Content uploaded by Jia-Rui Lin
Author content
All content in this area was uploaded by Jia-Rui Lin on May 31, 2022
Content may be subject to copyright.
Abstract— As an important contributor to GDP growth, the
construction industry is suffering from labor shortage due to
population ageing, COVID-19 pandemic, and harsh
environments. Considering the complexity and dynamics of
construction environment, it is still challenging to develop fully
automated robots. For a long time in the future, workers and
robots will coexist and collaborate with each other to build or
maintain a facility efficiently. As an emerging field,
human-robot collaboration (HRC) still faces various open
problems. To this end, this pioneer research introduces an
agent-based modeling approach to investigate the coupling effect
and scale effect of HRC in the bricklaying process. With multiple
experiments based on simulation, the dynamic and complex
nature of HRC is illustrated in two folds: 1) agents in HRC are
interdependent due to human factors of workers, features of
robots, and their collaboration behaviors; 2) different
parameters of HRC are correlated and have significant impacts
on construction productivity (CP). Accidentally and
interestingly, it is discovered that HRC has a scale effect on CP,
which means increasing the number of collaborated
human-robot teams will lead to higher CP even if the
human-robot ratio keeps unchanged. Overall, it is argued that
more investigations in HRC are needed for efficient construction,
occupational safety, etc.; and this research can be taken as a
stepstone for developing and evaluating new robots, optimizing
HRC processes, and even training future industrial workers in
the construction industry.
I. INTRODUCTION
As a pillar industry of the national economy, the
construction sector builds and maintains buildings and
infrastructures for the support of production, living,
transportation, and healthcare. Despite this, the construction
industry has always been criticized for its harsh working
environment, occupational accidents, and low productivity.
The willingness of workers to stay in the construction industry
continues to be low, and the labor shortage is increasing.
Meanwhile, the trend of global population ageing and the
COVID-19 pandemic further impact the fragile labor market
supply in the construction industry. Therefore, development of
new construction methods, i.e., construction robots, modular
construction, have become the cutting-edge trends in the
construction industry[1].
*Research supported by National Natural Science Foundation of China
(No. 51908323, No. 72091512) and the National Key R&D Program of China
(No. 2018YFD1100900).
Jia-Rui Lin is with the Department of Civil Engineering, Tsinghua
University, Beijing 100084, China (corresponding author, phone:
0086-6278-9225; e-mail: lin611@tsinghua.edu.cn).
Minghui Wu was with the Department of Civil Engineering, Tsinghua
University, Beijing 100084, China. He is now with the Department of Civil
and Environmental Engineering, University of Michigan, Ann Arbor, MI
48109, USA (e-mail: minghuiw@umich.edu).
However, the construction process is highly complex and
dynamic. And, the construction site is a typical unstructured
scene, where the surrounding, layout and the appearance of
building components change frequently. Therefore, compared
with the development and deployment of robots in structured
scenarios of manufacturing, developing fully automated
construction robots in an unstructured and dynamic
construction site is still challenging[2]. It is foreseeable that for
a long time, the construction site will be in a state where
humans and robots coexist and collaborate with each other.
Generally, workers and robots in the construction process
have their own characteristics and advantages. For example,
workers are highly adaptable and have a strong ability to cope
with complex environments, while they have specific
limitations in terms of ergonomics and occupational safety.
They are limited by human factors such as forgetting and
fatigue, as well as risk constraints such as human-machine
collision and harsh environments, which leading to unstable
performance in quality and efficiency. On the contrary, robots
have relatively stable performance in terms of quality and
efficiency, and is applicable in extreme environments.
However, their capacity is limited in complex and dynamic
scenarios due to insufficient adaptability. How to effectively
utilize the advantages of workers and robots to achieve
effective collaboration between the two and maximize
construction efficiency and quality is a key challenge for
human-robot collaboration in the construction industry. To
overcome this challenge, a series of open problems should be
addressed through tight collaboration with experts from area of
construction, robotics, informatics, and ergonomics, etc.
