Journal of Intelligent & Robotic Systems (2021) 102: 75

https://doi.org/10.1007/s10846-020-01266-1

REGULAR PAPER

A Climbing Robot for Steel Bridge Inspection

Son Thanh Nguyen1 · Hung Manh La2

Received: 8 January 2020 / Accepted: 2 October 2020 / Published online: 5 July 2021
© Springer Nature B.V. 2020

Abstract
As an effort of automating the bridge inspection process, this paper presents a new development of an adaptable tank-like
robot, which can climb on steel structures to collect data and perform inspection. While most current steel climbing mobile
robots are designed to work on flat steel surface, our proposed tank-like robot design is capable of climbing on different steel
structural shapes (e.g., cylinder, cube) by using reciprocating mechanism and magnetic roller-chains. The developed robot
can pass through the joints and transition from one surface to the other (e.g., from flat to curving surfaces). A prototype
robot integrating multiple sensors (hall-effects, IR, IMU, Eddy current and cameras), has been developed by coping with
variety of strict concerns including tight dimension, effective adhesive and climbing adaptation. Rigorous analysis of robot
kinematics, adhesive force, sliding and turn-over failure and motor power has been conducted to certify the stability of the
proposed design. The theory calculations can serve as an useful framework for designing future steel climbing robots. The
cameras and Eddy current sensor is integrated on Robot for visual and in-depth fatigue crack inspection of steel structures.
Experimental results and field deployments on more than twenty steel bridges confirm the adhesive, climbing, inspection
capability of the robot. Video of this deployment can be seen in this link: https://youtu.be/1Wl9Trd3EoM.

Keywords Field robotics · Climbing robots · Magnetic wheel robots · Steel bridge inspection · Structure health monitoring

1 Introduction Therefore, the bridges require more frequently inspection in

order to timely monitor their structural health. However, the
1.1 Motivation current inspection practice is mainly by visual (see Fig. 1)
and manual nondestructive testing/evaluation and hence can
Bridges in general and steel bridges in particular are not meet the demand for frequent and adequate inspection
vital parts of transport infrastructure. There are more and maintenance [1, 3].
than 607,380 bridges in the United States [1]. Of this The inspection on steel bridges often takes place at
total, 200,000 bridges are steel, and about 25% of these height, limited/confined space and dangerous environments
steel bridges are either deficient or functionally obsolete, or at inaccessible areas sometime (see Fig. 1). These
indicating a growing threat to the safety of transportation. works demand skillful inspectors, and it is time consuming
The collapse of several bridges recently (e.g., I-5 Skagit and also costly for training and operation. There is a
River Bridge collapse [2] in Washington State in May 2013) significant motivation for automating the bridge inspection
has shown significant impact on the safety of travelers. and evaluation through the development of climbing robots,
which can seamlessly support the inspectors to complete
their jobs in an efficient manner. Thus, the automated
 Hung Manh La inspection is an expected solution, in which robot is able
hla@unr.edu
to climb on steel bridges, carry cameras and nondestructive
Son Thanh Nguyen evaluation (NDE) sensors to collect data and update
thanhson@nevada.unr.edu infrastructure’s health. However, design and development
of a such climbing robot is challenging because it requires
1 Advanced Robotics and Automation (ARA) Lab,
that the robot is able to climb, adhere (without falling
University of Nevada, 1664 N. Virginia Street, Reno, off) and transition on complex steel structures with various
NV 89557, USA shapes. In addition, the robot should be able to perform
2 ARA Lab, University of Nevada, 1664 N. Virginia Street, data collection in the confined space. This paper presents
Reno, NV 89557, USA a new design and development of a climbing robot to
75 Page 2 of 21 J Intell Robot Syst (2021) 102: 75

