Remote sensing of 4D point cloud (4DPC) for digital twin
construction: A time-dynamic, occlusion-minimized data
source exploiting low-cost LiDAR sensors
5

Fan Xue 1*, Yibo Wang 2, Yijie Wu 3, Zhe Chen 4, Sou-Han Chen 5 and Dong Liang6
This is the authors’ version of the paper:
Xue, F., Wang, Y., Wu, Y., Chen, Z., Chen, S.-H. & Liang, D. (2025).
Remote sensing of 4D point cloud (4DPC) for digital twin construction: A
time-dynamic, occlusion-minimized data source exploiting low-cost LiDAR
sensors. Proceedings of the 29th International Symposium on Advancement of
Construction Management and Real Estate, Springer, in press.
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: [LINK_TO_SPRINGERLINK]. Any uses other than personal and academic
purposes must obtain appropriate permissions from Springer first.

10

Abstract: In the era of smart construction, many sensing technology systems, such as the Internet
of Things, AI cameras, and terrestrial laser scanning, have been applied to collect data on multiobjective and highly time-dynamic construction activities on site. However, these systems are
handicapped in computing for civil engineering by various issues, such as cost, data coverage,
accuracy, and battery life. This paper presents a novel time-dynamic 4D (x, y, z and time) point cloud
(4DPC) data streaming system for highly time-dynamic construction activities. The 4DPC sensing
device in this paper integrates an edge computer and self-driving car’s low-cost LiDAR sensor at
1* Xue, F., PhD

Corresponding author, 512 Knowles Building, Department of Real Estate and Construction, The University of
Hong Kong, Pokfulam, Hong Kong;
E-mail: xuef@hku.hk
Tel: +852 3917 4174
2 Wang, Y., MSc

Department of Real Estate and Construction, The University of Hong Kong, Pokfulam, Hong Kong; e-mail:
yibo@hku.hk
3 Wu, Y., MSc

Department of Real Estate and Construction, The University of Hong Kong, Pokfulam, Hong Kong; e-mail:
yijiewu@connect.hku.hk
4 Chen, Z., MSc

Department of Real Estate and Construction, The University of Hong Kong, Pokfulam, Hong Kong; e-mail:
cz77@connect.hku.hk

5 Chen, S.-H., MSc

Department of Real Estate and Construction, The University of Hong Kong, Pokfulam, Hong Kong; e-mail:
hankchen@connect.hku.hk
Liang, D., MSc
Department of Real Estate and Construction, The University of Hong Kong, Pokfulam, Hong Kong; e-mail:
leodong@connect.hku.hk

6

15

20

less than US$1,000. The 4DPC platform employs a cloud server collecting 4DPC data streams from
multiple devices offline or via Wi-Fi or 5G, and visualizes the data streams on a GIS platform. In
contrast with the existing technologies, the presented platform offers a novel, low-cost, timedynamic, and occlusion-minimized 4DPC data source that opens a new avenue to various digital
twin construction applications.
Keywords: 4D point cloud (4DPC); digital twin construction; time-dynamic LiDAR data source;
sensor fusion; cloud-edge system

1 Introduction

25

30

35

40

45

50

55

60

Smart construction aims to adopt and develop information technology to improve productivity,
reduce costs, and monitor risks at construction sites. Digital twin construction (DTC), which
promises understanding and reasoning, gained increasing attention in recent years (Zhao et al. 2023).
To facilitate the development of smart construction, researchers and experts focused on how to use
novel sensors such as the Internet of Things (IoT), Artificial Intelligence (AI) cameras, and
terrestrial laser scanning to replace a portion of management or survey work in the construction
industry (Štefanič & Stankovski 2018). However, IoT sensors and AI cameras have several
limitations (Liang & Xue 2022). IoT technology is vulnerable to poor networks and high costs (AlSharekh & Al-Shqeerat 2019), and AI cameras lack depth (Chen et al. 2017).
With the rapid development of sensing technologies, LiDAR sensing evolves swiftly for smart
city applications, such as robotics and self-driving cars. There are a variety of Light Detection And
Ranging (LiDAR) data, of which the time-dynamic 4D (xyz and time) point clouds (4DPC) can
simultaneously collect both spatial and temporal information (Liang & Xue 2022). The LiDAR
sensors also became affordable due to the scale effect. However, the volume of the 4DPC data can
be huge, thereby leading to high bandwidth during transmission or local storage. Furthermore, there
can be strong occlusion, or obstruction of laser ‘eyesight’, leading to large ‘shadow’ areas in the site
area. Also, the common 4DPC LiDAR sensors are designed for moving objects, such as robots and
cars. In summary, there is a gap in the literature on how to integrate 4DPC remote sensing into the
scenarios of digital twin construction.
This paper presents a 4DPC data streaming platform to address the three limitations above for
monitoring highly time-dynamic construction activities for DTC. This platform includes multiple
4DPC sensing edge devices and one cloud service. First, 4DPC data is captured by an edge computer
with a self-driving car’s low-cost LiDAR sensor. Then, multiple 4DPC edge devices are planned to
minimize occlusion before the construction activities. Finally, a cloud server collects the 4DPC data
streams offline or via WiFI or 5G with data compression options enabled. The platform's feasibility
was validated with a couple of real-world pilot projects. As far as we are concerned, the multi-sensor
fusion architecture of the presented 4DPC platform is the first of this kind of research, which
contributes to the body of knowledge of real-time site monitoring and data source of DTC.

