서지주요정보
Space-adaptive mutual space generation for mixed reality remote collaboration = 혼합 현실 원격 협업을 위한 공간 적응적 공유 공간 생성
서명 / 저자 Space-adaptive mutual space generation for mixed reality remote collaboration = 혼합 현실 원격 협업을 위한 공간 적응적 공유 공간 생성 / Dooyoung Kim.
발행사항 [대전 : 한국과학기술원, 2024].
Online Access 원문보기 원문인쇄

소장정보

등록번호

8042484

소장위치/청구기호

학술문화관(도서관)2층 학위논문

DGCT 24008

휴대폰 전송

도서상태

이용가능(대출불가)

사유안내

반납예정일

리뷰정보

초록정보

To achieve a sense of coexistence in Mixed Reality (MR) remote collaboration where Augmented Reality (AR) hosts and Virtual Reality (VR) clients are leveraged, it is necessary to overcome spatial differences between spaces and create a mutual space with a unified coordinate system. This work proposed a space-adjusting technique with redirected walking and multiple spatial matching algorithms to create a spatial affordance-aware mutual space from heterogeneous remote spaces where users could physically interact in connected virtual space. In the beginning, Study 1 determined the threshold range of relative translation gain (RTG) in redirected walking with two user studies (n = 64) and revealed that three components of spatial configuration, space size, object existence, and spatial layout, all significantly affected the users' visual perceptual RTG thresholds and estimated the allowable space adjustment constraints according to the spatial configuration. In the following, Study 2 proposed an edge-centric physical space rescaling using RTG as a space rescaling term and spatial registration algorithm to utilize basic geometric information from users' physical space in the virtual space. The evaluation results showed that RTG-based space registration could not only align planes and edges most but also secure the largest interactable area compared to other spatial registration methods. Study 3 proposed a method for extracting and matching interactable subspaces from multiple disparate remote spaces and generating mutual spaces using an optimization algorithm that determines the optimal initial positions of users. Experiments on 900 space combinations generated by varying the number of client spaces to two, four, and six proved that the spatial affordance-aware subspace allocation method could generate a mutual space suitable for collaborative contexts even if the number of client spaces increases. Finally, the proposed mutual space generation algorithm was implemented in the MR remote collaboration system, and mutual space generation guidelines for developers were derived. The results of the three phases of research showed that physical space transformation using translation gain-based redirected walking is effective in aligning heterogeneous spaces, and the spatial affordance-aware interactable subspace allocation approach is beneficial to generating a mutual space from multiple dissimilar spaces. The findings from the three-stage study will enable remote collaboration between distant users as if they are all in the same room, connecting people across spatial boundaries and contributing to the widespread adoption of remote work, immersive VR content creation, and addressing urban sprawl and rural decline.

