The result shows that the impact of media on legislation is different for each topic, and the social-related topic has a significant impact. Moreover, media reporting and focusing events can explain these different impacts by topic, and the result shows the prerequisites of the media’s influence on legislation. 많은 연구자들에 의해 입법과정에서 미디어와 사건사고는 중요한 역할을 한다는 것은 연구되어왔다. 그러나 대부분의 연구는 특정 사례에 국한되는 경우가 많았고, 법안 구성에서 간접적인 영향력만 측정됐다. 본 연구는 뉴스와 법 발의안 데이터 를 활용하여 주제별 법안형성에 대한 언론의 영향력을 측정한다. 먼저 word2vec 를 사용하여 뉴스와 청구서를 주제별로 분류한다. 다음으로, 우리는 VAR 모델을 사용하여 각 주제의 법률에 대한 미디어 영향을 측정한다. 마지막으로, 우리는 커버 리지 주기성과 사건의 영향에 따른 다양한 영향을 설명하는 회귀 모델을 구성한다. 그 결과 법안 형성에 대한 미디어의 영향력은 주제별로 차이를 보였으며, 사회와 관련된 주제에서 유의미한 영향을 미치는 것으로 나타났다. 또한 언론 보도 주기 와 사건사고의 강도는 토픽별로 다르게 나타나는 미디어의 영향력을 설명한다. 이 결과들을 통해 본 연구는 법안에 대한 언론의 영향력의 한계를 보여준다.

Study on the establishment and characterization of a novel CD56dimCD62L+ natural killer cell line with distinct immunostimulatory potential derived from extra-nodal NK/T lymphoma

