Applications of 4D Point Clouds (4DPC) in Digital Twin Construction: A SWOT Analysis
Dong Liang 1*, Fan Xue 2
This is the authors’ version of the paper:
Liang, D., & Xue, F. (2022). Applications of 4D point clouds (4DPC) in digital
twin construction: A SWOT analysis. Proceedings of the 27th International
Symposium on Advancement of Construction Management and Real Estate
(CRIOCM2022), pp. 1231–1238, Springer, DOI:10.1007/978-981-99-3626-7_95.
This file is shared for personal and academic use only, under the license CC
BY-NC-ND 4.0 (Non-Commercial, No Derivatives, and with an Attributed
citation when you use). The final published version of this paper can be found at:
https://doi.org/10.1007/978-981-99-3626-7_95.

5

10

15

20

Abstract: Digital twin construction (DTC), the virtual replica of physical construction,
is an essential enabler for the promised Construction 4.0. Thanks to its real-time virtualphysical synchronicity, many uncertainty-tolerant functions, and applications, such as
system integration, testing, monitoring, and maintenance, could be dynamically
simulated and optimized through the DTC. The data foundation for creating or updating
a DTC is real-time reality capture on construction sites. Despite a lot of efforts such as
the Internet of Things (IoT) and AI cameras, so far, it is very challenging to capture the
spatial-temporal information of the whole construction site. Time-dynamic 4D point
cloud (4DPC) is an emerging sensing technology consisting of three-dimensional point
cloud scans at a frequency, e.g., two frames per second. 4DPC has been successfully
used in several industries, such as autonomous driving and motion analysis in sports
science, due to its ability to capture dynamic objects and environments. This paper
presents a strengths, weaknesses, opportunities, and threats (SWOT) analysis of the
possible applications of 4DPC in DTC. In summary, 4DPC can effectively capture the
dynamic feature of construction progress, with strengths in real-time information
updating, covering more precise geometry dynamics, and providing localization and
mapping information, etc., which are in line with the development of DTC. Meanwhile,
the disadvantages include errors or noise in recognizing lign-permeable materials and
water, enormous data volume, challenges in modeling spatial-temporal structure, and
computation load in processing. The 4DPC receives several opportunities, such as
establishing a dynamic data foundation for construction automation and modeling of
workspace and integrating with 4D BIM for creating DTC; However, threats such as
LiDAR-hostile weather and emerging competitors are also identified. Overall, 4DPC is
a promising research direction that may interest DTC researchers and practitioners for

Dong Liang
Department of Real Estate and Construction, The University of Hong Kong, Hong Kong SAR, China
* Corresponding author, E-mail: leodong@connect.hku.hk
1

Fan Xue
Department of Real Estate and Construction, The University of Hong Kong, Hong Kong SAR, China
E-mail: xuef@hku.hk

2

25

various application situations in the construction industry.
Keywords: 4D point cloud; Construction; Digital twin; Digitalization and automation;
SWOT analysis.

1 Introduction

30

35

40

45

50

55

60

65

Digital twin, representing a physical system or process by a real-time virtual system to
enable simulations, analyses, and optimization, has been adopted by several industries,
such as the aircraft and manufacturing industry. The digital twin has been introduced
into construction to ensure more automatic, faster, and more collaborative processes in
recent years[1]. The purpose of the construction digital twin (DTC) is to improve
existing construction processes and enrich current building-related models (i.e.,
building information modeling (BIM))[2]. Within the context of cyber-physical
synchronicity, the digital twin models should have the ability to reflect the construction
of physical assets in real-time. However, most current DTC models, such as BIM
models, are static and cannot be updated in real-time, which cannot meet the demand
of the modern construction projects with the increasing complexity and increasing
number of involved parties. One of the main reasons is that current measures in
construction site sensing for reality capturing are restricted to routine laser scanning,
manual management updates. Some advanced digital modeling and simulation
technologies could not be fully utilized if the dynamic feature of construction progress
could not be effectively captured. Therefore, a real-time DTC modeling approach that
is fast, inexpensive, reliable, and automated is required. The technique should be able
to rapidly create and update accurate, comprehensive, semantically rich models in a
master format that can be used to any engineering situation and is broadly applicable to
all building projects.
Sensors for spatial data have developed over the last several decades from basic
range sensors based on sonar or IR, which provided a few bytes of information about
the environment, to ubiquitous cameras and laser scanners[3]. Recent advances in
LiDAR technology have made it possible for these sensors to produce high-quality 3D
reconstructions of the world: point clouds[4]. Point clouds, a three-dimensional
collection of data points or coordinates, provide more information than twodimensional pictures and are insensitive to light[5]. Compared to other geometric
representations like grids, point clouds do not experience quantization mistakes as
much[6]. In most cases, point clouds always refer to 3D point clouds, the phase that
comes before creating an accurate 3D model of the actual environment. 3D point clouds
have already been widely used in many research and industrial domains, such as
surveying, electricity, construction, and industry, due to their excellent ability to
represent our three-dimensional world. In the construction industry, the 3D point cloud
is currently used for various aims such as as-built building reconstruction and geometry
quality inspection[7-9]. It should be noted that the application of 3D point clouds is
limited in capturing some static objects due to only spatial information contained in the
data.

