ras: a robotic assembly system for steel structure ... · pdf fileregular paper ras: a robotic...

REGULAR PAPER

RAS: a robotic assembly system for steel structure erectionand assembly

Ci-Jyun Liang1 • Shih-Chung Kang1 • Meng-Hsueh Lee1

Received: 4 December 2016 / Accepted: 19 July 2017 / Published online: 13 September 2017

� Springer Nature Singapore Pte Ltd. 2017

Abstract This research focuses on a long-standing, yet

critical problem in the erection of steel structures. In the

current state of practice, steel workers must stand on an

unfinished structure to assist with the assembly of structural

elements manually. They must pull on the wire hanging

under the rigging elements to align the bolting holes of the

moving and fixed elements. This work is often performed

in high places, which can be very risky. Therefore, we have

developed a robotic assembly system (RAS) for steel beam

erection and assembly to prevent workers from having to

work in a high place. The RAS consists of four methods:

rotation, alignment, bolting, and unloading. The rotation

method involves a flywheel installed on top of the rigging

beam, which aims to rotate the beam to the assembly angle.

The alignment method includes both vertical and hori-

zontal alignment. The vertical alignment relies on a camera

and a marker on the column to align the beam altitude. The

horizontal alignment relies on a specially-designed beam,

which allows for it to be smoothly guided into the right

position. The bolting method is used to connect the beam to

a fixed element. We designed an additional guide hole

above each bolt hole. The bolt can be inserted in the guide

hole and slid to the bolt hole. The unloading method is used

to unload the crane cable and the RAS. We use a pin

mechanism for the beam-hook connection so it can easily

be unplugged by a motor. The system is built in a scaled

experimental construction site to validate its feasibility.

The results show that the RAS can operate the assembly

process without humans working at risky heights, and can

complete faster than the traditional method. In conclusion,

we have developed a robotics assembly system that can

help reduce the frequency of accidental falls during the

steel beam assembly process. The RAS adheres to the

process of the current erection method and can be broadly

introduced to existing construction sites.

Keywords Steel beam assembly � Construction robotics �Construction safety � Auto joint � Rotation method

1 Introduction

The steel beam erection and assembly process is always in

the critical construction path and accounts for a high per-

centage of the cost in a large high-rise steel structure

construction project (Chi et al. 2012; Chin et al. 2005;

Pavlovcic et al. 2004); however, it relies largely on manual

labor (Irizarry 2011), which means even a simple human

mistake might result in a serious delay of the entire con-

struction schedule and thus, extra costs to the project

(Peurifoy et al. 2011). Figure 1 illustrates the process of

steel beam erection and assembly. First, ground workers

connect the steel beam to the tower crane hook, then the

tower crane lifts and transports the steel beam to the

assembly position, as shown in Fig. 1a, b. Second, workers

at the construction height align the steel beam to the pre-

cise joint position by hand, by wire, or even by foot, as

shown in Fig. 1c. This step accounts for the highest per-

centage of time spent in the entire process (Chi and Kang

2010). Finally, workers assemble the steel beam with steel

plates and two or three bolts to achieve the temporary

connection, as shown in Fig. 1d. During the process, steel

workers have to stand on a narrow steel bracket or other

& Shih-Chung Kang

[email protected]

1 Department of Civil Engineering, National Taiwan

University, No. 1, Sec. 4, Roosevelt Road, Taipei 10617,

Taiwan

123

Int J Intell Robot Appl (2017) 1:459–476

https://doi.org/10.1007/s41315-017-0030-x

http://crossmark.crossref.org/dialog/?doi=10.1007/s41315-017-0030-x&domain=pdf

http://crossmark.crossref.org/dialog/?doi=10.1007/s41315-017-0030-x&domain=pdf

https://doi.org/10.1007/s41315-017-0030-x

steel beam at a substantial height with only a simple safety

cable. Accidents sometimes happen and can cause serious

injuries or fatalities (Beavers et al. 2009)—falling, being

crushed/struck/hit by an object, and being electrocuted are

the three main categories of fatal events common to the

task of crane erection. Furthermore, the efficiency of the

process is difficult to control due to the impact of manual

labor (Liang and Kang 2014). Therefore, preventing human

workers from having to work at heights is the primary goal

in improving the safety and efficiency of the steel beam

erection and assembly process.

1.1 Crane operation safety and efficiency

Safety and efficiency issues are very important in con-

struction projects (Zhou et al. 2012). Recent erection and

assembly related research has been focused on improving

crane operation (Kang and Miranda 2006) and steel worker

performance (Teizer et al. 2013). Since the crane operator

plays an important role in crane operation, several research

papers focused on crane operator training and blind spot

reduction (Cheng and Teizer 2014; CM Lab 2015; Huang

and Gau 2003). Juang et al. (2013) developed a stereo-

scopic kinesthetic crane training system to train the crane

operator in a realistic approximate approach. Chi et al.

(2012) developed a tele-operated crane interface for a

worker-free construction site. The interface demonstrated

the crane erection status and planning path and informed

the operator when a collision was about to happen. Lee

et al. utilized location tracking sensors and building

information modeling (BIM) (Volk et al. 2014) to set up a

crane navigation system to assist with blind lifting (Lee

et al. 2012). Ray and Teizer (2012) presented a mobile

crane operator head motion estimator to build a map of

dynamic blind spots with a range camera. In addition, path

planning algorithms and visualization techniques have

been utilized to help crane erection operation without

guidance from ground workers on construction site (Chang

et al. 2012; Kang and Miranda 2006; Kang and Miranda

2009; Wang et al. 2011). The genetic algorithm (Yoo et al.

2012) and configuration space (C-Space) (Kang et al. 2009)

methods are normally used in erection path planning.

Zhang and Hammad (2012) proposed the Rapidly-explor-

ing Random Trees Connect-Connect Modified (RRT-Con-

Con-Mod) and Dynamic RRT-Con-Con Modified (DRRT-

Con-Con-Mod) methods to improve the erection path

planning and re-planning. Lei et al. (2013) utilized the

Configuration Space Obstacle (C-Obstacle) method to

check the mobile crane lift path in order to prevent colli-

sions. Hung et al. (2016) proposed HBCD strategies

(Hoisting, Boundary, Capacity, and Direction) to accelerate

the computing time of the mobile crane path planning

algorithm.