양현걸 포항공과대학교 일반대학원 (융합생명공학부) 2019 국내박사

자연살해세포는 직접적인 세포 독성과 면역 조절 잠재력을 가진 선천성 림프구이다. 이렇게 특화된 기능들과 몸을 지키는 역할 때문에 자연살해세포에 대한 연구와 활용법에 대한 관심이 커져 왔다. 하지만 높아진 관심에도 불구하고, 자연상태의 자연살해세포들의 특징들, 이를 테면 혈액 속에 적게 존재하고, ex vivo 증식이 제한적이며, 순수한 세포들을 분리하기 위한 기술적 한계 등으로 인해 보다 심도 깊은 연구가 어려웠다. 따라서 영구적인 자연살해세포주를 만드는 것은 이러한 한계들을 극복할 수 있는 해결책이 될 수 있다. 순수한 자연살해세포들을 무한정 공급이 가능하며, 사용하기 쉬울 뿐 아니라, 윤리적 문제에서도 자유롭기 때문에 과학적인 연구뿐 아니라 바이오의학 분야에 있어서도 귀중한 도구로 사용될 수 있다. 연구의 첫번째 파트에서는 새롭게 구축된 자연살해세포주인 NK101을 형태학, 면역표현형, 세포독성, 사이토카인/키모카인 분비의 관점에서 종합적으로 분석하였다. 기본적으로 NK101의 경우 자연적인 자연살해세포와 유사하게 대형과립림프구 세포의 형태와 전형적인 표면 마커 프로필을 보일 뿐 아니라, 독성 과립의 내재 및 자가 변형/손실에 대한 인식 능력과 같은 다른 주요 특징들 역시 가지고 있었다. 흥미롭게도 NK101은 특이적인 CD56dimCD62L+ 표현형을 가지고 있었는데, 이는 자연살해세포의 분화 과정 중, 중간 단계에 해당하는 소그룹의 특징들로 알려져 왔다. 실제로 NK101의 경우 앞서 확인한 면역 표현형뿐만 아니라, 기능적인 부분 역시 CD56dimCD62L+자연살해세포 소그룹을 대표하는 다중 기능 작용기 특성과 유사한 것을 확인되었다. 다시 말해 NK101은 사이토카인 자극에 의해 향상된 분열능력 및 인터페론 감마 분비 촉진을 보일 뿐만 아니라, 암세포를 직접적으로 인지하고 죽일 수 있는 다중 기능 작용기를 가지고 있음을 확인하였다. 이러한 결과들은 NK101이 앞서 언급한 자연적인 자연살해세포들의 여러 한계로 인해 거의 연구되지 못했던 희귀한 다중 기능 자연살해세포 소그룹을 연구하는데 있어 유용한 모델로 쓰일 수 있음을 보여주고 있다. 연구의 두번째 파트에서는 NK101이 종양치료를 위한 세포치료제 플랫폼으로 가능성이 있는지를 연구하였다. 현재까지 임상시험에 들어간 자연살해세포주의 경우 NK-92가 유일하기 때문에, NK101을 세포독성, 사이토카인 분비 특성, 유전자 발현 프로필, 생산성 측면에서 NK-92와 직접적으로 비교하였다. NK101의 경우 NK-92와 비교하여 낮은 세포독성을 보였는데, 이는 상대적으로 낮은 perforin과 granzyme B의 발현 때문으로 보인다. 대신 NK101에서 인터페론 감마 및 TNF-α와 같은 면역 반응을 촉진하는 사이토카인들이 NK-92에 비해 높게 발현되는 것을 확인하였다. 반면, IL-1ra나 IL-10과 같이 면역 반응을 억제하는 사이토카인들의 경우 NK101에서는 거의 발현되지 않는 반면 NK-92에서 매우 높게 발현되었다. 유사한 맥락으로 백혈구의 증식을 긍정적으로 조절하는 유전자들이 NK101에서 높게 발현되는 반면, 반대의 역할, 즉 백혈구의 증식을 억제하는 유전자들의 경우 NK-92에서 높게 발현되는 것을 확인하였다. 이러한 기능성/발현양상의 차이는 면역력이 보존된 4T1 종양 모델에서 잘 나타났다. NK101의 경우 강한 종양-특이적 면역 반응과 함께 NK-92보다 강한 항암 효과를 보였다. 이뿐 아니라 생산성 측면에서 NK-92와 비교해, NK101은 해동 이후 회복이 훨씬 빠를 뿐 아니라, 20일 배양 기준 200배가 넘는 성장 프로필을 보여주었다. 종합적으로, 본 연구는 NK101이라는 새로운 자연살해세포주가 희귀한 CD56dimCD62L+ 소그룹으로서 가지는 차별화된 특징들을 강조할 뿐만 아니라, 이들이 면역항암요법의 새로운 세포치료제로써 가능성이 있음을 시사한다. Natural killer (NK) cells are innate lymphocytes endowed with direct cytotoxicity and immunomodulatory potential. Specialized functions and roles for the host defense gives rise to attention for NK cell study and its applications. However, despite elevated interest in understanding NK cells, characteristics of primary NK cells such as scarcity in blood, limited ex vivo life span, and the technical challenges in isolating pure population constrain further extensive study. Thus, establishing permanent NK cell line could become a solution overcoming those limitations. It is limitless in supply, easy-to-use, no ethical concerns, and homogeneous population, being an invaluable tool not only in scientific research, but also in the field of biomedicine. In the first part of the study, a newly established NK cell line, NK101, was comprehensively characterized with regard to morphology, immunophenotype, cytotoxicity, and cytokines/chemokines secretion. Basically, NK101 resembled major features of natural NK cells including large-granular-lymphocyte morphology, typical surface marker profile, inclusion of cytolytic granules, and capacity of ‘missing-self’ recognition. Interestingly, NK101 had a unique CD56dimCD62L+ phenotype, which has been known as a feature of NK subset in the intermediate stage of differentiation. In agreement with the immunophenotypes, NK101 was verified to have polyfunctional effector properties that are representative of CD56dimCD62L+ NK subset. It displayed enhanced proliferation and interferon-γ secretion upon cytokine stimulation as well as direct cytotoxicity against cancer cells. These findings suggest that NK101 provides a valuable model for studying a unique polyfunctional NK cell subset, which has been little studied due to several limitations of primary NK cells. In the second part of the study, I assessed a potential of NK101 as a cellular platform for cancer treatment. Since NK-92 is only available NK cell line entering clinical trials, NK101 was compared with NK-92 in terms of cytotoxicity, cytokine signature, gene expression profile and manufacturing potential. NK101 expressed lower levels of perforin and granzyme B that correlated with weaker cytotoxicity than NK-92, but produced higher levels of pro-inflammatory cytokines including IFN-γ and TNF-α. On the other hand, anti-inflammatory cytokines such as IL-1 receptor antagonist and IL-10 were highly produced by NK-92, which were nearly undetectable in NK101. Similarly, genes linked to the positive regulation of leukocyte proliferation were enriched in NK101, while those associated with opposite function were highly upregulated in NK-92. Such functional and expressional disparities were well-represented in immunocompetent 4T1 tumor model where NK101 showed more potent anti-tumor effects than those of NK-92, accompanied with stronger tumor-specific immune responses. Regarding manufacturing potential, NK101 not only recovered rapidly after thawing, but also exhibited faster growth profile than NK-92, yielding more than 200-fold higher cell numbers after 20-day culture. Overall, this study not only highlights the distinctive features of a novel NK cell line, NK101, as a unique polyfunctional CD56dimCD62L+ NK subset, but also addresses the capability of NK101 as a new platform for adoptive cancer immunotherapy.