증강 현실 호스트와 가상 현실 클라이언트들 간의 공존감 있는 혼합 현실 원격 협업을 위해서는 공간들 간의 구조적 차이를 극복하고 사용자들이 상호작용할 수 있는 하나의 공유 공간을 생성하는 것이 필요하다. 이를 위해 본 연구는 재조정된 걸음 기술을 사용하여 가상 현실 클라이언트들의 물리 공간들을 변형하고, 다수의 이질 공간을 정합하여 공간적 어포던스를 고려하여 사용자들이 물리적으로 상호작용 가능한 공유 공간을 생성하는 방법을 제안한다. 첫 번째 연구는 사용자 실험 (n = 64) 을 통해 공간의 세 가지 구성 요소인 공간의 크기, 객체의 유무, 공간의 구조가 모두 사용자의 상대적 길이 조정의 인지 임계값에 유의미한 영향을 미친다는 것을 밝혔으며, 공간의 구성에 따른 공간 변형 허용 범위를 밝혔다. 이어지는 두 번째 연구에서 상대적 길이 조정을 활용하여 가상 현실 사용자의 물리 공간의 기본 기하 정보인 모서리와 평면 정보를 가상 공간에서 활용할 수 있도록 모서리 중심 공간 변형 및 정합 알고리즘을 제안하고, 제안한 방법이 다른 공간 등록 방법들과 비교하여 현실-가상 공간 간의 모서리와 평면 정보를 가장 많이 정렬할 수 있을 뿐만 아니라 상호작용 가능한 면적도 가장 크게 확보할 수 있음을 보였다. 마지막 연구에서는 서로 다른 다수의 원격 공간들로부터 상호 작용 가능한 부분 공간을 추출 및 정합하고, 사용자들의 최적 초기 위치를 결정하는 최적화 알고리즘을 활용한 공유 공간을 생성하는 방법을 제안한다. 클라이언트 공간의 수를 2개, 4개, 6개로 달리해가며 생성한 900개의 공간 조합에 대한 실험을 통해 공간적 어포던스 기반 부분 공간 정합 방법이 클라이언트 공간의 수가 늘어나도 강건하게 협업 맥락에 적합한 공유 공간을 생성할 수 있음을 보였다. 최종적으로 제안한 공유 공간 생성 알고리즘을 혼합현실 원격 협업 시스템에 적용하였다. 총 세 단계의 연구 결과로부터 재조정된 걸음 기술을 활용한 물리 공간 변형이 이질 공간 간의 정합에 효과적이며, 공유 공간 생성시 공간적 어포던스를 고려하여 협업의 중심이 되는 부분 공간을 정합하는 것이 중요함을 보였다. 본 연구 결과를 활용하면 멀리 떨어진 사용자들이 마치 한 공간에 모여있는 것처럼 공존감 있는 원격 협업이 가능할 것이며, 공간의 한계를 넘어 사람들을 연결함으로써 원격 근무의 보급, 몰입형 가상 현실 컨텐츠 제작, 도시 집중화 및 지방 소멸 문제 해결에 기여할 수 있을 것이다.

서지기타정보

서지기타정보
청구기호 {DGCT 24008
형태사항 vii, 116 p. : 삽도 ; 30 cm
언어 영어
일반주기 저자명의 한글표기 : 김두영
지도교수의 영문표기 : Woontack Woo
지도교수의 한글표기 : 우운택
Including appendix
학위논문 학위논문(박사) - 한국과학기술원 : 문화기술대학원,
서지주기 References : p. 92-106
주제 Mixed reality
Virtual reality
Augmented reality
Remote collaboration
Spatial matching
Mutual space
Optimization
Redirected walking
Relative translation gains
Visual perceptual threshold
혼합 현실
가상 현실
증강 현실
원격 협업
공간 정합
공유 공간
최적화
재조정된 걸음
상대적 길이 조정
인지 임계값
QR CODE

책소개

전체보기

목차

전체보기

이 주제의 인기대출도서

A teaserimage of research motivation. This research aims to generate a mutual space from heteroge- neous spaces for Mixed Reality (MR) remote collaboration where AR/VR remote users connect to the AR host users reference space.

Research motivation which shows the limitation ofsimple floor-centric dissimilar space registration A) Due to the current VR HMD S user-defined empty floor-based playground generation, the VR user falls down when he tries to reach a VR ping-pong table as ifitisreal. B) VR user access from remote space passes through the AR user's physical table and breaks immersion because spatial registration doe

The overview of three studies. Study 1 investigated the effects ofspatial configuration on the relative translation gain (RTG) threshold. Study 2 proposed an edge-centric space rescaling technique with RTG and an optimization algorithm for registering dissimilar virtual-physical space. Study 3 presented a mutual space generation algorithm from multiple spaces and implemented an MR telepresence sys

The overview ofStudy 1. (A) From the target virtual environment, the effects ofspatial configuration (space size, object existence, and spatial layout) on the RTG threshold range with two user studies. (B) According to the distinguished spatial configuration, the proper RTG threshold range, which is obtained from user studies, could be set. (C) The selected RTG threshold range can be used for the

The overview of Study 2. From dissimilar 3D input spaces' top view, multiple polygons with corresponding semantic labels were extracted. Centered on the main object, a grid-based space rescaling technique was applied and the optimal position and RTGs for space registration could be obtained from the optimization algorithm.