Worker safety is also an important issue in erection and

assembly related research (Irizarry and Abraham 2006;

Kim and Kim 2012; Vijay et al. 2006). Irizarry (2011)

Fig. 1 Steel beam assembly

process: a lifting,

b transporting, c aligning and

d bolting

460 Ci-Jyun Liang et al.

123

analyzed the steel erection process and presented the fac-

tors that affect worker performance. Teizer et al. (2013)

utilized Ultra-Wideband sensors to track the ironworkers’

location in the training environment, and virtual reality

techniques to demonstrate the training process for

improving the ironworkers’ education and training method.

Park and Brilakis (2012) proposed a construction worker

detection algorithm to identify the construction worker in

video frames. Lee et al. (2012) utilized an RFID sensor to

monitor the workers’ location at the construction site.

1.2 Steel beam erection and assembly

Based on observations of the steel structure erection and

assembly at a real construction site, we separate the process

into three steps: rotation control, alignment, and bolting, as

shown in Fig. 2. First, the workers rotate the rigging beam

to the correct orientation. Second, the beam is aligned to

the correct position relative to the column. Third, the beam

is assembled to the column with two or three bolts to

complete the temporary connection. The rotation control of

the rigging beam is intended to rotate the beam to the

assigned position and maintain its orientation. A suspen-

sion unit controlled by gyroscopic moment (GYAPTS)

(Wakisaka et al. 2000) and a motor controllable hook block

(Lee and Lee 2014) are two different ways to achieve this

rotation control. The GYAPTS is a gyroscopic device that

attaches to the lifted beam and contains a flywheel. It can

stabilize the lifted beam (passive control) or rotate it to a

precise angle using the moment provided by the flywheel

(active control) (Gajamohan et al. 2012; Gams et al. 2007).

The GYAPTS was implemented in an automatic con-

struction system and used on a reinforced concrete building

site (Wakisaka et al. 2000). Alternately, the motor con-

trollable hook block provides a power resource from the

crane hook. It can simply rotate the rigging beam and

maintain the orientation with a motor connected to the

hook (Lee and Lee 2014; Lee et al. 2012).

In order to automate construction, a suitable manipulator

must be implemented (Gambao et al. 2000; Kahane and

Rosenfeld 2004; Yu et al. 2009). The main purpose of the

manipulator is to align the construction component to the

assigned position and connect it to fixed components. Feng

et al. (2015) developed a marker detecting algorithm for a

mobile robotic manipulator to identify, grasp, and assemble

the construction components. Garg and Kamat (2014)

designed a robotic fabrication mechanism for rebar cages

in concrete construction. Viscomi et al. (1994) utilized a

six degree-of-freedom Stewart platform crane—a three-

dimensional fully controllable manipulator—and an

ATLSS connection to attach the rigging beam. The ATLSS

connection is a joint for fast and easy assembly. In addi-

tion, Quicon� (The Steel Construction Institute 2004), plug

and play joints (Bijlaard et al. 2009), and ConX� (2015)

are all joint innovations well-known for fast connections.

Lee et al. (2012) presented a non-powered multi-beam

lifting system for improving efficiency of the steel beam

erection process. On the other hand, several construction

robot are also used in industry, such as Auto-Claw, Auto-

Clamp, Robotic End-Effector for Big Canopy, and Auto-

mated Building Construction System (ABCS) developed

by Obayashi Corporation (Bock and Linner 2016a). The

ABCS contains an alignment and accuracy measurement

system to check the alignment by vision and laser sensors.

Shimizu Corporation and Samsung Corporation developed

a robot crane end-effector Mighty Jack, Auto-Shackle, and

Mighty Shackle Ace for assisting with steel beam posi-

tioning and installing (Bock and Linner 2016b). Saidi et al.

(2006) proposed a RoboCrane system end-effector to

manipulate rigging beam precisely.

Jung et al. (2013a) developed a robot-based construction

automation system for high-rise buildings. The system

included a construction factory (Kim et al. 2009), a scissor

jack-type manipulator (Jung et al. 2008), and the robotic

beam assembly system (Jung et al. 2013b). The construc-

tion factory is a large and safe workspace, also named Sky

Factory, which is assembled outside the unconstructed

building and can move vertically during the construction

process, like a tower crane, carrying workers and con-

struction machinery. ABCS (Obayashi Corporation), Aka-

tuki 21 (Fujita Corporation), FACES (Goyo), MCCS

(Maeda Corporation), SMART (Shimizu Corporation), and

T-Up (Taisei Corporation) are well-known construction

systems in industry featuring construction factory (Bock

and Linner 2016b). The scissor jack-type manipulator can

lift the steel beam to the assembly location and the robotic

beam assembly system will assemble the steel beam to the

existing column. The robotic beam assembly system

includes a teleoperation system (Jung et al. 2013b), a

transport mechanism (Jung et al. 2013a), a robotic bolting

device (Chu et al. 2013), and a specially-designed steelFig. 2 Three steps for typical steel beam erection and assembly

RAS: a robotic assembly system for steel structure erection and assembly 461

123

beam with an automatic guide rope (Kim et al. 2016). Nam

et al. (1946) introduced a boom-mounted, combined

robotic system and wire-suspended positioning system for

automatic steel beam assembly.

A key aspect of the manipulator is the bolting robot. The

main purpose of the bolting robot is to attach the steel

beam to the column with bolts. The bolting robot utilizes a

camera and a computer vision method to detect bolt holes

(Mo et al. 2014), once detected, the robotic bolting device

will install the steel bolts (Chu et al. 2013).

1.3 Research goal

In this study, we develop a robotic assembly system (RAS)

for steel beam erection and assembly, which aims to

improve the safety and the efficiency of the steel beam

assembly process. The RAS can rotate, align, and bolt the

steel beam without help from steel workers at the height of

construction. Removing steel workers from heights on the

construction site during the steel beam erection and

assembly process prevents falls and injuries when struc-

tures fail, which are the design implications of the pro-

posed robotic assembling system. In addition, the

efficiency of the operation can easily be controlled since

the manual factor has been excluded from the process. In

comparison with previous research, this system is easily-

removable and light-weight, which meets the requirements

of the current erection method and can be broadly intro-

duced to existing construction sites. The system is vali-

dated by a scaled physical experiment in our laboratory.

