Diverse and Scalable Skill Acquisition for Robot Manipulation

Xu, Zhenjia ProQuest Dissertations & Theses Columbia Universit 2024 해외박사(DDOD)

The acquisition of capable robot manipulation skills is a critical prerequisite for the widespread deployment of robots in real-world environments, from household tasks to industrial applications. However, current robot manipulation systems remain limited in their ability to handle the diversity of objects, materials, and manipulation actions required in the real world. Data-driven methods have shown impressive results toward generalizing across a variety of problems, but existing approaches often require costly data collection using real robot platforms, hindering the scalability of skill acquisition.In this dissertation, we aim to push the limits of robotic manipulation task diversity by providing mechanisms to acquire new skills in a scalable manner. Achieving the "right" data with large quantity and high quality is of vital importance. We approach this problem by leveraging physics simulators. Different from commonly used rigid body simulators, we have customized simulators to support deformable objects with diverse materials and dynamics. Aerodynamics and fracture effects are also included to enable a wider range of manipulation actions such as blowing and cutting. With sophisticated system design, including proper representation selection and customized hardware design, the policies trained in simulation can be seamlessly applied to real robots.More specifically, this dissertation presents a series of works to address the challenges of diversity and scalability in robot manipulation skill acquisition. First, we introduce UMPNet, a universal policy network that can infer closed-loop action sequences for manipulating a wide range of articulated objects using only visual input. Second, we present DextAIRity, a system that leverages active airflow to enable safe and effective deformable object manipulation, expanding the repertoire of skills beyond traditional contact-based methods. Third, we describe RoboNinja, a cutting system for multi-material objects. With an interactive state estimator and an adaptive cutting policy, RoboNinja successfully removes the soft part of an object while preserving the rigid core.

Applying Video-based Instruction on Employment Skill Acquisition among Secondary Students with Intellectual and Developmental Disabilities

Sun, Xiaoning The Ohio State University ProQuest Dissertations & 2022 해외박사(DDOD)

This dissertation explores the theme of applying video-based instruction on employment skill acquisition among secondary students with intellectual and developmental disabilities (IDD). It consists of an introduction, three stand-alone papers, and a research statement. Chapter 1 presents the overview of the current competitive integrated environment (CIE) among individuals with IDD and video-based instruction which lays a foundation for the development of Chapters 2 and 3. Chapter 2 presents a literature review on the efficacy of VBI on employment skill acquisition of secondary students with IDD. Chapter 3 extends this literature by describing an experimental analysis of self-directed video prompting on vocational skills. Chapter 4 is a practitioner paper that maps a framework for preparing and implementing self-directed video prompting. Chapter 5 provides a research statement that discusses how my project findings fit in with the current literature and how my future research will contribute to promoting CIE of individuals with IDD.

Essays on the Consequences of Innovative Industries

Niswonger, Benjamin Harvard University ProQuest Dissertations & Theses 2023 해외박사(DDOD)

This dissertation develops novel modeling techniques, theoretical insights and empirical estimations in order to uncover a diverse set of downstream consequences from innovative industries.Chapter 1, which is co-authored with Elhanan Helpman, provides a theoretical framework which allows us to analyze the dynamics of industries constituted by a continuum of small firms and a finite number of large multi-product firms. The presence of a competitive fringe of small firms allows us to provide analytical results which are consistent with observed patterns of increased markups and concentration and a decreased labor share. Furthermore, the model predicts an inverted-U relationship between product span and productivity with implications for industry dynamics.Chapter 2 focuses on the response to changes in skill demand precipitated by the proliferation of skill-biased technologies. The main focus is on the spatial winners and losers from rising skill premiums. This chapter highlights the impact that skill acquisition which is biased in favor of local-skill demand can have on spatial inequality and economy-wide productivity. The main takeaway is that there is large scope for a trade-off between static productivity gains from agglomeration and dynamic gains from local signaling.Chapter 3, co-authored with Zoe Hitzig, considers how the rise of new industries might affect the optimal regulation of bilateral contracts. We begin by building a mechanism design framework which theoretically motivates the use of default and immutable laws in the regulation of contracts. We consider a setting with observable but unverifiable information which is common knowledge to two agents which efficiently bargain over their joint surplus. Our mechanism design framework makes clear how the principal optimally uses defaults in order to achieve equity objectives and sets immutable laws to internalize externalities. We then apply this framework to the current debate on the classification of gig-workers in order to highlight the tradeoffs between these two forms of regulation when the regulator cannot achieve first best.