2 Related Works
There exist schools of smart technologies at construction sites, especially for large civil engineering
projects. The management and coordination of construction projects using an IoT cloud-based
platform were validated by real-world data under realistic conditions (Bucchiarone et al. 2019). In
addition, real-time video surveillance captured by AI cameras was applied to avoid the arisen of
construction accidents (Luo et al. 2020). An edge-based solution for video surveillance in the smart
construction site assisted by a graph neural network was presented to address the limitations of video
surveillance (Ming et al. 2022). However, the high-resolution camera has an inevitable limitation in
that it cannot work at night because of the need for illumination (Vargas et al. 2021). Point clouds
generated from photogrammetry and laser scanning were compared and utilized for smart heavy
equipment planning (Moon et al. 2019). Besides, the benefits of the use of 5G technology in the
construction industry were analyzed and a global framework for the application of 5G technology
was also presented (Mendoza et al. 2021).

65

70

75

Compared with traditional 3D point clouds collected by fixed-point laser scanning (Tang et al.
2022; Zhang & Arditi 2020), 4DPC collected by compact and portable LiDAR sensors can
simultaneously capture the spatial and temporal information of highly time-dynamic activities.
Some sensor models, such as DJI Livox Mid-70, have detection range at 260 m, angular resolution
at 0.1°, and 10~20 frames per second. Therefore, 4DPC gained increasing attention in various
industries. The construction site monitoring typically needs in-person inspections, thereby resulting
in high labor costs. To overcome this limitation, 4DPC can be used for automatic monitoring (Cheng
et al. 2022). 4DPC data can also be used to conduct semantic registration for monitoring cranerelated activities on the construction site (Liang et al. 2023). However, the previous research mainly
focused on the usage and postprocessing of 4DPC, and the capture of 4DPC data is still on-site. It
should be noted that 5G technologies can be employed to improve the data transmission speed,
thereby facilitating data collection remotely. Therefore, remote sensing technologies of 4DPC data
in the construction industry can be further explored and researched.

3 The proposed 4D point cloud (4DPC) sensing platform for DTC
3.1 The general framework

80

Figure 1 shows the architecture of the proposed 4D point cloud (4DPC) sensing platform. In general,
the architecture of the platform has three layers, as shown in Figure 1. The following subsections
describe the details.

Figure 1. Architecture of the proposed 4D point cloud (4DPC) sensing platform.

3.2 4DPC sensing device
85

90

95

Each 4DPC sensing device has one sensor and one edge device. There are multiple choices in the
market for collecting 4DPC. We selected a low-cost LiDAR sensor model, the DJI Livox Mid-70.
The sensor has an appropriate view angle (about 70° cone field of view) and depth (up to 260m). It
emits an infrared laser for range detection, so it works well for nighttime construction work. The
edge device was a Raspberry Pi 4. A set of device cost about USD 993 (Liang et al. 2024). The edge
device runs a lightweight Linux OS and is good for programable extensions. The 4DPC data
protocol was the Livox SDK (https://github.com/Livox-SDK/Livox-SDK) over HTTP protocol. The
CPU load and temperature are also collected and reported in addition to 4DPC data.
Multiple edge devices can be installed at different sensor locations (xi, yi, zi) and heading
directions to. 4DPC data from different edge devices can be merged into the universal xyz-time
coordinate, according to the sensor pose and synchronized time. An optimized multi-sensor fusion
plan can minimize occlusion. Indoor-outdoor fusion is also enabled by a multi-sensor fusion plan.