70

75

80

85

90

The advent of the 4D point cloud, consisting of three dimensions and time
information, has expanded the application of the point cloud into a dynamic world[10].
The most significant improvement of a 4D point cloud compared with a 3D point cloud
is that temporal structure is added to data. The 4D point cloud, a window into the
dynamics of the 3D world we live in, is a rich source of visual data that demonstrates
how objects move against backdrops and what occurs when we take action[11]. There
are some reasons why the 4D point cloud has gained popularity since it appeared. First
of all, the ability to comprehend a changing 3D world is essential for robotic agents and
a variety of other applications. In addition, various identification tasks, such as
calculating a moving object's acceleration or identifying human activities, benefit from
temporal data sequences longer than two frames[12]. The current popular 4D point cloud
research and application fields include but are not limited to autonomous driving, robot
vision, human-robotics interaction, simultaneous localization and mapping (SLAM),
and motion analysis in sports science. For instance, autonomous agents need to use 4D
point clouds to understand the dynamic 3D environment in many applications, such as
robotic manipulation[13].
As an effective sensing technology for dynamic environments, the 4D point cloud
(4DPC) has the potential to promote the development of the digital twin of the
construction industry. The 4DPC could help achieve the 4D modeling, making it
possible to represent, examine, and evaluate all processes associated with construction,
such as graphical models, management, resources, and safety concerns, from a temporal
viewpoint. The dynamic digital twin model may provide a new way to view some
applications such as lean construction, site monitoring, safety or environmental aspects.
Given its great potential for application in construction, as briefly illustrated in
Figure 1, this paper analyzes strengths, weaknesses, opportunities, and threats of 4DPC
in construction application sceneries.

Figure 1 Summary of key findings from SWOT analysis
95

3 Strength

100

105

110

As 4DPC data records the additional time dimension compared with the 3D point cloud,
it can help update real-time environment information[12], which is essential for
continuous and disorderly changing construction scene. To facilitate machine
automation and digitalization management in construction site, it is crucial to
understand dynamic 3D environments for robotics agents and many other applications.
4DPC could offer an effective measure to construct and update the dynamic
environment modeling, which could serve for the automation application.
Compared with conventional videos that are widely used in monitoring and
detection work, the 4DPC shows its advantages in covering more precise geometry
dynamics[11]. The reason behind that is point clouds do not suffer as much from
quantization errors compared to other geometric representations such as grids.
Therefore, point cloud video is more suitable for some applications requiring dynamic
detail detection, and typical instances include calculating the acceleration of a moving
object or identifying human activities. For human action recognition, one typical
application of 4DPC video is to analyze the technical action of the professional athlete,
and another one is to help robotics understand human-to-robot gesture instructions in
human-computer interaction.
Another advantage of point cloud video compared with conventional video is good

115

120

125

130

135

adaptability in poor visibility environments[11]. Point cloud video exhibits its advantage
over conventional image-based video for some applications, including low-light work
environments at night, such as autonomous driving and construction site monitoring.
The strength is that point cloud Lidar is an active system generating light, while
conventional video requires an external light source.
In addition, the 4DPC could provide localization and mapping information due to
its principle of signal acquisition and the coordinate information contained[14].
Specifically, it could serve for simultaneous localization and mapping (SLAM), an
algorithmic effort to solve the issue of mapping an unknown area while navigating that
environment using the map.
It is easier and more efficient to acquire the full-view spatial-temporal data of the
whole environment using a set of 4DPC LiDAR. On the construction site, mechanical
equipment, material accumulation, and as-built buildings can cause visual blind spots
if a single sensing device such as a camera or LiDAR is used. Therefore, multiple
LiDARs must be combined to eliminate the visible bind spots. Compared with other
video resources, such as image-based video, the alignment complexity for multi-source
point cloud data is lower.
Finally, compared with other sensing devices such as 360° virtual reality cameras,
Internet of Things (IoT), and Real-time kinematic positioning (RTK), the cost of
common 4D LiDAR Sensors such as Livox is low. For example, the price of a highperformance LiDAR designer for L3/L4 autonomous driving is about 1299 USD, while
the price of the basic version of the virtual reality camera (Insta360 Pro 2) is around
5280 USD. Therefore, the low cost and minimal infrastructure requirements could be a
strength of the 4DPC to be used in the construction sector.