We compare the RAS with the traditional method on a

basis of operation space and operation time. In Sect. 2, we

describe the system architecture of the RAS. Details of the

assembly method are illustrated in Sect. 3, Sect. 4, and

Sect. 5. In Sect. 6, we introduce the scaled physical

experiment for validation. The experimental results are

shown and discussed in Sect. 7. Finally, we discuss the

limitation and conclusion of the study in Sect. 8.

2 Robotic assembly system architecture

The system was designed by observing and reproducing the

current steel structure erection and assembly process, as

shown in Fig. 2. We utilize a rotation method to rotate the

rigging beam to the right angle. A vertical and horizontal

alignment method is developed to align the beam; a bolting

method is developed to attach the beam; and an unloading

method is developed to unload the crane cable. We will

discuss all the methods in the following section.

Two workers are required for the RAS, one is the ground

operator and the other is the tower crane operator. The

detailed procedure of steel beam erection and assembly

using the RAS is illustrated in Fig. 3, with a comparison

of current process. The RAS consists of four key methods:

the rotation method, the vertical and horizontal alignment

method, the bolting method, and the unloading method.

First, the ground operator attaches the steel beam to the

tower crane hook and prepares for the erection. The rig-

ging beam and the RAS must be adjusted such that they

are fully horizontal. Second, the tower crane operator

transports the beam to the assembly position, and aligns

roughly. Third (the vertical alignment method), the RAS

helps the operator to align the height of the beam to a

proper level, such that the beam can later be successfully

connected to the column. The ground operator has to

double check whether the beam is aligned correctly

through the camera. Fourth (the rotation method), the RAS

rotates the beam to the assembly orientation. Fifth (the

horizontal alignment method), the crane operator adjusts

the horizontal position of the beam accurately. The ground

operator has to check that all the bolts are in the correct

bolt holes through the camera. Notice that if the beam

fails to get to the right position, the RAS has to go back to

the rotation step and repeat the process. Sixth (the bolting

method), the beam is attached with bolts, and the tem-

porary connection is completed. The ground operator has

to check whether all bolts have been installed correctly.

The rough alignment step must be repeated if the RAS

failed to install any of the bolts. Seventh (the unloading

method), the RAS unloads the beam-hook connecting

cable. Finally, the tower crane operator removes the RAS

and repositions for the next beam. We will provide

detailed descriptions of the four methods listed above in

the following sections.

3 Rotation method

We employ the principle of conservation of angular

momentum to realize the rotation method. A rotation box

with a flywheel is installed on top of the rigging beam to

generate angular momentum and the beam generates an

inverse angular momentum. Figure 4 shows a side per-

spective of the rotation box. The flywheel is rotated by a

motor through an axle and gears. A motor controller and a

wireless router are used to control the flywheel by the

ground operator. After the beam has arrived at the proper

position, the operator turns on the flywheel until the beam

rotates to the correct angle. The rotation box is connected

by two pairs of connecter bracket that clip to the beam

during the process. The camera on the rotation box is used

to realize the alignment method.

Figure 5 shows the mathematical model of the rotation

method. Since the friction of the bearing between the crane

hook and the block can be minimized, we simply assume


123

the friction is zero. We also neglect the effect of the wind

given the massive weight of the rigging beam. The angular

velocity of the beam given by the conservation of angular

momentum equation is

xb ¼Iw

Ibxw ð1Þ

where Iw, Ib represent the moment of inertia of the flywheel

and the rigged beam, and xw, xb represent the angular

velocity of the flywheel and the rigged beam.

The angular velocity of the flywheel is provided by a

motor inside the rotation box, the maximum revolution per

Fig. 3 The procedure of beam

erection and assembly with:

a current process, and b with the

RAS

Fig. 4 The side perspective view of the rotation box

Fig. 5 The mathematical model of the rotation method (side view)


123

minute of which is xm. The angular velocity of the fly-

wheel can be split into two periods. First is the accelerating

period and second is the constant velocity period. In the

accelerating period, the angular velocity is xw ¼ ata,where ta is the accelerating time to reach the maximum

revolution per minute xm. In the constant velocity period,

we assume the angular velocity of the motor always

reaches the maximum revolution per unit time. Therefore,

the angular velocity of the rigged beam can be derived

from (1) and the angular velocity of the flywheel, as shown

in Fig. 6.

In order to select a proper motor for the rotating system,

we have to calculate the maximum power Pmax of the

motor. The angular acceleration a is given from the motor

s ¼ Iwa ð2Þ

where s is the torque of the motor. We can then calculate

the power P by (1) and (2)

P ¼ sxm ¼ Iwað Þxm ¼ Iwxw

taxm ð3Þ

Knowing that when xw ¼ xm, P is the maximum value

P ¼ Pmax ¼ Iwx2

m

tað4Þ

From (4) and Fig. 6, we find that the maximum revo-

lution per minute of the motor and the accelerating time

influence the rotation time of the rigging beam and the type

of the motor we must select.

4 Alignment method

The alignment method consists of the vertical alignment

and the horizontal alignment. The objective of the vertical

alignment is to check whether the rigging beam reaches the

right height. We use a camera to detect the marker on the

column and inform the crane operator whether the beam

reaches the right height by a transmission signal to the

control cabin. Figure 7 illustrated the vertical alignment

method. If the marker lies at the center of the camera

frame, the vertical alignment is completed. In order to

attach the beam, the vertical alignment position of the

rigging beam must be slightly higher than the bracket.

Length d represents the distance from the center of the

camera lens to the beam top surface. Length d is the ver-

tical distance between the beam and the bracket, which is

also the distance from the center of the guide hole to the

bolt hole, as shown in Fig. 8. Therefore, the centroid of the

marker is d þ d higher than the bracket. Length L is the

distance from central of the camera lens to the column,

which we will use to determine the marker size.

