Long Cheng
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Long Cheng has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Andreas Kunz
Innovation Center Virtual Reality, ETH Zurich, Switzerland
3 shared publications - Damla Welti
Innovation Center Virtual Reality, ETH Zurich, Switzerland
1 shared publication - Gilberto Mazzola
Innovation Center Virtual Reality, ETH Zurich, Switzerland
1 shared publication Lipeng Qin
School of Automation and Electrical Engineering, Linyi University,Linyi,Shandong Province,CN
1 shared publication- Mathieu Lutfallah
Innovation Center Virtual Reality, ETH Zurich, Switzerland
1 shared publication - Michael Schreiner
Digital Innovation Lab, RhySearch, Switzerland
1 shared publication
Latest Publications
- 7 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 32 (5) , 3497-3507
Spatial Perspective Taking in Interior Design: A Comparison of World-in-Miniature and Cross Reality
Visualization of complex architectural and interior design proposals demands spatial perspective taking (SPT), an ability that can be difficult to acquire. Mixed Reality (MR) is particularly effective in architectural visualization because it overlays virtual models onto already existing real environments, allowing for precise spatial accuracy and aesthetic assessment. However, MR confines users to viewpoints tied to locations they can physically reach in the real-world. To overcome this constraint, we designed an applied, visibility-driven SPT task and explored two methods to provide additional perspective cues in MR: the World-in-Miniature (WIM) mode and the Cross Reality (CR) mode. The WIM mode provides a scaled-down 3D replica of the virtual environment, allowing users to explore a design at a reduced scale. Conversely, the CR mode allows switching from MR to VR to hide the real-world completely, enabling users to teleport to physically unreachable viewpoints and assess a design at an accurate scale. We conducted a pilot user study with 14 participants and a formal user study with 80 participants on an object-placement task. Our results revealed that the CR mode achieved success scores as high as those of the WIM mode. The WIM mode was faster and perceived as more user-friendly. However, participants preferred the CR mode for its accurate spatial representation and effective presentation of design ideas. The CR mode recreates the physical room as a scale-accurate virtual counterpart, preserving geometric cues critical for spatial visualization while removing the need for manual room modeling. Our findings highlight the potential of these modes to transform 3D design visualization and set new standards in the industry.
- 7 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 32 (5) , 3497-3507
Spatial Perspective Taking in Interior Design: A Comparison of World-in-Miniature and Cross Reality
Visualization of complex architectural and interior design proposals demands spatial perspective taking (SPT), an ability that can be difficult to acquire. Mixed Reality (MR) is particularly effective in architectural visualization because it overlays virtual models onto already existing real environments, allowing for precise spatial accuracy and aesthetic assessment. However, MR confines users to viewpoints tied to locations they can physically reach in the real-world. To overcome this constraint, we designed an applied, visibility-driven SPT task and explored two methods to provide additional perspective cues in MR: the World-in-Miniature (WIM) mode and the Cross Reality (CR) mode. The WIM mode provides a scaled-down 3D replica of the virtual environment, allowing users to explore a design at a reduced scale. Conversely, the CR mode allows switching from MR to VR to hide the real-world completely, enabling users to teleport to physically unreachable viewpoints and assess a design at an accurate scale. We conducted a pilot user study with 14 participants and a formal user study with 80 participants on an object-placement task. Our results revealed that the CR mode achieved success scores as high as those of the WIM mode. The WIM mode was faster and perceived as more user-friendly. However, participants preferred the CR mode for its accurate spatial representation and effective presentation of design ideas. The CR mode recreates the physical room as a scale-accurate virtual counterpart, preserving geometric cues critical for spatial visualization while removing the need for manual room modeling. Our findings highlight the potential of these modes to transform 3D design visualization and set new standards in the industry.