3 Weaknesses
140

145

150

155

For some building materials, such as glass, LiDAR waves may pass wholly or partially
transmit through the material. The possible result is no reflected wave is collected or a
large number of virtual points are collected, which depends on the light transmittance
of the material[15]. If no point is generated, point cloud data cannot cover the object
composed with a light-permeable material. If mobile robotics relying on LiDAR for
navigation cannot recognize them as obstacles, some unexpected collision can happen.
In addition, virtual points' reflection noise may negatively impact 3D reconstruction
and other related techniques.
In addition, LiDAR waves may experience refraction when meeting water bodies;
thus, the acquired point cloud is not accurate in those locations[16]. The problem is first
discussed in autonomous driving since the road's puddle is always considered an infinite
black hole. Because there is often stagnant water after rain on the pavement or
component surface in the construction site, the use of 4DPC also needs to be concerned
about such issues. Otherwise, some unexpected errors may be caused without
awareness.
Enormous data volume is a critical issue in the 4DPC[17]. For LiDAR with higher
scanning density and scanning frequency, the amount of data acquired per unit time is

160

165

170

175

180

185

190

195

also more significant. Although much research on point cloud video systems focuses
on data compression of the point cloud, the large size of 4DPC still puts higher
requirements for storage systems and transmission systems.
The processing of stationary 3D point clouds may be done using a variety of
methods. However, 4DPCs have unique characteristics that cannot be handled by these
conventional methods[18]. Modeling the Spatial-temporal structure, such as 3D action
recognition or 4D semantic segmentation, in a 4DPC is highly challenging because
random and unordered coordinate sets inconstant point cloud density of individual
regions over time and inconsistent points emerging across different sets/frames.
Meanwhile, achieving real-time processing of 4DPC is compute-intensive due to
large data volumes and complex modeling processes[19]. For a disordered environment
such as a construction site, even some basic processing procedures such as point cloud
segmentation can be challenging to achieve real-time or near-real-time completion.

4 Opportunities

Digital 3D environments are already crucial components of construction and are on the
critical route to complete end-to-end site autonomy, but existing solutions lack realtime functionality[20]. Using 4DPC, the digitalization construction environment could
be real-time established and updated. Therefore, the 4DPC provides a chance to create
a digital twin of the construction site. First, It may serve as a platform for the digitization
and automation of the construction industry, which is necessary to encourage the use of
automation and autonomy in the sector.
Second, a series of 4DPC LiDAR could help establish dynamic workspace
modeling of the construction site. The real-time updating modeling is crucial for giving
operators of heavy equipment, such as cranes, real-time support, including danger
identification, collision avoidance, and 3D visualization[21]. The real-time environment
updating enables operators to quickly detect their surroundings and perform
manipulation activities safely and effectively. In addition, a site safety alert system
could be established using dynamic modeling based on a 4DPC, and it has the potential
to help reduce the construction fatalities such as being struck by a moving vehicle.
Third, the 4DPC could help create the 4D building information modeling (BIM),
which enable users represent, evaluate, and analyze all aspects of the BIM process
(graphical models, management, costs, resources, safety issues, etc.) from a temporal
perspective[2]. The 4D BIM could help achieve dynamic management assignments such
as progress monitoring and checking of construction projects. In recent years, although
BIM models have been widely adopted in construction projects and have become
necessary in the design stage in many countries and regions, they are not effectively
used in the construction stage. One important reason is that there is no effective method
to bridge the dynamic construction site with the BIM models. The use of 4D BIM will
improve the current situation.
In addition, by integrating with Unity3D, a versatile 3D engine to enable the user
to view and interact with the environment, the 4DPC could provide a virtual reality (VR)
platform for remote managers to have an immerse experience in the construction site.

200

205

Such experience could help increase the collaboration efficiency of different parties and
help them make optimal decisions according to the site's real-time situation.
Finally, 4DPC records could be used for productivity assessment and optimization.
Benefiting from full-view monitoring of construction sites, the movement path of
workers and machine facilities could be extracted and analyzed. The productivity of
different roles could be assessed, and optimization work could be conducted for a
similar construction project in the future.