The camera captures the image and searches for the

marker, as shown in Fig. 9. If the beam reaches the right

level, the marker can be found on the image and the

operator will be informed. The marker size is influenced by

the erection swag. Figure 10 shows the mathematical

model of the influence on the marker size due to the

erection swag. The marker length D is

D ¼ 2L tan h ð5Þ

where L represents the distance between the camera and the

column and h represents the pendulum angle. The pendu-

lum equation, according to Kuo and Kang (2014), is

d2hdt2

¼ a

lcos h� g

lsin h ð6Þ

where a represents the crane operation acceleration, l

represents the crane cable length and g represents the

gravity. The marker width B is

B ¼ 2l sin h ð7Þ

Therefore, we can determine the marker length D and

width B with (5), (6) and (7). The marker size and location

needs to be set on a correct column location in the manu-

facturing factory before delivering to construction site. A

minor adjustment based on the environmental condition is

also required on-site before starting the erection and

assembly process.

The camera has to stay at the right orientation during the

rough positioning and vertical alignment, in other words,

facing the column; we use a gyro sensor and a motor to

control the orientation of the camera. Before the vertical

alignment step starts, the motor will rotate the camera in

the direction of the column.

The objective of the horizontal alignment is to adjust the

rigging beam to the assigned position. Since the beam has

been aligned at the correct height during vertical align-

ment, the horizontal alignment will only consider planar

positioning. We change the shape of the flange plates to

parallelograms so that the beam will not get stuck during

Fig. 6 Schematic diagram of angular velocity of the rigging beam as

a function of rotating time


123

the rotation process. In addition, this shape allows the beam

to be easily controlled by the tower crane in case the beam

is not at the right position. Figure 11 shows the horizontal

alignment method. The bolting steel plates are used to

validate the accuracy of the alignment. The operator has to

check whether all bolts are positioned in the guide holes.

5 Bolting and unloading method

For providing a faster bolting process, we use a ‘‘plug and

play’’ method instead of a traditional ‘‘tightening bolts’’

method. Figure 8 shows the front view of the bolting steel

plate. We add two additional guide holes through the bolt

holes because only two bolts are needed for temporary

connection. The bolting steel plates are attached to the

bracket and the rigging beam before erection. After fin-

ishing the horizontal alignment, the bolts have been posi-

tioned in the guide holes. The crane operator then releases

the rigged beam and the bolts slide into the bolt holes, as

shown in Fig. 12, completing the temporary bolt attach-

ment step. We have designed a new nut for this method.

The nut has two parts: the sliding part and the attachment

part. The sliding part is used to connect the bolting steel

plate to the beam and slide down through the guide hole to

the bolt hole. These newly designed nuts will be assembled

and welded before the erection process in order to prevent

detachment at the assembly elevation. Then the beam will

be assembled by the attachment part to fully achieve the

temporary connection.

The unloading method is used to remove the RAS and

unload the crane cable. The rotation box is connected to the

crane hook by the cable before the erection process. We

also utilized a simple gripping mechanism to mount the

rotation box on the rigged beam. Therefore, the RAS can

simply be removed by the tower crane during the unloading

step. The cable connecting the rigging beam and the beam

hook also needs to be unloaded during this step. We use a

pin mechanism, cable, and motor to realize the unloading

operation. The cable attaches to the pin bar and the motor.

Fig. 7 The vertical alignment

method (partial side view)

Fig. 8 The bolting steel plate

(front view)

Fig. 9 The camera captures the image and searches for the marker


123

After the temporary bolting assembly step is completed, the

motors start to roll the cable and extract the pin bar from

the pin hole. Then, the cable will release and unload the

rigged beam. Finally, the tower crane will remove the RAS

from the attached beam and reposition for the next target.

6 Scaled physical experiment

For validating the RAS, we implemented a scaled physical

experiment, which includes a tower crane and a steel

structure. We used KUKA KR 16 CR (KUKA 2005) to

simulate the tower crane, as shown in Fig. 13a. The KUKA

is a six degree of freedom industrial robot arm which

connects with the cable and the hook on the end effector.

We used block board and steel bracket to build the steel

structure. The steel structure model is an experimental

structure from the National Center for Research on Earth-

quake Engineering (Lin et al. 2013), which contains one

beam and two columns, and was scaled with length ratio

a ¼ 0:4. Figure 13b shows the scaled steel structure. The

bolting steel plates with guide and bolt holes were

manufactured by OMAX 2652 JetMachining� Center

(OMAX 2016).

The rotation box was also implemented in the scaled

physical experiment. Figure 13c shows the scaled rotation

box. We used plywood to fabricate the outer covering. The

flywheel, the motor, the controller and the connector were

demonstrated by the TETRIX� (2014) and LEGO�

Mindstorms NXT (2014). The TETRIX� is a robotic

toolkit which contains metal members, motors, controllers

and batteries. We used the metal member to fabricate the

connector and the flywheel, which was connected to the

motor, as well as the motor controller and the battery. The

LEGO Mindstorms NXT was utilized as a process unit. We

connected to the LEGO Mindstorms NXT through Blue-

tooth to control the motor revolution velocity. LabVIEW

(2014) was used to program the controller software. For the

vertical alignment, we used the GHI. Net Gadgeteer kit

(2014) and green paint as marker. The Gadgeteer kit con-

tains a mainboard, a camera module, and a multicolor LED

module. We used the camera module to capture the column

image. When the camera detects the green paint, the

multicolor LED module will start to flash and inform the

Fig. 10 The mathematical model of the influence on marker size of the erection swag: a left side view and b back view

Fig. 11 The horizontal

alignment method: a top view

and b side view


123

operator that the beam has reached the right height. We

utilized green paint since the camera is most sensitive to

green color (Brown 2004). The detailed specifications of

the scaled experimental scenario are listed in Table 1.

7 Scaled experiment result

The results of the scaled physical experiment are illustrated

in the following section. We discuss the comparison

between the traditional method and the proposed method

based on two factors: the operation space and the operation

time.

7.1 Result

The procedure of the experiment follows the process of

beam erection and assembly with the RAS, as shown in

Fig. 3. First, the ground operator prepares for the beam

erection and assembly process, as shown in Fig. 14a. The

rotation box must be set up and installed on the top of the

rigging beam. The ground operator also checks the hori-

zontal status of the beam before transporting it, to ensure

that all bolts can be positioned in the guide holes later, as

shown in Fig. 14b. Second, the crane operator transports

the beam to the assembly position, as shown in Fig. 14c.

Third, the beam is roughly aligned above the assembly

position, as shown in Fig. 14d.

Fourth, the crane operator adjusts the altitude of the

beam using the vertical alignment method. The camera on

the rotation box is rotated to the column orientation and

captures the image, as shown in Fig. 15a. We use image

processing to detect the color at the center of the image. If

the camera detects the green color, the LED light will start

to flash and inform the crane operator, as shown in

Fig. 15b.