The overview of Study 3. (A) From polygonal space inputs from heterogeneous spaces, (B)optimal physical-virtual space registration results can be obtained. (C) To generate a mutual space from heterogeneous spaces, the spatial affordance-aware subspace allocation algorithm and user instantiation algorithm were used. (D After augmenting proper obstacles for the semantically unmatched areas, mutual s

The overview of Study 1.

The concept image ofthe relative translation gains (RTGs). Black dotted lines indicate real-world translation (Treal) that the user physically walked in the real world and green lines indicate virtual-world translation (T.rtual) that the user walked along visually in the virtual world. The translation gain 8T 1S TviriuallTeal and the RTG threshold range QT (2D translation ratio) can be represented

The AngleofDeclination (AoD)between the participant's eye-gaze and forward-gaze. Forward-gaze refers to the orthogonal projection vector ofthe eye-gaze.

Top-view ofthe pathin (A) Medium X Empty and (B) Medium X Furnished conditionin Study 1-1. Participants walk along thepath repeatedly in order ofpath 1,path2, and path3 (SZ= Start Zone,Z1 =Zone 1,Z2 =Zone2).

Three virtual room size conditions* AoD settings

(A) Three size conditions used in Study 1-1. (B) Top-view ofthe Empty and the Furnished condition in the Small size condition.

Screenshot ofthe six conditions in Study 1-1. The combined paths and furniture are placed at the center of each condition.

(A) Two size conditions (Small, Large) used in Study 1-2. (B) Top-view ofthe four spatial layout conditions (Empty, Peripheral, Scattered, Centered) in the Small condition.

Screenshot ofthe eight experimental conditions in Study 1-2. The inner floor area expressed in the Large condition is the same size as the Small condition's entire floor area and is located at the center ofthe room Combined paths and exhibit objects are placed in it.

Participant information for user Study 1-1

Participant information for user Study 1-2

The participant's view through the VR HMD during the experiment. (A)Large X Empty, (B)Large X Furnished.

The effects ofSize and Object Existence on the probabilities of "faster' responses (0: mean

Theeffects of Size and Spatial Layout on the probabilities of "faster' responses (0: mean)

Fitted psychometric functions of mean estimated threshold values from (A) Study 1-1 and (B) Study 1-2.

Relative translation gain thresholds according to virtual room configurations in Study 1-1.

Relative translation gain thresholds according to virtual room configurations in Study 1-2.

(A) The density distribution of AoD according to VEconfigurations in Study 1-1. (B) The colored area refers to where users could perceive the virtual horizon for (B1)Large, (B2) Medium, and (B3) Small conditions

The density distribution of AoD according to VE configurations in Study 1-2. The area where users could perceive the virtual horizon in Large is drawn with a red dashed area and in Small is drawn with a blue filled area.

Thetop-view of the heatmap from subjects' gaze distribution in Large conditions from Study Sequential paths are drawn as dashed lines, and objects placed in the rooms are represented as colored square b

The top-view of the heatmap from subjects' gaze distribution in Small conditions from Study Sequential paths are drawn as dashed lines, and objects placed in the rooms are represented as colored square bc

The post-SSQ total scores in Study 1-1 and Study 1-2.

Front views ofthree different virtual spaces generated for the same movable space in the real world Compared with (A) the base state, the adjusted movable space to which relative translation gains are applied can be increased when (B) the perceived movable space is larger than the adjusted movable space and (C) objects are placed.

The overview of Study 2.

Top-view ofthe physical space forrescaling. Grids were divided by the extension ofthe orange-colored main object (Grid 5)'s four edges. The dashed area indicates the linear gain change area.

Illustrations about four space registration evaluation metrics' numerator. Green lines and dashed areas refer to matched edges and planes between physical and virtual spaces. Red areas in Vsem refer to semantically unmatched areas. Each metric can be computed with described matched areas or lengths.

Five input 3D indoor space models for evaluation. Each model is generated from real indoor environments

(E) Office has aboutone-third ofthe size compared to the XR Studio, and also, the table size is about half ofthat of XR Studio. In every spatial aspect,spatial matching with XR Studio and Officeis expected to be the highest among all space pairs.