The Effect of Sequence Learning on Sensorimotor Adaptation

Liu, Yang ProQuest Dissertations & Theses Indiana University 2020 해외박사(DDOD)

Background: Sensorimotor learning and sequence learning are the two processes of motor skill acquisition. These two processes typically occur together in nature and share much in common. However, they are almost always studied in isolation.Objectives: The current study implemented both sequence learning and sensorimotor learning into a reaching task to study the influence of sequence component on sensorimotor learning.Materials and Methods: In both experiments, a distance-dependent force field was used to induce force field in a reaching task. In experiment 1, sequence of target positions was added to the force field adaptation task. There were 12 different reaching directions in one sequence. In experiment 2, sequence of force directions was applied and there were 3 different force directions.Results: Sensorimotor learning and sequence learning occurred in both experiments. In experiment 1, learning the sequence of target positions did not facilitate or impair the adaptation performance. In experiment 2, learning the sequence of force directions increased retention and made the re-learning of the force field faster.Conclusions: When the process of learning a sequence and a sensorimotor force field occur simultaneously, we found the sequence component did not occupy the sensorimotor learning resources, and both processes were able to take place. Certain forms of the sequence learning could facilitate the sensorimotor adaptation rate and increase retention of learning.

대한민국명장의 숙련형성에 관한 내러티브 연구

김용이 한국기술교육대학교 2025 국내박사

본 연구는 대한민국 산업현장에서 최고 수준의 기술력을 보유한 ‘대한민국명장’들의 숙련형성 과정에서의 경험과 그 의미를 내러티브 접근을 통해 탐색하였다. 연구목적은 명장들이 어떤 경로와 방식으로 최고 수준의 숙련을 형성해 왔는지 그 경험과 과정을 실증적으로 분석하고, 이를 통해 숙련형성의 본질적 특성을 이해하는 데 있다. 이를 위한 연구문제는 다음과 같다. 첫째, 대한민국명장의 숙련형성 과정에서의 경험은 어떠한가? 둘째, 대한민국명장이 숙련형성 과정에서 경험한 주요 경험과 그에 담긴 의미는 무엇인가? 연구문제를 탐색하고 연구목적을 달성하기 위하여 질적연구방법 중 하나인 내러티브 연구(Narrative Study)를 활용하여, 명장 10명의 생애사적 경험을 수집하고 그 의미를 해석하였다. 연구참여자는 자동차정비, 보일러, 통신, 철도차량, 전자기기 등 8개 기계·기술분야에서 15년 이상 산업현장에 종사하며 명장으로 선정된 기술자들로 구성하였다. 연구자는 2018년부터 2023년까지 장기간에 걸쳐 이들과의 심층면담을 반복적으로 진행하였고, 수집된 면담 자료와 연구자의 현장노트, 참여자 제공 자료 등을 바탕으로 내러티브를 구성하였다. 연구결과, 대한민국명장의 숙련형성 과정에서의 주요 경험은 다음과 같다. 첫째, 현장 중심의 직접 체험을 통해 교과서로는 습득할 수 없는 암묵적 지식을 내재화하였다. 둘째, 반복적 실천과 체계적 복습을 통해 지식과 기술을 체화하였다. 셋째, 공식 교육보다는 선배, 스승, 비공식 학습 커뮤니티를 통해 핵심 기술을 습득하였다. 넷째, 지속적 성찰과 문제해결 중심의 접근을 통해 전문성을 심화하였다. 다섯째, 실무경험(암묵지)과 이론적 지식(명시지)의 통합을 통해 독창적 전문성을 발전시켰다. 여섯째, 오랜 경험과 실천을 바탕으로 혁신적 해결책을 도출하였다. 일곱째, 암묵지의 후세 전수를 중요한 책임으로 인식하였다. 또한 대한민국명장의 숙련형성 과정은 입문, 적응, 성숙, 전환의 네 단계를 거치며 발전하는 것으로 나타났다. 입문 단계에서 명장들은 계획보다는 우연과 기회, 환경적 요인에 의해 전문분야에 입문하였으며, 현장에서의 직접적인 경험을 통해 기초역량을 습득하였다. 특히 형식적 교육보다는 현장에서의 관찰과 실습, 시행착오를 통해 암묵지를 형성하였다. 적응 단계에서는 현장에서의 역할 확장과 다양한 문제해결 경험을 통해 실질적 역량을 키웠으며, 문제를 분석하고 개선점을 발견하는 안목을 발달시켰다. 또한 자격 취득과 지속적인 학습을 통해 전문성을 향상시켰다. 성숙 단계에서는 자신의 전문성을 집대성하고 체계화하는 과정을 거쳤으며, 현장 맞춤형 시스템과 솔루션을 개발하였다. 또한 사회적 인정과 위상을 확립하고, 통합적 문제해결과 혁신을 통해 전문성을 더욱 발전시켰다. 전환 단계에서는 국가직무능력표준(NCS) 개발 참여, 마이스터고 교재 개발 등 국가 차원의 제도적 기여를 하였으며, 명장기술교육원 설립 등을 통해 후학 양성과 기술 전수에 힘썼다. 또한 학문과 사회적 영역으로 활동을 확장하고, 지속적 혁신과 미래 기술을 탐구하였다. 본 연구는 대한민국명장들의 숙련형성 과정을 입문, 적응, 성숙, 전환이라는 발달적 관점에서 분석함으로써, 각 단계별 특징과 전환점을 파악하고, 숙련형성에 영향을 미치는 핵심 요소들을 식별했다는 점에서 의의가 있다. 연구결과는 현장 중심의 직업교육 강화, 암묵지 전수를 위한 체계적 지원, 기술 분야의 사회적 인식 제고, 평생학습과 재교육 시스템 강화 등에 중요한 시사점을 제공한다. 특히 4차 산업혁명 시대에 전통적 숙련기술과 디지털 기술의 융합, 숙련기술의 재정의와 발전 방향 모색 등이 중요함을 강조하였다. 향후 연구에서는 다양한 산업 분야의 비교 연구, 디지털 전환 시대의 숙련기술 재정의, 그리고 암묵지 전수의 효과적 방법론 개발 등이 필요할 것이다. This study explored the experiences and meanings in the skill formation process of Korean Myungjang, who possess the highest level of technical expertise in Korea's industrial fields, through a Narrative Study approach. The research purpose was to empirically analyze the pathways and methods through which Korean Myungjang have developed their expertise, and to understand the essential characteristics of skill formation. The research questions were as follows: First, what are the experiences in the skill formation process of Korean Myungjang? Second, what are the main activities experienced by Korean Myungjang during their skill formation process and what meanings do they