Currently, research on construction robots still focuses on
the development of robots and algorithms for specific
construction tasks, which aim to improve the performance of a
single robot[3]. Less attention is paid to the worker side, and
there still lacks investigations on HRC at a macro level.
To this end, this research conducts a pioneer exploratory
work in HRC for the construction industry. Due to the
complexity, high-risk and irreproducibility of the real site
experiments, this study introduces an agent-based modeling
and simulation approach, to systematically analyze the
interdependency of human factors, robot features, and HRC
parameters as well as their impacts on construction
productivity (CP) in the bricklaying scenario.
II. METHODOLOGY
In view of the flexibility of the agent-based modeling and
simulation[4], this study introduces an agent-based HRC
modeling and simulation approach (Fig. 1), which consists of
four steps: parameter extraction, agent-based modeling,
Exploiting the Power of Human-Robot Collaboration: Coupling and
Scale Effects in Bricklaying*
Jia-Rui Lin, and Minghui Wu
experimental design, and analysis and insights. The following will explain each step in detail.
Parameter Extraction
Site Observation
Literature Review
Agent-based Modeling
Model Architecture
Behaviors & Rules
Analysis & Insights
472
474
476
478
480
482
484
486
488
490
050 100 150 200 250 300
Construction time(hour)
Supplement Limit (SL)
CI=3900 CI=4200 CI=4500 CI=4800
440
445
450
455
460
465
470
475
480
485
490
050 100 150 200 250 300
Construction time(hour)
Supplement Limit (SL)
CI=2700 CI=3000 CI=3300 CI=3600
430
440
450
460
470
480
490
500
050 100 150 200 250 300
Construction time(hour)
Supplement Limit (SL)
CI=1500 CI=1800 CI=2100 CI=2400
430
440
450
460
470
480
490
050 100 150 200 250 300
Construction time(hour)
Supplement Limit (SL)
CI=300 CI=600 CI=900 CI=1200
Coupling Effect
Scale Effect
Theoretical Explanation
Experimental Design
Variable Control
Scenario Design
Simple
Complex
Figure 1. Proposed methodology
A. Parameter Extraction
This step deals with the acquisition of simulation
parameters, and is mainly achieved from two aspects. On one
hand, observation of real site experiment is conducted, video
capture or other techniques are utilized to record the process of
HRC. Later, humans and robots are collaborated can be
extracted from the recorded data. These may include the
sequence of tasks of workers and robots, performance of robots,
frequency and duration of human-robot interaction, etc. on the
other hand, literature review is also adopted to extract data
from published articles. For example, CP of robot-based
bricklaying, and ergonomic models such as forgetting and
fatigue are obtained from literatures [5] and [6].
B. Agent-based Modeling
Based on the parameters extracted above, Anylogic, a
widely used modeling software, is adopted to create the
agent-based model for HRC. The developed model mainly
includes four kinds of agents: workers, robots, bricks, and site
environment recorders. The main relationships of them are
illustrated in Fig. 2. In general, the following parameters are
considered when modeling the HRC process in bricklaying.
• Environment: location and capacity of temporary
storage, location and length of walls, etc.;
• Worker: role, position, speed, mortar removing
performance, fatigue and forgetting model, etc.;
• Robot: position, movement speed, bricklaying
performance, brick capacity, safety space, etc.;
• Collaboration: check interval (CI) that for workers to
check the robot, and supply limit (SL) when the
worker starts supplying bricks, proactive interaction,
etc.
More details could on parameter definition and HRC
modeling can be found in our papers [7] and [8].
Figure 2. Developed agent-based model for human-robot collaboration
C. Experimental Design
In this step, a series of simulation experiments are designed
based on control variable method. That is, simulations based
on different CIs, SLs, and different human-robot interaction
modes, etc., are created and implemented to generate data for
future analysis. Meanwhile, different HRC scenarios, i.e.,
single robot-single worker scenario, multiple robot-multiple
worker scenario, are also designed and simulated.