Fiveinput spaces for evaluation. The polygonal input data was generated from the top view of3D indoor space models. The polygon's color refers to the following semantic labeled polygon: Black floor, orange - the main object (table), red - obstacles.

The table shows the optimization results ofthree registration methods for four evaluation conditions: object (table) edgesync ratio (Vobj), wall sync ratio (Vwall), semantic match ratio (Vsem), registered space size and registered table area ratio. Optimal RTGs (G) and optimal position (o) for rescaling and registering physical space are also written.

Space registering results offour input space pairs. The transparent polygons are an intersection oftwo spaces and the black rectangle overlayed on it refers to the registered area between physical and virtual spaces The green lines indicate the aligned table edge and the wall surface, the orange areas indicate the registered table surface, and the red areas indicate the registered obstacles or lab

CR [143], ENI [145], SD, and SMD for given space pairs

The proposed space-rescaling technique aligns edges and planes larger than the original scale of physical space, which enables the utilization ofbasic geometric information from physical space in virtual space The user A in his tracked physical space (left), connected to the dissimilar target virtual space (right) through his avatar and physically interacted with the registered floor, table, and w

(A) Three interactable mutual subspaces were highlighted with corresponding collaboration context: (a) table-centric (yellow), (b) wall-centric (red), and (c) floor-centric (blue). (B) AR/VR clients spatial affordance- aware instantiation positions in their remote spaces according to the collaboration context.

The overview of Study3.

The overview offour sequential procedures for MR mutual space generation. A) Selecta target object from each client's space corresponding to the host's target object with a scene graph, B)collaboration context-aware spatial matching, C) spatial affordance-aware user instantiation, and D) interactable subspace extraction and allocation.

Topview offour host spaces (m - meeting room) and ten client spaces (h home, 0 office) - -

The user instantiation success rate in each condition with a corresponding number ofhost users. A H1-C2,B)HI-C4, and C) H1-C6.

The mean ofinteractable space area and obstacle area in each condition. A) Mean total mutual space area in A-1)H1-C2.A-2) H1-C4,and A-3)H1-C6. B-1) Mean interactable area per client and B-2) mean obstacle area per client.

The representative spatial matching results with SA-ISA (SA-Table, SA-Wall, SA-Floor) and two comparison groups, S-ISA and S-TI, in H1-C2 conditions. H1, H2,and H3 referto host users 1.2, and3. C1 and C2refer to clients from each space. The gray areas in each space refer to semantically unmatched areas.

The representative spatial matching results with SA-ISA (SA-Table, SA-Wall, SA-Floor) and two comparison groups, S-ISA and S-TI in H1-C4conditions. H1, H2,and H3 refer to hostusers 1. 2, and3. Cl to C4 refers to clients from each space. The gray areas in each space refer to semantically unmatched areas, and the case ofthe user on the gray area refers to the sittable area from the client's space.

The representative spatial matching results with SA-ISA (SA-Table, SA-Wall, SA-Floor) and two comparison groups, S-ISA and S-TIin H1-C6conditions. H1, H2, and H3 refer to host users 1 2, and 3. C1to C6 refers to clients from each space. The gray areas in each space refer to semantically unmatched areas, and the case of the user on the gray area refers to the sittable area from the client's space.

System flow chart for creating a mutual space from heterogeneous spaces. From 3D space inputs. top-viewed polygons with semantic labels were extracted, and the RTG threshold range was set according to the spatial configuration of the host's space. Then, the proposed space-matching algorithm will find optimal values for VR client space rescaling and multi-space registration. Finally, augment proper

Screenshot ofMR telepresence system with AR host user with a Hololens2 and two VR clients with Quest Pro. (A) The AR host user in KISTI and (B) the VR client user in the KAIST N5 building.

The overview ofthree studies. Study 1 investigated the effects ofspatial configuration on the relative translation gain (RTG) threshold. Study 2 proposed an edge-centric space rescaling technique with RTG and an optimization algorithm for registering dissimilar virtual-physical space. Study 3 presented a mutual space generation algorithm from heterogeneous spaces.