Fig. 12 The concept of the

bolting method: a, c Before

releasing and b, d after

releasing


123

Fifth, the ground operator starts the rotation box motor

and rotates the beam to the assembly angle, as shown in

Fig. 16. The rotating beam is stopped when the steel

bolting plates reaches the proper assembly position. Sixth,

the crane operator adjusts the horizontal position of the

beam with the horizontal alignment method. The ground

operator has to check through the camera that the beam is

at the right position and that all bolts are in the guide holes,

as shown in Fig. 17.

The rotation time of the proposed system is illustrated in

Table 2. The Beam I is the real size of the experiment

structure (Lin et al. 2013) and the Beam II is a steel beam

from a real steel structure. We use a motor with 1500 rpm

and a 1500 kg m2 flywheel for the real rotation box. The

accelerating time is 10 s. The angle of the rotation is set at

90 degrees. The resulting rotation time and motor power,

calculated by (4) and Fig. 6, are listed in Table 2.

Seventh, the crane operator releases the beam and lets

the bolts slide into the bolt holes, as shown in Fig. 18. The

ground operator must check that all the bolts are in the bolt

holes and completely fastened. Eighth, the RAS unloads

the pin mechanism of the beam-cable connection and the

rotation box. Then the rotation box is repositioned by the

tower crane and prepares for the next beam attachment

process, as shown in Fig. 19.

To validate the RAS, we compared the RAS with the

traditional method. Table 3 shows the operation time

results. The traditional method is tested on a thirty-floor

Fig. 13 The scaled physical experiment: a the scaled tower crane, b the scaled steel structure, and c the scaled rotation box

Table 1 The detail

specification of the scaled

physical experiment

Items Components Specification

Construction equipment Tower crane KUKA KR 16 CR robot arm

Steel structure Steel beam

Steel column

H 290 9 160 9 15 9 15 9 1720 (mm)

BOX 280 9 280 9 15 9 1000 (mm)

Rotation box Outer covering 400 9 160 9 220 (mm)

Flywheel Cuboid 290 9 225 9 30 (mm)

Motor TETRIX� DC Gear Motor

12-Volt, 152 rpm and 300 oz-in

Motor controller TETRIX� DC Motor Controller

Process unit LEGO Mindstorms NXT

Vertical alignment GHI.Net Gadgeteer

72 MHz. 32-bit ARM Mainboard

320 9 240 20FPS Camera module

Multicolor LED Module

Battery TETRIX� 12-Volt NiMH Battery

Software Control program NI LabVIEW 2010


123

steel reinforced concrete construction site located in Tai-

wan. The steel beam size is illustrated in Table 2 Beam II.

We recorded the steel beam erection and assembly process,

counted the operation time for ten times, and calculated the

average operation time. We performed two different tasks.

The first was the low-level operation, which took place at

the level of the 3rd floor, and the second was the high-level

operation, which took place at the level of the 20th floor. In

the RAS low-level task, the ground operator can directly

monitor the whole operation and inform the crane operator

when the alignment and bolting are completed. Conversely,

for the high-level task, the ground operator can only

monitor the whole operation through a camera. In addition,

the rotation time is affected by the scale ratio. According to

Kuo and Kang (2014),

a ¼ c2 ð8Þ

where a represents length ratio and c represents time ratio.

Thus, the time ratio c ¼ 0:63.

The results show that the traditional method took 501 s

to position and attach one steel beam at the low-level, while

the RAS took 55 s, which amounts to a reduction in oper-

ation time by about 89%. The rotation method of the RAS

took 19 s, which are similar to the calculation result from

Table 2 applied with time ratio c. The alignment and bolt-

ing operation shows a significant improvement with the

Fig. 14 The procedure: a preparing for the beam erection and assembly, b checking the horizontal status, c transporting, and d rough alignment


123

assistance of camera alignment and since the RAS only

needed to release the beam for bolting. The unloading

method is also reduced to a simple unplugging process in

comparison with traditional loosing bolt process. At the

high-level, the traditional method took 514 s to position and

attach one steel beam and the RAS took 69 s, reducing the

operation time by about 86%. The alignment method took

almost double the time to achieve since the ground operator

can only utilize the camera to check the accomplishment.

7.2 Discussion

We discuss the comparison between the traditional method

and the proposed system considering two main factors: the

operation space and the operation time. The operation

space is the space for operating the process; the operation

time is the time for operating the process.

7.2.1 Operation space

The operation space is the space size for operating the

process, including the rigging path, the rotation and

alignment area and the steel workers working area. For the

rigging path, the traditional method and the proposed

system are almost the same. They simply have to transport

the beam to the assembly position. For the rotation and

alignment operation, the proposed method is significantly

shorter than the traditional method. The alignment method

in the proposed system reduces the unnecessarily manual

alignment process. For the steel workers working area, the

traditional method needs a working area on the bracket and

the beam for steel workers. The proposed system does not

need the steel workers working area. Therefore, the pro-

posed system needs less operation space than the tradi-

tional method.

Fig. 15 The vertical alignment: a adjusting the altitude of the beam and b achieving the vertical alignment

Fig. 16 Rotating the beam


123

7.2.2 Operation time

The operation time is the time for operating the process.

The operation time of the proposed system is listed in

Table 2 and Table 3. In the rotation step, the traditional

method uses manual drag to rotate the beam, which is

time-consuming and requires more human workers. In the

alignment step, the traditional method relies on human

workers to align. In the bolting step, the traditional

method needs much more time than the proposed system

since in this system we utilize the plug and play method

instead of the tightening of bolts method. In the unloading

step, the pin mechanism can accelerate the unloading

process.