contain? To explore these research questions and achieve the research purpose, Narrative Study, a qualitative research method, was employed to collect and interpret the life-historical experiences of ten Korean Myungjang. The research participants consisted of technicians who had worked in industrial fields for more than 15 years across eight mechanical and technical areas including automotive maintenance, boilers, telecommunications, railway vehicles, and electronic equipment, and were designated as Korean Myungjang. The researcher conducted repeated in-depth interviews with them over an extended period from 2018 to 2023, and constructed narratives based on the collected interview data, field notes, and materials provided by participants. The research findings revealed the following key experiences in the skill formation process of Korean Myungjang: First, they internalized tacit knowledge that cannot be acquired through textbooks through direct field-centered experiences. Second, they embodied knowledge and skills through repetitive practice and systematic review. Third, they acquired core technologies through seniors, mentors, and informal learning communities rather than formal education. Fourth, they deepened their expertise through continuous reflection and problem-solving-centered approaches. Fifth, they developed original expertise through the integration of practical experience (tacit knowledge) and theoretical knowledge (explicit knowledge). Sixth, they derived innovative solutions based on extensive experience and practice. Seventh, they recognized the transmission of tacit knowledge to future generations as an important responsibility. The skill formation process of Korean Myungjang was found to develop through four developmental stages: entry, adaptation, maturity, and transition. In the entry stage, Korean Myungjang entered their professional fields through chance, opportunity, and environmental factors rather than planning, acquiring basic competencies through direct field experience. They formed tacit knowledge particularly through field observation, practice, and trial and error rather than formal education. In the adaptation stage, they developed practical capabilities through role expansion and diverse problem-solving experiences in the field, developing insight to analyze problems and discover improvements. They also enhanced their expertise through qualification acquisition and continuous learning. In the maturity stage, they went through processes of consolidating and systematizing their expertise, developing field-customized systems and solutions. They also established social recognition and status, further developing their expertise through integrated problem-solving and innovation. In the transition stage, they made institutional contributions at the national level through participation in National Competency Standards (NCS) development and Meister school textbook development, and devoted efforts to nurturing successors and transmitting technology through establishing Korean Myungjang training institutes. They also expanded their activities to academic and social domains, exploring continuous innovation and future technologies. This study is significant in that it analyzed the skill formation process of Korean Myungjang from a developmental perspective of entry, adaptation, maturity, and transition, identifying characteristics and turning points of each stage and core factors influencing skill formation. The research findings provide important implications for strengthening field-centered vocational education, systematic support for tacit knowledge transmission, improving social recognition of technical fields, and strengthening lifelong learning and retraining systems. The study particularly emphasized the importance of convergence between traditional skilled techniques and digital technology, and exploring redefinition and development directions of skilled techniques in the Fourth Industrial Revolution era. Future research should include comparative studies across various industrial sectors, redefinition of skilled techniques in the digital transformation era, and development of effective methodologies for tacit knowledge transmission