5 Threats

210

215

220

225

230

235

Point cloud Lidar is mainly used to detect environmental information in different
scenarios in unmanned applications. However, point cloud lidar would have a poor
performance in rain, fog, and dust environments due to its inherent characteristics of
Lidar[22]. It will recognize rain, fog, and dust as physical things such as obstacles. The
root cause of this problem is that there will be reflections when the beams of several
Lidars hit rain, fog and dust, LiDAR observation would become noisy owing to the
reflection and attenuation of raindrops, or even LiDAR will judge that the rain, fog, and
dust are obstacles, resulting in misidentification. Misidentification caused by rain, fog,
and dust has become problematic in Lidar applications. The above are the common
problems of point clouds, which also occur in 4DPCs. On construction site, dust and
rain may be the main threat that needs to be considered for 4DPC applications. In some
special scenarios, such as the earth excavation stage or rainy day, 4DPCs should be used
with caution, or other sensors need to be supplemented.
Some other emerging technologies, such as AR cameras, maybe a potential
competitor of 4DPC LiDAR for reality capturing. AR camera has already been used in
many areas, such as navigation and education applications[23], and also has the potential
to model the real-time digital twin of the construction site.
In addition, there is no construction standard or policy for establishing a dynamic
digital twin platform-based 4DPC in Hong Kong and mainland of China. The dynamic
4DPC-based digital twin will revolute the traditional schedule-based construction
management strategy. The benefit and necessity of the application of the 4DPC in the
construction industry need to be further explored and discussed from economic and
social aspects.
Acceptance of workers and managers is also one of the factors that should be
considered. In the process of 5G technology promotion, there have been civilian
boycotts in many parts of the world. The main reason is that people worry about their
health being affected by the signal radiation from the 5G infrastructure. A similar
situation may also happen in the application of a 4DPC on the construction site since
LiDAR would emit light waves toward workers.

6 Summary

240

In summary, the 4DPC appears to be a promising applied technology to promote the
development of digital twins for the construction sector. Due to its spatial-temporal data
structure, the 4DPC could update the environmental information in real time and
provide more precise geometry dynamics. In addition, the unique perception

245

250

255

mechanism of LiDAR provides a 4DPC with good adaptability in a poor visibility
environment and localization and mapping functions. Meanwhile, low cost and minimal
infrastructure requirements are potential advantages for application in the construction
industry. However, practitioners and researchers should also be aware of the
limitations, including the inaccurate result when meeting water and light permeable
construction materials, large data volume, and the challenging and compute-intensive
modeling process of the 4DPC. After fully understanding its advantages and
disadvantages, potential opportunities are proposed from five aspects, i.e., for
facilitating construction digitalization and automation, workspace assistance, 4D BIMbased construction management, virtual environment interaction, and productivity
assessment and optimization. The final analyzes the possible threats of the 4DPC in
construction scenarios in terms of the adaptability of the construction site, competitors'
acceptance of workers and managers, and related policies.
On the one hand, as stated by Düking[24], the SWOT analysis process is inherently
subjective and has several inherent limitations. On the other hand, it is anticipated that
the results of this work will aid researchers and practitioners in determining the
suitability and feasibility of the 4DPC for various application situations in the
construction industry.

References
260

265

270

275

280

285

[1] Opoku, D. G. J., Perera, S., Osei-Kyei, R., & Rashidi, M. (2021). Digital twin
application in the construction industry: A literature review. Journal of Building
Engineering, 40, 102726.
[2] Boje, C., Guerriero, A., Kubicki, S., & Rezgui, Y. (2020). Towards a semantic
Construction Digital Twin: Directions for future research. Automation in
Construction, 114, 103179.
[3] Salami, D., Palipana, S., Kodali, M., & Sigg, S. (2020, September). Motion pattern
recognition in 4d point clouds. In 2020 IEEE 30th International Workshop on
Machine Learning for Signal Processing (MLSP) (pp. 1-6). IEEE.
[4] Xue, F., Lu, W., Chen, K., & Webster, C. J. (2019). BIM reconstruction from 3D
point clouds: A semantic registration approach based on multimodal optimization
and architectural design knowledge. Advanced engineering informatics, 42,
100965.
[5] Bhople, A. R., Shrivastava, A. M., & Prakash, S. (2021). Point cloud based deep
convolutional neural network for 3d face recognition. Multimedia Tools and
Applications, 80(20), 30237-30259.
[6] Zhang, Y., Zhou, Z., David, P., Yue, X., Xi, Z., Gong, B., & Foroosh, H. (2020).
Polarnet: An improved grid representation for online lidar point clouds semantic
segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision
and Pattern Recognition (pp. 9601-9610).
[7] Xue, F., Lu, W., Webster, C. J., & Chen, K. (2019). A derivative-free optimizationbased approach for detecting architectural symmetries from 3D point clouds.
ISPRS Journal of Photogrammetry and Remote Sensing, 148, 32-40.
[8] Wu, Y., Shang, J., & Xue, F. (2021). Regard: Symmetry-based coarse registration
of smartphone’s colorful point clouds with cad drawings for low-cost digital twin
buildings. Remote Sensing, 13(10), 1882.