D. Analysis and Insights
At last, generated data from simulations are visualized,
analyzed to obtained insights on how human and robots are
collaborated. Meanwhile, theoretical model is also developed
to explain the underlying mechanism of HRC.
III. RESULTS
Based on the proposed method, many simulation
experiments were carried out in this study. By data analysis
and theoretical explanation, the following results related to the
interdependency of HRC parameters and the impacts of HRC
on CP were obtained. Note that CP is estimated based on the
construction time of a certain quantity of work.
A. Dynamics and Coupling Effect of HRC
According to Fig. 3, the behavior and workload of different
workers are significantly different due their differences in roles.
The worker who removes the mortar (EMR worker) is always
in a state with intensive workload because his/her performance
is lower than the performance of the robot for bricklaying;
while the brick supplement worker (BS workers) often need to
repeatedly check whether the bricks should be supplied. If the
check interval is not selected properly, there will be many
redundant checking tasks. In addition, due to the safety
distance defined between the robot and the worker, the low
efficiency of the EMR worker may hinders the bricklaying
process of the robot, resulting in intermittent interruptions time
by time.
Redundant
Checking
Intermittent
Interruption
Intensive
Workload
Working
Idle
Setup
Moving
Robot
BS
Worker
Supplying
Idle
Setup
Checking
Removing
Idle
Setup
Figure 3. State changes of workers and robot
The experiment also shows that when the EMR worker
enters a state of fatigue after a long time of intensive work,
his/her efficiency will drop significantly, which may also lead
to intermittent interruptions of robots (Fig. 4). Note that to
demonstrate the results in Fig.3 and Fig. 4, workers are set to
work for a long time, i.e., more than 8 hours a day.
In summary, states of workers and robot are interdependent,
and the performance of workers or robot are influenced by
different parameters, making it difficult to understand and
optimize the HRC process and improve the productivity.
Worker
Fatigue
Intermittent
Interruption
Working
Idle
Setup
Moving
Figure 4. Impact of worker fatigue on the performance of robot
B. Impact of HRC on Construction Productivity
Based on the simulation data of different scenarios, this
research systematically analyzed the effect of different HRC
parameters (including SL, CI, interaction modes) on CP.
For the single robot-single worker (SRSW) scenarios,
when SL are in different ranges, the relationship between CI
and CP follows different rules. As shown in Fig. 5(a) and Fig.
5(b), when the value of SL is large, the total construction time
increases linearly as CI increases. However, when the value of
SL is small, although the total construction time increases as
CI increases, there is a flattening or oscillation range in the
middle. Meanwhile, for the multiple robot-single worker
(MRSW) scenarios in Fig. 5(c) and Fig. 5(d), the relationship
between CI, SL and CP are quite different comparing to SRSW
scenarios. That is, when the SL is small, the construction time
decreases as CI increases; while when the SL is large, the
construction time increases as CI increases.
Figure 5. Impacts of CI and SL on productivity
The analysis shows that there are complex
interdependencies and constraints between various parameters
of the HRC process, which may have different effects on CP.
This also implies that parameters of HRC are tightly coupled
with each other, and more in-depth investigations are needed to
clearly understand and further optimize the HRC processes.
Except for the scenarios where the workers passively check
the status of robots, we also introduced sensors to help robot
proactive feedback its states to the workers. As show in Fig. 6,
comparing to passive interaction mode, proactive interaction
mode between workers and robot can significantly improve the
CP of both SRSW scenarios and MRSW scenarios. In most
cases, the CP can be improved by more than 20%. This shows
that the active communication and interaction between robots
and workers has a very positive significance for improving
efficiency, which means that robots with self-awareness are
much better than traditional ones.