A Comparison of Student Nurse Anesthetists’ Perceptions of Regional Anesthesia Skill and Knowledge Acquisition Utilizing Cadavers Versus Non-Cadaver Models

Burzumato, Hannah Hall ProQuest Dissertations & Theses Georgetown Univers 2020 해외박사(DDOD)

Due to the expansion of healthcare to a larger population and the increased need for more anesthesia providers, Certified Registered Nurse Anesthetists (CRNAs) are practicing to the full extent of their education and providing a wider range of anesthetic techniques to the general population. One anesthetic technique widely utilized by CRNAs is regional anesthesia. Regional anesthesia is a recognized method of managing pain in the perioperative arena. It has therefore become an integral part of the curriculum in Student Registered Nurse Anesthetists’ (SRNAs’) educational programs. CRNA faculty utilize various models and methods to teach these regional anesthesia techniques and pertinent educational content. Some use cadavers to assist with skill and knowledge mastery while others employ non-cadaveric models. The SRNAs’ perception of regional anesthesia skill and knowledge acquisition has not been determined when either method is used. The purpose of this study is to compare student nurse anesthetists’ self-perception of skill and knowledge acquisition in programs utilizing cadaveric models versus those utilizing non-cadaveric models. To answer the research questions and achieve the study’s purpose, the study utilized a concurrent Convergent Triangulation Design, a typed of Mixed Methods research design that contains a quantitative component followed by a qualitative component within a survey. This survey was distributed nationally via the American Association of Nurse Anesthetists (AANA) Research Services and Assistance Department. Descriptive statistics and chi-square test for independence were used to examine the data for significant relationships between factors, and categorical themes were identified through content analysis of the open-ended qualitative questions. From the quantitative data gathered, there was no statistically significant difference between the two broad groups of cadaveric versus non-cadaveric methods of instruction (p = 0.508 and p = 0.847), however the qualitative analysis yielded narrative components that elaborated on the vague quantitative findings. Through the qualitative analysis, it can be concluded students felt more confident and prepared in their knowledge and skill acquisition when the teachings were integrative and student-centered. This data also provides direction to nurse anesthesia instructors, emphasizing the best methods for regional anesthesia may not be the traditional methods of lecture typically used.

Constructive Learning from Demonstration (CLfD): Application of Constructionism to Computer Vision and Direct Teaching for Robotic Culinary Skill Acquisition

강경진 포항공과대학교 융합대학원 2025 국내석사

We apply Constructive Learning from Demonstration (CLfD), an approach grounded in constructionism—a learning theory traditionally employed in human education— to robotic culinary skill acquisition by comparatively analyzing two distinct teaching methodologies: Direct Teaching and Computer Vision-based imitation. The first method involves Direct Teaching, capturing precise coordinate data from the robot arm's end -effector, ensuring accurate physical reproduction of culinary tasks. In contrast, the second method leverages computer vision techniques, utilizing depth cameras to capture detailed human hand movements while physically guiding the robot arm during cooking instruction, subsequently converting these observations into robotic-compatible coordinate for imitation learning. Both methodologies incorporate Convolutional Neural Network (CNN) networks to effectively extract and generate optimized feature representations, facilitating robotic imitation-based learning processes. A comparative evaluation was conducted on a culinary task involving tofu cutting, measuring physical outcomes such as the number and thickness of slices, alongside operational metrics including total robotic arm movement distance and task completion time. Results indicated that Direct Teaching, especially with ample training data, led to more accurate motion execution and reflected a greater impact of constructionist learning. By comparison, the computer vision-based method was less influenced by constructionist principles. These findings suggest that applying constructivist learning principles significantly enhances robotic skill acquisition, presenting substantial implications for the advancement of food robotics and computer vision disciplines.