Instagram Post Lifestyle Classification Model for Effective Influencer Marketing

한유정 포항공과대학교 융합대학원 2024 국내석사

With the advent of social media, the 'influencer' has emerged. An influencer is an individual with a significant number of followers on their personal social media account. However, from a corporate perspective, finding multiple influencers suitable for their brand and products remains challenging. Most research into the compatibility between brands and influencers has been predominantly confined to methods such as surveys and case studies to verify effectiveness. In other words, there is a notable absence of technical research focused on identifying the 'aesthetic harmony' between brands and influencers, which is a vital component in influencer marketing. Therefore, this study proposes a scenario-based application case, which involves developing a model that classifies Instagram post lifestyles, using a fine-tuned CLIP model with crawled data. Specifically, for the task of lifestyle classification, we crawled images and texts that can describe five lifestyles using seventy-one adjectives as keywords based on the word-image scale. Secondly, for efficient natural language supervised learning during CLIP fine-tuning, text preprocessing is performed. Ultimately, we demonstrate that our proposed model achieves high performance with 87% accuracy compared to four baseline models. Furthermore, we proposed a process for ranking influencers suitable for brands using the style model, thereby providing a practical guide for its application.

A Study on Improving Recommendation for Unpopular Fashion Items Using Brand Similarity

박수현 포항공과대학교 융합대학원 2024 국내석사

This research aims to improve recommendation performance for unpopular items in the fashion domain. Through the proposed method, we intend to improve the recommendation performance for unpopular items without compromising the recommendation performance for popular items. Our approach uses Word2vec, commonly used for computing not only word similarity but also item similarity, and incorporate brand similarity that is not influenced by individual item popularity. By calculating a weighted average of item similarities and brand similarities associated with those items , we aim to improve recommendation accuracy for rarely clicked items or items with low sales that align with users' preferences, even if they are less popular. Our approach allows us to recommend unpopular items effectively while maintaining the recommendation performance for popular items.

An Analysis of the Effect of Occupational Structure on Employment Growth in the City

이지수 포항공과대학교 융합대학원 2022 국내석사

In this study, we assess factors that affect urban development. As the efficiency of the microscope perspective observing the labor market structure is confirmed, this study traces the external effects of jobs owing to the agglomeration economy. We aim to empirically analyze characteristics of the job structure that affect urban growth and derive implications. We analyzed the labor market with more detailed job and job skills than industry by applying the network methodology and extracted the characteristics of the job structure. Because analyzing the job structure of each metropolitan statistical area (MSA) in the United States by dividing it into three characteristics: job connectivity, clustering coefficient, and diversity, the degree of clustering was confirmed to significantly affect the growth of the employment rate. The above results revealed that some characteristics of the job structure have positive external effects and can be used to create general regional employment development policies. 어떤 요인이 도시 발전에 큰 영향을 끼칠까? 도시의 노동시장의 어떤 특징이 도시 성장에 영향을 끼쳐왔는지에 관한 연구가 활발히 지속되어 오고 있다. 특히, 산업이 도시에 밀집함으로써 얻는 특징에 따라 외부효과를 분석한 연구가 많았다. 노동시장 구조를 미세히 관측하는 것의 효율성을 확인함에 따라, 본 연구에서는 직업이 집적 경제로 인한 외부효과에 대해 관측한다. 직업 구조의 어떤 특징이 도시 성장에 영향을 끼치는지 실증적으로 분석하고 시사점을 도출하는 데에 목적과 의의가 있다. 우리는 네트워크 방법론을 적용하여 노동시장을 산업보다 더 자세한 직업과 직업 스킬로 분석하였고, 여기서 직업 구조의 특징을 추출하였다. 미국 MSA별 직업구조를 직업 연결성, 군집화 계수, 다양성 세 가지 특징으로 구분하여 분석한 결과, 군집화 정도가 고용률 성장에 유의한 영향을 끼친다는 것을 확인하였다. 이상의 결과를 통해 직업 구조의 일부 특징이 긍정적인 외부효과를 가지며, 일반적인 지역 고용 발전 정책을 만드는 데에 활용할 수 있다는 것을 밝혔다.