3.3 4DPC data transmission and streaming

100

The collected 4DPC data can be streamed to a cloud server via WiFi or 5G. If Internet is not available
at the site, the data can also be transmitted offline via a USB drive. Whether online streaming or
offline transmission, the data volume is enormous, leading to possibly expensive 5G data plans.
The platform implements two value functions to handle the costs of uploading massive volume
of 4DPC data. The first function is a Zip/7-Zip-based 4DPC file-level compression. The file-level
compression does not change the bytes of Livox SDK’s format, but does reduce the file size and

105

bandwidth considerably. The other function is a streaming-on-demand function that streams live
data only for user-demanded (short) periods of time, such as five-minute crane hoisting events or 1minute daily progress checking (in the early morning every day).

3.4 Integrated 4DPC source and exchange

110

A cloud server integrates and exchanges the 4DPC data with DTC and end users. First, it receives
4DPC from edge devices with the sensor poses. The sensor poses are presumed to be fixed locations,
as demonstrated in Figure 2a. A user can set up the sensor pose according standard coordinate system,
e.g., the Hong Kong 1980 Grid System (EPSG:2326) shown in Figure 2a. The edge device’s CPU
load and temperature are also readable from the cloud server, as shown in Figure 2b.

Figure 2. Integration of multiple 4DPC edge devices

115

120

Then, the cloud server employs a technological pipeline, as shown in Figure 3, to convert the
4DPC data format. The conversion starts from the .7z archive format and unzips the Livox native
files (.lvx). Then, the 4dPC data is registered to standard coordinates concerning the sensor pose, as
shown in Figure 2a, to be converted to a common LiDAR format .las for exchange. A part of this
conversion was reported in Liang et al. (2023). In addition, the 4DPC data is geo-referenced and
converted to the .pnts format for live visualization on the Cesium platform. The visualization effects
were reported in Liang et al. (2024).

Figure 3. 4DPC data format conversion pipeline on the platform

4 Pilot Tests
4.1 Cases description
125

Two cases studies were conducted in Hong Kong. The first case was a footbridge construction
project at midnight on June 28, 2022. The aims of first case test were edge device planning for
occlusion minimization, offline transmission, and visualization. The other case was an excavation
project in November 2023, where the aims of test included unattended 4DPC streaming and remote
edge device control.

130

4.2 Results of occlusion minimization, transmission and visualization in pilot 1
We planned two edge devices to cover the target site area, as shown in Figure 4. The site was small,
so that the sensor detection range achieved far beyond the site. More details can be found in Liang
et al. (2023; 2024).

135

140

145

Figure 4. 4DPC edge device planning for occlusion minimization (Liang et al. 2023; 2024)

Each edge device recorded about 1.4 MB of 4DPC data per second. However, we found one
device did not work stably. The No. 1 edge device attached to the fourth floor’s 220V power supply
recorded the whole period of 6.5 hours of construction activities. In contrast, the same edge device
(No. 2) attached to the site’s 220V power rebooted from time to time. The main reason could be the
voltage surges caused by heavy construction equipment.
Figure 5 shows the visualization of 4DPC on the Cesium. The offline transmitted 4DPC data
was integrated to Hong Kong’s local grid system. The 4DPC files were successfully converted to .las
formats and .pnts formats for exchange and visualization. In addition, 3D building information
models and Geographical Information System objects were programable compatible with the timedynamic 4DPC data.

Figure 5. Visualization of 4DPC of pilot case #1 on Cesium

4.3 Results of unattended 4DPC streaming and remote device control in pilot 2
150

Table 1 lists the comparison of online 4DPC data streaming methods. The original Livox file format
(.lvx) was about 1.4 megabytes (MB) per second, or 11.4 megabits per second (Mbps) bandwidth.
In contrast, the Zip files achieved a 1:0.334 compression ratio, without any data loss. The best results
were from 7-Zip and XZ Utils. Both achieved a 1:0.235 compression ratio, which indicated a 76.5%

saving in terms of 5D data plan.