Comparing Reading Comprehension Interventions for Middle School Students With Autism: An Alternating Treatment Study

Hunt, Steven M The Ohio State University ProQuest Dissertations & 2024 해외박사(DDOD)

Reading comprehension is a critical academic skill that permeates all academic subjects. Students with autism spectrum disorder are more likely to struggle with reading comprehension, especially as they enter secondary education. Therefore, evidence-based interventions to support middle school students with autism in developing reading comprehension skills are needed. The current alternating treatment study directly compared two research-supported interventions in a group setting: a graphic organizer (story map) and a cognitive strategy (think before, while, and after reading; TWA). Participants demonstrated greater accuracy in responding to comprehension questions while using the interventions in comparison to baseline and no-intervention. Two students demonstrated greater performance while using the story map in comparison to the TWA strategy. No significant difference between the two interventions was identified for three participants. Interventions were equally beneficial for improving factual comprehension performance. The story map was associated with greater performance in answering inferential questions two participants, while the TWA strategy was associated with greater inferential comprehension for one participant. The current study suggests both interventions are promising for this population. However, more research is needed to identify ways in which skill acquisition can be supported while using either intervention strategy.

인지양식, 학습맥락, 시간압력이 시각변별기술의 습득 및 전이에 미치는 효과

박정민 연세대학교 대학원 2003 국내석사

본 연구는 개인의 고유한 인지 특성인 인지양식, 과제의 난이도에 따른 학습맥락, 상황적 영향력인 시간압력이 시각변별과제의 기술습득 및 전이에 어떠한 영향을 주는지 알아보고자 하였다. 총 3개의 실험에서 전체적 ? 분석적 인지양식과 과제의 난이도의 상호작용의 효과를 살펴보았고, 특히 실험 2와 3에서는 인지특성이 아닌 상황적 영향력을 검증하기 위해 각각 1.5초와 0.5초 동안에 시각변별을 하게 하는 시간압력 조건을 추가하였다. 시간압력이 미약했던 실험 1에서는, 분석적 처리자가 훈련 세션의 첫 블록에서 자극변별의 높은 정확도와 느린 반응시간을 보였다. 하지만, 1.5와 0.5초 내에 자극을 변별하도록 한 실험 2와 3에서는, 분석적 처리자가 훈련세션의 초기 블록 동안 낮은 시각변별의 정확도와 자극의 복잡성에 상관없는 빠른 반응시간을 보였다. 이러한 결과는 정보처리의 시간적 제약 때문에 분석적 처리자가 자극의 세부 특징들을 일일이 비교하는 자신의 특성을 이용하지 못하였기 때문이라고 할 수 있다. 하지만, 블록이 진행될수록, 분석적 처리자의 반응시간의 기울기가 과제특성에 따라 변화하는 것이 보여, 분석적 처리자는 시간압력 상황에서의 학습이 전체적 처리 양식을 유도함에도 불구하고, 일대일 비교를 하는 분석적 양식의 특성을 고수하려 한다는 것을 알 수 있었다. 하지만 실험 3에서는 어려운 학습맥락 조건의 분석적 처리자의 경우, 전체적 처리자와 유사한 전략의 개발을 보였다. 즉, 시간제약이 극심해서 상황적 영향력이 증가한 경우에는 자신의 인지양식을 사용하지 못하고, 상황에 순응적으로 전략을 변화시킨 것이라고 할 수 있다. 이러한 결과들을 통해 시각변별의 기술습득은 인지양식에 매개하여 인지전략이 개발됨으로써 이루어지지만, 자신의 인지양식을 사용하기 어려운 극심한 시간압력과 같은 상황적 영향력 하에서는 상황에 순응적으로 전략을 변화시킨다는 결론을 내릴 수 있다. This study investigated how cognitive styles, training contexts and time pressure influenced visual discrimination skill acquisition and transfer. In this study, three experiments were conducted. Each of experiments identified the effects of the interaction of the holistic and analytic style with varying levels of training difficulty on transfer performance. Especially, in experiment 2 and experiment 3, time constraints of each 1.5 seconds and 0.5 seconds provided to verify the situational influences. In experiment 1 that time pressure was feeble, analytic participants were not affected by the hard and easy training strategies during transfer performance. However, in experiment 2 and 3 that do to discriminate stimuli in 1.5 and 0.5 seconds, analytic participants showed low accuracy and low reaction time that was independent in complexity of stimuli during early block of training session. This results show analytic participants are difficult to use own characteristic that compares stimuli one by one because of time pressure. However, in experiment 2, as block was progressed, analytic participants' slopes of reaction time increased, suggesting that their own analytic style emerged, even though time pressure induced them to the holistic strategies. In experiment 3, analytic group showed similar strategic development to holistic group. According to the results of this study, Individuals differently develop strategic processing skills depending on cognitive styles, and cognitive strategies change adaptably in influential circumstance such as extreme time pressure that hard to develop own cognitive strategies.

인지 기술 학습에서 상위인지의 역할 : 학습 난이도에 따른 Judgment of Learning (JOL) 변화가 전이 수행에 미치는 영향

윤종성 연세대학교 대학원 2007 국내석사