7.3 Limitation

The limitations of the RAS are listed in this section. First,

we have assumed that the rigging beam can remain fully

horizontal at all times; however, the beam might be not

horizontal during the process and this would cause the RAS

fail. We will design a horizontal mechanism to address this

issue in future work. Second, the RAS only operates the

process until the temporary connection is completed. The

full connection of the beam still requires human workers to

finish. Third, when the rotation method is operated manu-

ally, the overshoot issue will happen and cause structural

damage; thus, a suitable rotation controlling method needs

to be implemented in the future. In addition, utilizing a

Fig. 17 The horizontal alignment: a checking that the beam is at the right position and b checking that all the bolts are in the guide holes

Table 2 The comparison of the

rotation timeBeam I Beam II Scaled Beam

Beam section (mm) H 800 9 400 9 22 9 32 H 800 9 300 9 14 9 26 H 290 9 160 9 15 9 15

L (mm) 5000 10,000 1720

Weight (kg) 1630 2099 5

Ib (kg m2) 3396 17,492 1.989

xm (rpm) 1500 1500 152

Iw (kg m2) 2 2 8.982 9 10-3

xb (�/s) 5.298 1.032 7.717

ta (s) 10 10 7. 9 10-2

Rotation time (s) 21.99 92.21 11.74

Power (W) 4934.80 4934.80 33.72


123

dual flywheel system instead of the single flywheel system

could also maintain the rigging beam and reduce the

swaying effect. Fourth, the marker detection for the vertical

alignment can be influenced by outdoor light conditions

causing the alignment to be unstable. In our method, we

used color detection for tracking the marker, instead a

suitable marker tracking algorithm could be applied in the

RAS, such as AprilTag (Olson 2011) or KEG tracker (Feng

and Kamat 2013). These trackers can work well under

outdoor light conditions. Fifth, the vertical alignment is

operated by the crane operator, who relies on signal feed-

back from the rotation box. The delay of the signal might

cause the vertical alignment to be insufficient or overshoot.

Therefore, the motion controller combined with the marker

tracking algorithm for vertical alignment needs to be fur-

ther developed to address this issue. Sixth, the scaled

experiment result is based on the indoor environment. We

neglected some outdoor environmental factors such as

wind and weather issues.

With the aspect from two experts in academic structural

engineering and one in construction industry, the special

cutting shape needs further verification for welding speci-

fication. In addition, it is not cost-effectiveness for material

using. Therefore, in the future work, we need to develop a

robust beam orientation controlling system for assembling

a regular shape beam.

8 Conclusion

We developed a robotic assembly system (RAS) for steel

structures. The rotation method, the alignment method, the

bolting method, and the unloading method are the four

main operations performed by the RAS. The rotation

method utilizes a flywheel and the conservation of angular

momentum to rotate the rigging element. The alignment

method utilizes a camera and a marker on the column to

ensure the altitude of the beam is correct. By using a

parallelogram flange plate, the beam can be easily aligned.

The bolting method uses a plug and play method. We add

an additional guide hole above the bolt hole on the steel

bolting plate; therefore, the bolt can plug into the guide

Fig. 18 Assembling the bolts


123

Fig. 19 Unloading and repositioning

Table 3 Comparison of the traditional method and the RAS operation time

Operation time (s) Task 1 Task 2

Low level (3rd floor) High level (20th floor)

Traditional (T) RAS (R) Reduction rate (%) Traditional (T) RAS (R) Reduction rate

Rotation 76 19 75.00 82 20 75.61

Alignment 201 27 86.57 197 41 79.19

Bolting 191 1 99.48 195 1 99.49

Unloading 33 8 75.76 40 7 82.50

Total 501 55 89.02 514 69 86.58


123

hole and slide to the bolt hole. The unloading method is a

pin mechanism and can be easily unloaded. A scaled

physical experiment was implemented to verify the feasi-

bility of the system. We found that the RAS can operate the

steel beam placement and connection process without steel

workers having to be in high places. In addition, the RAS

needs less operation space and can finish the process faster.

To sum up this research, the system described is intended

to replace the human workers in high-rise building con-

struction. This could greatly reduce accidental falls as well

as improve the efficiency of the steel structure assembly

process.

References

Beavers, J.E., Moore, J.R., Schriver, W.R.: Steel erection fatalities in

the construction industry. J. Constr. Eng. Manag. 135(3),227–234 (2009)

Bijlaard, F.S.K., Coelho, A.M.G.O., MagalhA£es, V.J.D.A.: Innova-

tive joints in steel construction. Steel Constr. 2(4), 243–247

(2009)

Bock, T., Linner, T.: Construction robots: elementary technologies

and single-task construction robots. Cambridge University Press

(2016a)

Bock, T., Linner, T.: Site automation: automated/robotic on-site

factories. Cambridge University Press (2016b)

Brown, M.: Advanced digital photography. Media Publishing Pvt,

Limited, Australia (2004)

CM Lab: Vortex Master. http://www.cm-labs.com/market/construc

tion/products/vortex-master (2015). Accessed 10 Apr 2017

Chang, Y.-C., Hung, W.-H., Kang, S.-C.: A fast path planning method

for single and dual crane erections. Autom. Constr. 22, 468–480(2012)

Cheng, T., Teizer, J.: Modeling tower crane operator visibility to

minimize the risk of limited situational awareness. J. Comput.

Civ. Eng. 28(3), 04014004 (2014)

Chi, H.-L., Chen, Y.-C., Kang, S.-C., Hsieh, S.-H.: Development of

user interface for tele-operated cranes. Adv. Eng. Inf. 26(3),641–652 (2012)

Chi, H.-L., Kang, S.-C.: A physics-based simulation approach for

cooperative erection activities. Autom. Constr. 19(6), 750–761(2010)

Chin, S., Yoon, S., Kim, Y.-S., Ryu, J., Choi, C., Cho, C.-Y.:

Realtime 4D CAD ? RFID for project progress management.

Construction Research Congress (CRC): broadening perspec-

tives, San Diego, California. 331–340 (2005)

Chu, B., Jung, K., Lim, M.-T., Hong, D.: Robot-based construction

automation: an application to steel beam assembly (Part I).

Autom. Constr. 32, 46–61 (2013)

ConXtech: ConX system. http://www.conxtech.com/conx-system/

(2015). Accessed 10 Apr 2017

Feng, C., Kamat, V.R.: Plane registration leveraged by global

constraints for context-aware AEC application. Aided Civ.