A Study on Occupational Mismatch of Korean Labor Market: Labor Supply Estimation Model with Qualitative Variables based on Deep Learning

이지인 포항공과대학교 융합대학원 2023 국내석사

This study aims to find the link between people’s perspectives about occupations and labor demand, regarding the labor mismatch situation. Labor mismatch has been one of the prominent issues to be considered in the Korean labor market. The study devised a labor demand estimation model as the solution for the mismatch problem. In the modeling process, people’s perceptions of occupations were regarded as a qualitative factor. Being one of the features used to define labor mismatch, labor demand tends to be affected by the image of occupations, since people tend to consider those images in their future or current job-seeking process. The study first compared the time-variant job image and their matching efficiency scores from 2010 to 2022. Then, it tested the GRU and LSTM-based labor demand prediction model with the qualitative factors and compared the result with the models with just quantitative factors. This study contributes to the field by handling labor mismatch regarding the perceived job image, furthermore, establishing a deep learning-based model that contains those qualitative factors as the solution for the mismatch.

Measuring Democratic Values Oriented by AI News Recommendation Algorithms

황수현 포항공과대학교 융합대학원 2023 국내석사

The market for news in Korea is increasingly using AI. The Korea Press Foundation estimated that 79.2% of Koreans would view news online portal sites in 2021, and Naver and Kakao had already transitioned from human to AI editing of news recommendations in 2017, accounting for more than 90% of the market share for portals. The truth, however, is that there is no established yardstick for judging AI news, and this is accompanied by a lack of awareness of AI’s market share in the news. The purpose of this study was to evaluate AI news recommendations in light of the democratic values that Korean society pursues. We propose three models of democracy, liberal, participatory, and deliberative, to achieve this purpose, in addition to evaluating how well Naver’s and Kakao’s AI-recommended news adhered to the principles of each democracy model, i.e., freedom, social involvement, and discussion. We identified that Kakao is closer to the deliberative democracy model, whereas Naver is closer to the liberal model, such that Naver’s news recommendation algorithm is better suited to guaranteeing people’s freedom of news selection. On the other side, we interpret that under Kakao’s algorithm, members are contributing substantially more to attaining agreement via debate. All democratic values are relative, and we cannot discern between excellent and terrible algorithms merely because they concentrate on a certain value. The key is to understand where algorithms fit into the map of values that each culture pursues, and this study is an effort to do just that. This is also crucial because it sets a precedent for the ultimate objective of putting in place an algorithm that complies with Korean society’s values. 한국 뉴스 시장에서 AI가 차지하는 비중은 갈수록 증가하고 있다. 한국언론진흥재단에 따르면 2021년 한국인의 79.2%는 뉴스를 볼 때 인터넷 포털 사이트를 이용하며, 포털 시장 점유율 90% 이상을 차지하는 네이버와 카카오는 2017년에 이미 뉴스 추천을 사람 편집에서 AI 편집으로 전환했다. 그러나 뉴스 시장에서 AI가 차지하는 비중에 대한 인식이 부족할뿐더러, AI 뉴스를 평가하는 합의된 기준 또한 부재한 것이 현실이다. 이 연구는 AI가 추천한 뉴스를 우리 사회가 추구하는 민주적 가치에 기준하여 평가하기 위한 것이다. 이를 위해 3가지 민주주의 모델 – 자유주의, 참여주의, 심의주의 – 을 상정하고, 네이버와 카카오의 AI 추천 뉴스가 각 민주주의 모델이 추구하는 가치 – 자유, 사회 참여, 토론 및 합의 – 와 얼마나 가까운지, 혹은 먼지를 측정하였다. 실험 결과 네이버는 자유주의 모델에, 카카오는 참여주의와 심의주의 모델에 더 가까운 것으로 나타났다. 이는 네이버 뉴스 추천 알고리즘은 개인이 뉴스를 선택할 자유를 보장하는 데 비교적 더 많이 기여하는 반면, 카카오 뉴스 추천 알고리즘은 구성원들이 사회의 다양한 사안에 참여하고, 토론하여 합의를 이루는 데 상대적으로 더 많이 기여하고 있다는 의미로 해석할 수 있다. 민주적 가치는 모두 상대적이며, 특정 가치에 집중한다는 이유로 좋은 알고리즘과 나쁜 알고리즘을 가를 수 없다. 중요한 것은 각 사회가 추구하는 가치의 지형도에 알고리즘이 어떻게 위치하고 있는지를 인지하는 것이며, 이 연구는 그 인지를 위한 작업이다. 이는 또한 한국 사회의 가치에 부합하는 알고리즘 구현이라는 최종 목표의 선행 작업이라는 데 의의가 있다.