155

160

165

Table 1. Comparison of online data streaming with file compression methods
Group
Method
File size (KB) Compression ratio
Original .lvx
–
1388
1:1
Gzip
558
1:0.402
Zip
bzip2
463
1:0.334
XZ Utils
326
1:0.235
Advanced zip
7-Zip
326
1:0.235

The remote sensing device control functions in Figure 2 worked well as designed. An edge
device can be set up the sensor pose and checked for the latest system loads. A user could request
all sensors to stream 4DPC data for a period of time, e.g., 3 seconds, 1 minute, or 3 minutes, by
clicking the buttons on the platform UI.
However, we encountered one issue of accumulated latency of on-demand 4DPC streaming,
as shown in Figure 6. The 30-second 4DPC data started streaming with a latency at 6 seconds, and
ended with a latency increased to 30 seconds. The latency showed a continuous increase in the test.
Possible reasons included the complex electromagnetic environment at the construciton site and
unable 4G/5G connection.

Figure 6. Accumulated latency in seconds using 4G/5G network (screenshot of cloud database)

4.4 Discussions
170

175

The two tests validated the feasibility of the proposed 4DPC sensing platform. In summary, the first
pilot test showed that the proposed 4DPC platform employs edge sensing devices for time-dynamic
4DPC data, integrates multiple devices for occlusion minimization, and converts 4DPC data into
various formats for exchange and visualization. The second test confirmed the 4DPC streaming with
a latency issue and remote edge device control.
In comparison to traditional AI camera surveillance solutions, the 4DPC platform has several
advantages. First, the 4DPC data offers accurate depth and 3D information. Also, the 4DPC sensors
work perfectly without good lighting at midnight. The fusion of 4DPC data streams from multiple
sensors is natural and precise, thanks to standard geospatial coordinate systems.
Disadvantages of the 4DPC platform, nevertheless, were also found. The first was the vast data

180

185

190

195

200

volume of 4DPC, which led to large disk size and live streaming latency. Although the 7-Zip was
proven useful in saving 76.5% of the size, the compressed data volume streaming was still
unsatisfactory and accumulated over time. Thus, stronger and domain-specific 4DPC compression
is a promising future research direction. Another one was the low-level point density outside the
center of the cone field of view of LiDAR sensors. The fast-evolving LiDAR hardware may resolve
this problem in the future. The third drawback is the lack of colors in the 4DPC in this paper. Industry
practitioners can supplement a camera to the 4DPC edge device in this paper.
Several challenges have also been identified in downstream DTC applications. First is the
difficulty in object detection and activity monitoring using the 4DPC data source. One recent work
can be referred to (Liang et al. 2024). Another one is the complexity of DTC, barring proactive risk
identification and smart responses.

5 Conclusion
The time-dynamic 4DPC stream is a promising data source for DTC. This paper presents a cloudedge system architecture for construction sites’ 4DPC sensing that exploits self-driving cars’ lowcost LiDAR sensors. The platform employs three layers: (i)_4DPC sensing edge devices, (ii) data
transmission and streaming, and (iii) 4DPC integration and exchange. Two pilot tests confirmed that
the platform collects occlusion-minimized 4DPC data and works unintendedly for 4DPC streaming.
The proposed 4DPC streaming platform is the first of its kind in the literature, as far as we are
concerned. In contrast with the existing technologies, the presented platform offers a novel, lowcost, time-dynamic, and free-of-occlusion 4DPC data source that opens a new avenue to various
digital twin construction applications. Future work includes testing advanced 4DPC sensors,
advanced 4DPC data compression methods, integration with RGB camera, semantic object
detection and pose estimation for construction activities in 4DPC, and proactive risk identification
and smart responses.

Acknowledgement
This study in this paper was supported by The University of Hong Kong (No. 2307102469).
205