Infrastruct. Eng. 28, 325–343 (2013)

Feng, C., Xiao, Y., Willette, A., McGee, W., Kamat, V.R.: Vision

guided autonomous robotic assembly and as-built scanning on

unstructured construction sites. Autom. Constr. 59, 128–138

(2015)

GHI Electronics LLC:.NET Gadgeteer. https://www.ghielectronics.

com/catalog/category/265 (2014). Accessed 10 Apr 2017

Gajamohan, M., Merz, M., Thommen, I., D’Andrea, R.: The cubli: a

cube that can jump up and balance. International Conference on

Intelligent Robots and Systems, Vilamoura, Algarve, Portugal.

pp. 3722–3727 (2012)

Gambao, E., Balaguer, C., Gebhart, F.: Robot assembly system for

computer-integrated construction. Autom. Constr. 9(5–6),479–487 (2000)

Gams, A., Zlajpah, L., Lenarcic, J.: Imitating human acceleration of a

gyroscopic device. Robotica 25(04), 501–509 (2007)

Garg, A., Kamat, V.R.: Virtual prototyping for robotic fabrication of

rebar cages in manufactured concrete construction. J. Archit.

Eng. 20(1), 06013002 (2014)

Huang, J.-Y., Gau, C.-Y.: Modelling and designing a low-cost high-

fidelity mobile crane simulator. Int. J. Hum.-Comput. St. 58(2),151–176 (2003)

Hung, W.-H., Liu, C.-W., Liang, C.-J., Kang, S.-C.: Strategies to

accelerate the computation of erection paths for construction

cranes. Autom. Constr. 62, 1–13 (2016)

Irizarry, J.: Application of methods analysis to steel erection tasks:

identification of factors affecting labor performance. Int. J. Con-

str. Manag. 11(4), 31–47 (2011)

Irizarry, J., Abraham, D.M.: Assessment of risk perception of

ironworkers. J. Constr. Res. 7(1/2), 111–132 (2006)

Juang, J.R., Hung, W.H., Kang, S.C.: SimCrane 3D?: a crane

simulator with kinesthetic and stereoscopic vision. Adv. Eng.

Inf. 27(4), 506–518 (2013)

Jung, K., Chu, B., Hong, D.: Robot-based construction automation: an

application to steel beam assembly (Part II). Autom. Constr. 32,62–79 (2013a)

Jung, K., Chu, B., Park, S., Hong, D.: An implementation of a

teleoperation system for robotic beam assembly in construction.

Int. J. Precis. Eng. Manuf. 14(3), 351–358 (2013b)

Jung, K., Chu, B., Bae, K., Lee, Y., Hong, D., Park, S., Lim, M.-T.:

Development of automation system for steel construction based

on robotic crane. International Conference on Smart Manufac-

turing Application, Kintex, Gyeonggi-do, Korea. 486–489

(2008)

KUKA: KR 16 CR v2 technical data, KUKA Robotics Corporation

(2005)

Kahane, B., Rosenfeld, Y.: Real-time ‘‘sense-and-act’’ operation for

construction robots. Autom. Constr. 13(6), 751–764 (2004)

Kang, S.-C., Chi, H.-L., Miranda, E.: Three-dimensional simulation

and visualization of crane assisted construction erection pro-

cesses. J. Comput. Civ. Eng. 23(6), 363–371 (2009)

Kang, S.C., Miranda, E.: Planning and visualization for automated

robotic crane erection processes in construction. Autom. Constr.

15(4), 398–414 (2006)

Kang, S.-C., Miranda, E.: Numerical methods to simulate and

visualize detailed crane activities. Comput. Aided Civ Inf. 24(3),169–185 (2009)

Kim, D.-W., An, S.-H., Cho, H., Jeong, J.-W., Lee, B.-H., Dohe, N.L.,

Kang, K.-I.: Development of conceptual model of construction

factory for automated construction. Build. Environ. 44(8),1634–1642 (2009)

Kim, K., Kim, M.: RFID-based location-sensing system for safety

management. Pers. Ubiquit. Comput. 16(3), 235–243 (2012)

Kim, C.-W., Kim, T., Lee, U.-K., Cho, H., Kang, K.-I.: Advanced

steel beam assembly approach for improving safety of structural

steel workers. J. Constr. Eng. Manage. 142(4), 05015019 (2016)

Kuo, T.-Y., Kang, S.-C.: Control of fast crane operation. Autom.

Constr. 42, 25–35 (2014)

LEGO: LEGO Mindstroms. http://www.lego.com/en-us/mindstorms/

(2014). Accessed 10 Apr 2017

Lee, G., Cho, J., Ham, S., Lee, T., Lee, G., Yun, S.-H., Yang, H.-J.: A

BIM- and sensor-based tower crane navigation system for blind

lifts. Autom. Constr. 26, 1–10 (2012a)


123

http://www.cm-labs.com/market/construction/products/vortex-master

http://www.cm-labs.com/market/construction/products/vortex-master

http://www.conxtech.com/conx-system/

https://www.ghielectronics.com/catalog/category/265

https://www.ghielectronics.com/catalog/category/265

http://www.lego.com/en-us/mindstorms/

Lee, C., Lee, G.: Feasibility of beam erection with a motorized hook-

block,’’. Autom. Constr. 41, 25–32 (2014)

Lee, G., Lee, G., Lee, T.: Field applicability and feasibility of the

automated non-powered multi-beam lifting system. Autom.

Constr. 28, 26–35 (2012d)

Lee, H.-S., Lee, K.-P., Park, M., Baek, Y., Lee, S.: RFID-based real-

time locating system for construction safety management.

J. Comput. Civ. Eng. 26(3), 366–377 (2012b)

Lee, C., Lee, G., Park, S., Cho, J.: Analysis of field applicability of

the rotation-controllable tower-crane hook block. Autom. Con-

str. 21, 81–88 (2012c)

Lei, Z., Taghaddos, H., Hermann, U., Al-Hussein, M.: A methodology

for mobile crane lift path checking in heavy industrial projects.