Figure 6. Productivity improvement through proactive interaction
C. Scale Effect of HRC
At last, scenarios related to multiple robot-multiple worker
(MRMW) are also considered and investigated. According to
Fig. 7, the general distribution of construction time in MRMW
scenario is similar to the SRSW scenario. Though dark blue
area is similar between the, some green area in Fig. 7(a) is
replaced by the blue area in Fig. 7(b). This indicates that
productivity increases although the ratio between robots and
workers remains the same. That is, there exists a scale effect
between the collaboration of robots and BS workers, especially
for scenarios with larger CI. This implies that largescale
promotion and application of construction robots on
construction sites may increase the performance of
construction dramatically.
Figure 7. Scale effect when increasing the number of human-robot teams
The underlying mechanism behind scale effect is that the
large portion of overlap between check intervals makes that
the general checking interval (GCI) smaller than the CI
determined by the supplement strategy (Fig. 8). In other words,
when there are two human-robot teams working together,
though the CI of each BS worker keeps unchanged, they can
make GCI smaller by helping their partners check the robot.
Thus, each robot can be checked in a shorter interval, or GCI
in this case. Here we defined this mechanism as mutual help
of workers, which could explain the scale effect theoretically.
Figure 8. Mutual help of workers – a theoretical explanation of scale effect
IV. CONCLUSIONS
In this research, an agent-based modeling and simulation
approach is introduced to investigate the human-robot
collaboration (HRC) process and its impact on construction
performance. With the developed model and various
experiments, it is showed that: 1) multiple parameters related
to HRC, from the human side, the robot side and their
interactions, are highly interdependent and coupled with each
other; 2) impacts of HRC on construction productivity are
complex, parameters should be carefully chosen and proactive
interaction mode is recommended; 3) HRC has a scale effect
on productivity due to mutual help mechanism between
workers, and large scale adoption robots may get a much better
performance than linearly adding multiple human-robot teams.
Overall, this research contributes an integrated approach to
simulate the HRC process and evaluate its impacts on
construction productivity. we also demonstrate a new
paradigm to understand the HRC process in various
robot-based construction scenarios, which can be taken as a
stepstone for the evaluation and development of new robots,
optimization of HRC process to maximize construction
performance and occupational health, and even training of
skilled workers for future construction.
REFERENCES
[1] T. Bock, “The future of construction automation: technological
disruption and the upcoming ubiquity of robotics,” Automat. Constr.,
vol. 59, pp. 113-121, 2015.
[2] G. Carra, A. Argiolas, A. Bellissima, M. Niccolini, and M. Ragaglia,
“Robotics in the construction industry: state of the art and future
opportunities,” In Proc. 35th Int. Symp. Automat. Robot. Constr., Berlin,
Germany, 2018, pp. 866-873.
[3] K. Jung, B. Chu, S. Park, and D. Hong, “An implementation of a
teleoperation system for robotic beam assembly in construction,” Int. J.
Precis. Eng. Manuf., vol. 14, no.3, pp. 351-358, 2013.
[4] J.L. Adler, and V.J. Blue, “A cooperative multi-agent transportation
management and route guidance system,” Transp. Res. Part C: Emerg.
Technol., vol. 10, no. 5-6, pp. 433-454, 2002.
[5] Z. Sun, C. Zhang, and P. Tang, “Modeling and simulating the impact of
forgetting and communication errors on delays in civil infrastructure
shutdowns,” Front. Eng. Manag., vol. 8, pp. 109-121, 2021.
[6] J. Seo, S. Lee, and J. Seo, “Simulation-based assessment of workers’
muscle fatigue and its impact on construction operations,” J. Constr.
Eng. Manag., vol. 142, no. 11, pp. 04016063, 2016.
[7] M.H. Wu, and J.R. Lin, “An agent-based approach for modeling
human-robot collaboration in bricklaying”, in Proc. 37th Int. Symp.
Automat. Robot. Constr., Kitakyshu, Japan, 2020, pp. 797-804.
[8] M.H. Wu, J.R. Lin, and X.H. Zhang, “How human-robot collaboration
impacts construction productivity: an agent-based multi-fidelity
modeling approach,” Adv. Eng. Inform., vol. 52, pp. 101589, 2022.