290

295

300

305

310

315

320

325

330

[9] Yuan, L., Guo, J., & Wang, Q. (2020). Automatic classification of common building
materials from 3D terrestrial laser scan data. Automation in Construction, 110,
103017.
[10] Wen, H., Liu, Y., Huang, J., Duan, B., & Yi, L. (2022). Point Primitive
Transformer for Long-Term 4D Point Cloud Video Understanding. arXiv preprint
arXiv:2208.00281..
[11] Fan, H., Yang, Y., & Kankanhalli, M. (2021). Point 4d transformer networks for
spatio-temporal modeling in point cloud videos. In Proceedings of the IEEE/CVF
Conference on Computer Vision and Pattern Recognition (pp. 14204-14213).
[12] Fan, H., Yu, X., Ding, Y., Yang, Y., & Kankanhalli, M. (2022). PSTNet: Point
spatio-temporal convolution on point cloud sequences. arXiv preprint
arXiv:2205.13713.
[13] Shi, H., Lin, G., Wang, H., Hung, T. Y., & Wang, Z. (2020). Spsequencenet:
Semantic segmentation network on 4d point clouds. In Proceedings of the
IEEE/CVF conference on computer vision and pattern recognition (pp. 45744583).
[14] Wang, Z., Li, W., Shen, Y., & Cai, B. (2020). 4-D SLAM: An efficient dynamic
Bayes network-based approach for dynamic scene understanding. IEEE Access, 8,
219996-220014.
[15] Gao, R., Li, M., Yang, S. J., & Cho, K. (2022). Reflective Noise Filtering of LargeScale Point Cloud Using Transformer. Remote Sensing, 14(3), 577.
[16] Chen, L., Yang, J., & Kong, H. (2017, May). Lidar-histogram for fast road and
obstacle detection. In 2017 IEEE international conference on robotics and
automation (ICRA) (pp. 1343-1348). IEEE.
[17] Li, L., Li, Z., Zakharchenko, V., Chen, J., & Li, H. (2019). Advanced 3D motion
prediction for video-based dynamic point cloud compression. IEEE Transactions
on Image Processing, 29, 289-302.
[18] Salami, D., Palipana, S., Kodali, M., & Sigg, S. (2020, September). Motion pattern
recognition in 4d point clouds. In 2020 IEEE 30th International Workshop on
Machine Learning for Signal Processing (MLSP) (pp. 1-6). IEEE.
[19] Liu, Z., Li, Q., Chen, X., Wu, C., Ishihara, S., Li, J., & Ji, Y. (2021). Point cloud
video streaming: Challenges and solutions. IEEE Network, 35(5), 202-209.
[20] Walker, R., Smith, S., & Bosché, F. (2021). Enabling operational autonomy in
earth-moving with real-time 3D environment modelling. In ISARC. Proceedings
of the International Symposium on Automation and Robotics in Construction (Vol.
38, pp. 145-152). IAARC Publications.
[21] Fang, Y., Cho, Y. K., & Chen, J. (2016). A framework for real-time pro-active
safety assistance for mobile crane lifting operations. Automation in Construction,
72, 367-379..
[22] Heinzler, R., Piewak, F., Schindler, P., & Stork, W. (2020). Cnn-based lidar point
cloud de-noising in adverse weather. IEEE Robotics and Automation Letters, 5(2),
2514-2521.
[23] Chidsin, W., Gu, Y., & Goncharenko, I. (2021). AR-based navigation using RGBD camera and hybrid map. Sustainability, 13(10), 5585.
[24] Düking, P., Holmberg, H. C., & Sperlich, B. (2018). The potential usefulness of
virtual reality systems for athletes: a short SWOT analysis. Frontiers in Physiology,
9, 128..