- 7 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 32 (5) , 3314-3324
U-NEXT: A User-Centered New Prototype for Walking Excavator Simulator
The walking excavator is a highly complex construction machine, characterized by challenges in control, instruction, and operational safety. In this work, we implemented a virtual reality (VR) walking excavator simulator that integrates real hardware components with kinematic physics simulation. Using the simulator, we conducted two user studies. The first involved 45 participants and examined the effects of user perspectives on performance and preference. The second involved 12 participants and investigated the influence of video passthrough of the controllers. Results show that the choice of perspective should be task-dependent. Third-person view outperformed first-person view in the undercarriage task in which complex leg movements are needed. It supported better planning and allowed complete body visualization. However, it did not reduce motion sickness and was less preferred than first-person view for training. From the second user study, passthrough significantly improved controller learning compared to animated controllers, with participants making fewer errors when using the passthrough design. Based on these findings, we provide design recommendations for user perspectives and controller representation in VR-based training simulators.
- 7 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 32 (5) , 3314-3324
U-NEXT: A User-Centered New Prototype for Walking Excavator Simulator
The walking excavator is a highly complex construction machine, characterized by challenges in control, instruction, and operational safety. In this work, we implemented a virtual reality (VR) walking excavator simulator that integrates real hardware components with kinematic physics simulation. Using the simulator, we conducted two user studies. The first involved 45 participants and examined the effects of user perspectives on performance and preference. The second involved 12 participants and investigated the influence of video passthrough of the controllers. Results show that the choice of perspective should be task-dependent. Third-person view outperformed first-person view in the undercarriage task in which complex leg movements are needed. It supported better planning and allowed complete body visualization. However, it did not reduce motion sickness and was less preferred than first-person view for training. From the second user study, passthrough significantly improved controller learning compared to animated controllers, with participants making fewer errors when using the passthrough design. Based on these findings, we provide design recommendations for user perspectives and controller representation in VR-based training simulators.
- Open Access
- 9 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Comparing Tracking Accuracy in Standalone MR-HMDs: Apple Vision Pro, Hololens 2, Meta Quest 3, and Pico 4 Pro
Modern Mixed Reality Head-Mounted Displays (MR-HMDs) can track user movements across large spaces without external markers. This study evaluates the tracking accuracy and the loop closure capabilities of four commercially available MR-HMDs across four distinct scenarios. We found consistent tracking performance in well-lit and expansive environments for all devices. Tracking accuracy remained stable even in outdoor nighttime conditions. Furthermore, most HMDs demonstrated effective error correction during loop closure, with errors in non-loop scenarios consistently exceeding those in loop scenarios.
- 9 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Open AccessComparing Tracking Accuracy in Standalone MR-HMDs: Apple Vision Pro, Hololens 2, Meta Quest 3, and Pico 4 Pro
Modern Mixed Reality Head-Mounted Displays (MR-HMDs) can track user movements across large spaces without external markers. This study evaluates the tracking accuracy and the loop closure capabilities of four commercially available MR-HMDs across four distinct scenarios. We found consistent tracking performance in well-lit and expansive environments for all devices. Tracking accuracy remained stable even in outdoor nighttime conditions. Furthermore, most HMDs demonstrated effective error correction during loop closure, with errors in non-loop scenarios consistently exceeding those in loop scenarios.
- 11 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 285-290
Inertial Sensor Binding Position and Combination in Badminton Action Recognition Using Machine Learning
During the coaching of both novice and seasoned badminton players, a comprehensive analysis of their physical training regimens has gained paramount importance. To achieve the objective of recognizing badminton actions based on inertial sensors, we developed a data acquisition platform that focuses specifically on badminton actions. This paper primarily employs three recognition algorithms, namely K-nearest neighbors (KNN), decision tree C4.5, and support vector machine (SVM), to classify various badminton actions. We have conducted multiple experiments to scrutinize and comprehend the influence of recognition outcomes under different sensor quantities and diverse sensor placement positions.
- 11 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 285-290
Inertial Sensor Binding Position and Combination in Badminton Action Recognition Using Machine Learning
During the coaching of both novice and seasoned badminton players, a comprehensive analysis of their physical training regimens has gained paramount importance. To achieve the objective of recognizing badminton actions based on inertial sensors, we developed a data acquisition platform that focuses specifically on badminton actions. This paper primarily employs three recognition algorithms, namely K-nearest neighbors (KNN), decision tree C4.5, and support vector machine (SVM), to classify various badminton actions. We have conducted multiple experiments to scrutinize and comprehend the influence of recognition outcomes under different sensor quantities and diverse sensor placement positions.