210

215

220

225

References
Al-Sharekh, S. I. & Al-Shqeerat, K. H. (2019). Security Challenges and Limitations in IoT
Environments. IJCSNS International Journal of Computer Science and Network Security,
19(2), 193-199.
Bucchiarone, A., Sanctis, M. D., Hevesi, P., Hirsch, M., Abancens, F. J., Vivanco, P. F., Amiraslanov,
O. & Lukowicz, P. (2019). Smart Construction: Remote and Adaptable Management of
Construction Sites through IoT. IEEE internet of things magazine, 2(3), 38-45.
doi:10.1109/IOTM.0001.1900044
Chen, J., Fang, Y. & Cho, Y. K. (2017). Real-time 3D crane workspace update using a hybrid
visualization approach. Journal of Computing in Civil Engineering, 31(5), 04017049.
doi:10.1061/(ASCE)CP.1943-5487.0000698
Cheng, S. Y., Liu, L., Hou, W., Hart, J. R. & Yong, Y. M. (2022). Smart Construction Monitoring
Using Photogrammetry and LiDAR-derived 4D Digital Model: A Case Study from the
Tung Chung New Town Development of Hong Kong. AIJR Proceedings (pp. 129-140).
AIJR. doi:10.21467/proceedings.133.11
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 (pp. 1231-1238). Singapore: Springer.
doi:10.1007/978-981-99-3626-7_95
Liang, D., Chen, S.-H., Chen, Z., Wu, Y., Chu, L. Y. & Xue, F. (2024). 4D point cloud-based spatialtemporal semantic registration for monitoring mobile crane construction activities.
Automation in Construction, 165, 105576. doi:10.1016/j.autcon.2024.105576
Liang, D., Chen, Z., Kong, L., Wu, Y., Chen, S.-H. & Xue, F. (2023). 4D Point Cloud (4DPC)-driven
real-time monitoring of construction mobile cranes. European Conference on Computing
in Construction (pp. 0-0). Flemish Region, Belgium: European Council on Computing in

230

235

240

245

250

255

260

Construction (EC3). doi:10.35490/EC3.2023.258
Luo, H., Liu, J., Fang, W., Love, P. E., Yu, Q. & Lu, Z. (2020). Real-time smart video surveillance
to manage safety: A case study of a transport mega-project. Advanced Engineering
Informatics, 45. doi:10.1016/j.aei.2020.101100
Mendoza, J., de-la-Bandera, I., Álvarez-Merino, C. S., Khatib, E. J., Alonso, J., Casalderrey-Díaz,
S. & Barco, R. (2021). 5G for Construction: Use Cases and Solutions. Electronics, 10(14),
1713. doi:10.3390/electronics10141713
Ming, Z., Chen, J., Cui, L., Yang, S., Pan, Y., Xiao, W. & Zhou, L. (2022). Edge-Based Video
Surveillance With Graph-Assisted Reinforcement Learning in Smart Construction. IEEE
Internet of Things Journal, 9(12), 9249 - 9265. doi:10.1109/JIOT.2021.3090513
Moon, D., Chung, S., Kwon, S., Seo, J. & Shin, J. (2019). Comparison and utilization of point cloud
generated from photogrammetry and laser scanning: 3D world model for smart heavy
equipment
planning.
Automation
in
Construction,
98,
322-331.
doi:10.1016/j.autcon.2018.07.020
Štefanič, M. & Stankovski, V. (2018). A review of technologies and applications for smart
construction. Proceedings of the Institution of Civil Engineers-Civil Engineering, 172(2),
83-87. doi:10.1680/jcien.17.00050
Tang, X., Wang, M., Wang, Q., Guo, J. & Zhang, J. (2022). Benefits of Terrestrial Laser Scanning
for Construction QA/QC: A Time and Cost Analysis. Journal of Management in
Engineering, 38(2). doi:10.1061/(ASCE)ME.1943-5479.0001012
Vargas, J., Alsweiss, S., Toker, O., Razdan, R. & Santos, J. (2021). An Overview of Autonomous
Vehicles Sensors and Their Vulnerability to Weather Conditions. Sensors, 21(16), 5397.
doi:10.3390/s21165397
Yang, J., Park, M.-W., Vela, P. A. & Golparvar-Fard, M. (2015). Construction performance
monitoring via still images, time-lapse photos, and video streams: Now, tomorrow, and the
future. Advanced Engineering Informatics, 29(2), 211-224. doi:10.1016/j.aei.2015.01.011
Zhang, C. & Arditi, D. (2020). Advanced Progress Control of Infrastructure Construction Projects
Using Terrestrial Laser Scanning Technology. Infrastructures, 5(10), 83.
doi:10.3390/infrastructures5100083
Zhao, R., Chen, Z. & Xue, F. (2023). A blockchain 3.0 paradigm for digital twins in construction
project
management.
Automation
in
Construction,
145,
104645.
doi:10.1016/j.autcon.2022.104645