Autom. Constr. 31, 41–53 (2013)

Liang, C.-J., Kang, S.-C.: Development of a steel beam hauling

system for automatic steel beam assembly. International Con-

ference on Computing in Civil and Building Engineering

(ICCCBE), Orlando, Florida. 1295–1302 (2014)

Lin, K.-C., Jhuang, S.-J., Lin, C.-H.: Study on seismic performance of

steel shop-fabricated moment connections, National Center for

Research on Earthquake Engineering (2013)

Mo, Y.-H., Kang, T.-K., Zhang, H.-Z., Hong, D.-H., Lim, M.-T.:

Development of 3D camera-based robust bolt-hole detection

system for bolting cabin. Autom. Constr. 44, 1–11 (2014)

Nam, H., Choi, W., Ryu, D., Lee, Y., Lee, S.-H., Ryu, B.: Design of a

bolting robot for constructing steel structure. International

Conference on Control, Automation and Systems, COEX, Seoul,

Korea. 1946–1949 (2007)

National Instruments Corporation: LabVIEW system design software.

http://www.ni.com/labview/ (2014). Accessed 10 Apr 2017

OMAX Corporation: OMAX 2652 JetMachining� Center. https://

www.omax.com/omax-machine/2652 (2016). Accessed 10 Apr

2017

Olson, E.: AprilTag: a robust and flexible visual fiducial system. IEEE

International Conference on Robotics and Automation (ICRA),

Shanghai, China. (2011)

Park, M.-W., Brilakis, I.: Construction worker detection in video

frames for initializing vision trackers. Autom. Constr. 28, 15–25(2012)

Pavlovcic, L., Krajnc, A., Beg, D.: Cost function analysis in the

structural optimization of steel frames. Struct. Multidiscip. O.

28(4), 286–295 (2004)

Peurifoy, R. L., Schexnayder, C. J., Shapira, A., Schmitt, R. L.:

Construction planning, equipment, and methods, McGraw-Hill,

New York (2011)

Pitsco Education: TETRIX�. http://www.tetrixrobotics.com/ (2014).

Accessed 10 Apr 2017

Ray, S.J., Teizer, J.: Coarse head pose estimation of construction

equipment operators to formulate dynamic blind spots. Adv.

Eng. Inf. 26(1), 117–130 (2012)

Saidi, K.S., Bunch, R., Lytle, A. M., Proctor F.: Development of a real-

time control system architecture for automated steel construction.

23th International Symposium on Automation and Robotics in

Construction (ISARC), Tokyo, Japan, 412–417 (2006)

Teizer, J., Cheng, T., Fang, Y.: Location tracking and data visual-

ization technology to advance construction ironworkers’ educa-

tion and training in safety and productivity. Autom. Constr. 35,53–68 (2013)

The Steel Construction Institute: Quicon� design guide to BS 5950-1.

http://www.newsteelconstruction.com/wp/quicon-design-guide-

to-bs-5950-1/ (2004). Accessed 10 Apr 2017

Vijay, V., Sen, R.N., Richardson, S.: Risk perception of structural

steel erection workers in Malaysia—Ergonomic implications for

occupational safety and health. International Symposium on

Past, Present and Future Ergonomics, Occupational Safety and

Health, Bali, Indonesia. pp. 100–105 (2006)

Viscomi, V.B., Michalerya, W.D., Lu, L.-W.: Automated construction

in the ATLSS integrated building systems. Autom. Constr. 3(1),35–43 (1994)

Volk, R., Stengel, J., Schultmann, F.: Building Information Modeling

(BIM) for existing buildings—literature review and future needs.

Autom. Constr. 38, 109–127 (2014)

Wakisaka, T., Furuya, N., Inoue, Y., Shiokawa, T.: Automated

construction system for high-rise reinforced concrete buildings.

Autom. Constr. 9(3), 229–250 (2000)

Wang, X., Zhang, Y.Y., Wu, D., Gao, S.D.: Collision-free path

planning for mobile cranes based on ant colony algorithm. Key

Eng. Mat. 467–469, 1108–1115 (2011)

Yoo, W.S., Lee, H.-J., Kim, D.-I., Kang, K.-I., Cho, H.: Genetic

algorithm-based steel erection planning model for a construction

automation system. Autom. Constr. 24, 30–39 (2012)

Yu, S.-N., Ryu, B.-G., Lim, S.-J., Kim, C.-J., Kang, M.-K., Han, C.-

S.: Feasibility verification of brick-laying robot using manipu-

lation trajectory and the laying pattern optimization. Autom.

Constr. 18(5), 644–655 (2009)

Zhang, C., Hammad, A.: Improving lifting motion planning and re-

planning of cranes with consideration for safety and efficiency.

Adv. Eng. Inf. 26(2), 396–410 (2012)

Zhou, W., Whyte, J., Sacks, R.: Construction safety and digital

design: a review. Autom. Constr. 22, 102–111 (2012)

Ci-Jyun Liang is currently a

graduate student in Robotics

Institute at University of

Michigan, Ann Arbor. He is

working with Prof. Vineet

Kamat in the Laboratory for

Interactive Visualization in

Engineering (LIVE). His

research interests include

autonomous construction site

robot, computer vision, machine

learning, and virtual/augmented

reality. He also received a M.S.

and B.S. in Civil Engineering

from National Taiwan

University.

Prof. Shih-Chung Kang,Ph.D. of Stanford University, is

currently a professor in depart-

ment of civil engineering in

National Taiwan University. His

research focuses on the

advanced visual and robotics

tools for engineering purposes.

He developed multiple methods

on the automation and simula-

tion for crane operations. He is

now the Editor-in-Chief of

Visualization in Engineering.

He edited multiple special

issues on the topics related to

the visualization and robotics applications. He is also an active

researcher on innovative engineering education. He offers courses on

engineering graphics, game development, data visualization and

robotics and was awarded NTU excellent teacher. His course on

engineering graphic was ranked top 2% MOOCs among Chinese

learners.


123

http://www.ni.com/labview/

https://www.omax.com/omax-machine/2652

https://www.omax.com/omax-machine/2652

http://www.tetrixrobotics.com/

http://www.newsteelconstruction.com/wp/quicon-design-guide-to-bs-5950-1/

http://www.newsteelconstruction.com/wp/quicon-design-guide-to-bs-5950-1/

Dr. Meng-Hsueh Lee is a Post

Doc researcher in Center for

Weather Climate and Disaster

Research at National Taiwan

University. He received his

Ph.D. and M.S. from the

Department of Civil Engineer-

ing at National Taiwan Univer-

sity in 2009 and 2004, and his

B.S. in Department of Civil and

Construction Engineering at

National Taiwan University of

Science and Technology in

2002. His research interests

include construction knowledge

management map, construction enterprise resource planning, seman-

tic analysis, and data mining.


123

ras: a robotic assembly system for steel structure ... · pdf fileregular paper ras: a robotic...

Documents