Minimum Useful Fidelity of Streamed Large-scale 3D Models over 5G

Large scale 3D models and point clouds play an important role in many fields ranging from simulation to Product Lifecycle Management to digital twins deployed on a plethora of AR-capable devices. However, due to systems or resource limitations, these models are not always as close to the reality as they purport to be.

Using AR connected to 5G networks, it is possible to stream these models to a device to assist processes performed by human operators or technicians. Generally speaking, the simpler the model (i.e. the lower the fidelity), the less resource is required for streaming. However, if simplification goes too far, utility suffers, possibly preventing the operator from performing the intended task.

The research scope should include comparative studies measuring the exact impact of automatic model and point cloud simplification due to streaming, possibly expressed by the decrease in polygon count vs the time required and error rates achieved in a particular operations. Measurement methods would be developed to ensure accuracy within 5% margin of error.

Stakeholders

Operations leaders, financial management, OEM manufacturers, Independent Software Vendors, Cloud Service Providers

Possible Methodologies

This research project will make a statistically significant number of comparisons between versions of different fidelity of the same large-scale 3D model and point clouds and will examine their impact on the performance aided by said models. Examining the relationship between model fidelity and human performance will inform the investment process for large-scale 3D models. New methods could be developed to compare usefulness at different levels of fidelity in various industries to clarify why some industries require different levels of fidelity to meet their objectives.

Research Program

This research topic can be combined with other projects examining the tradeoffs between fidelity and function of large-scale 3D models.

Miscellaneous Notes

Several AREA members are already working in this field and the 7th AREA research project performed a comparison of commercial solutions available for model decimation although did not study the streaming aspect of this research topic. The MxD is interested in collaborating to use its mmWave 5G infrastructure in a factory setting to study this topic further.

Keywords

Point cloud, large-scale 3D models, polygons, low-poly, high-poly, realistic, surfaces, finite element method, computer simulation, digital simulation, computational mechanics, computer aided engineering, discrete event simulation, virtualization

Research Agenda Categories

Technology, Business

Expected Impact Timeframe

Medium

Related Publications

Using the words in this topic description and Natural Language Processing analysis of publications in the AREA FindAR database, the references below have the highest number of matches with this topic:

• Fradi, A., Louhichi, B., Mahjoub, M. A., & Eynard, B. (2017). 3D Object Retrieval Based on Similarity Calculation in 3D Computer Aided Design Systems. 2017 IEEE/ACS 14th International Conference on Computer Systems and Applications (AICCSA).
• Xue, Y., Xu, S., Wang, L., Dai, C., & Wu, Y. (2019). Dual-Model Approach for Engineering Collision Detection in the CAVE Environment. 2019 IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct).
• Hayashida, H., Funashima, H., Katayama-Yoshida, H., & Nomakuchi, T. (2017). Opening the Door for the New Methodology for Optimizing Functional Material Development in Technology Management Framework II. 2017 Portland International Conference on Management of Engineering and Technology (PICMET).
• Dorit Borrmann; Andreas Nuechter; Thomas Wiemann .. Large-Scale 3D Point Cloud Processing for Mixed and Augmented Reality. (2018). 2018 IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct).
• Wang, W., Ceylan, D., Mech, R., & Neumann, U. (2019). 3DN: 3D Deformation Network. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

More publications can be explored using the AREA FindAR research tool.

Author

Peter Orban

Last Published (yyyy-mm-dd)

2021-08-31

Go to Enterprise AR Research Topic Interactive Dashboard