Fig. 1 Current practice of steel

bridge inspection: rope certified
bridge engineers inspecting the
Golden Gate bridge, San
Francisco, California , source:
NBC April 2018. Detail
available here: https://www.
nbcbayarea.com/on-air/as-seen-
on/Inspections-Continue-at-Golden-Gate-
Bridge Bay-Area-312301361.html

tackle these challenges. The difficult maneuverability on mechanism was developed by Inuktun company for steel
complex steel structures is overcome through an integrated structure inspection using camera for visual observation
design of reciprocating mechanism and permanent magnetic [34].
roller-chains. The NDE data collection is eased by the In addition to using permanent magnet, electromagnet
development and integration of two 5 degree-of-freedom was also used to develop a climbing robot. For instance,
(DOF) robotic arms (one on the robot and one on the electromagnet was embedded on a spider robot’s legs [35]
operator’s hand) to allow the operator/user to observe and to utilize the climbing ability of the creature on complex
manage data collection. structures.
Apart from steel bridge inspection, the great effort for
1.2 Literature Review automating concrete bridge deck inspection and evaluation
has been reported in [36–46]. The leading work in the
As an effort to automate the inspection process, there has automated concrete bridge deck inspection is the Robotics
been variety of implementations of climbing robots for Assisted Bridge Inspection Tool (RABITT M ), which was
inspection [4–8]. A notable development of steel climbing developed by the the Rutgers University’s Center for
robots is summarized in Fig. 2. A roller chain, integrated Advanced Infrastructure and Transportation. RABIT was
permanent magnets robot was reported in [9]. A legged integrated with multiple NDE sensors including ground
robot that can transition across structure members for penetrating radar, electrical resistivity, impact echo, ultra
steel bridge inspection was developed [10]. The robot surface wave and high resolution cameras [37]. RABIT can
uses permanent magnets integrated with each foot to move autonomously on bridge deck to collect visual and
allow it to hang from a steel bar. In another case, NDE data by using the Extended Kalman Filter (EKF)-
a magnetic wheeled robot, which can carry magneto based localization and the artificial potential field-based
resistive sensor array for detecting corrosion and cracks, motion planning [40, 47].
was developed [11]. Similarly, several climbing permanent Last, but not least, drones are also contributed in
magnet-robots [12–20] were designed to carry non- this area. In [48], the author introduced the multicopter
destructive evaluation (NDE) devices to detect corrosion, carrying high resolution camera for bridge inspection tasks.
weld defects and cracks, and these robots can be applied The remote building inspection and monitoring by drone
for inspecting steel structures and bridges. Other efforts application mentioned in [49]. The deeper researches for
have shown for development of climbing robots for power energy optimization and visual data quality was reported in
plant inspection [21] and bridge cable inspection [22]. [50–52]. Recently, Intel Corp. also involved with the drone
Additionally, significant development of climbing robots for development for bridge inspection [53].
steel structure and bridge inspection has been reported in [4, In summary, most existing designs are developed for
23–33]. Furthermore, with industry involvement, a surface particular application with limited functions. Some provide
adapting tank-like robot with untouched permanent magnet visual inspection only, and some use untouched magnets,
J Intell Robot Syst (2021) 102: 75 Page 3 of 21 75

Fig. 2 The notable steel structure climbing robots

which cause the robot much heavier. Most existing designs real bridges that require adaptable, light and effectively data
have fixed distance between the magnet and surface, thus collecting robot. Drones still get the limitation with energy
may not work on different types of surface contours. They issue and are able to perform visual/shallow inspection
might be difficult to apply on complicated structures of only.
75 Page 4 of 21 J Intell Robot Syst (2021) 102: 75

Fig. 3 System interface: Operator controls robot remotely from manipulator holding Eddy current sensor’s probe is controlled by a
ground station via radio links. Xbee modules manage to send robot haptic device while a keypad handles robot movement. Visual and
locomotion commands and receive data from sensor. A 5-DOF sensor data are live on a PC screen

Fig. 4 Steel climbing robotic system. (Left) a simulated robotic arm both camera and Eddy current sensor for visual and non-destructive
on the ground station for remote control. (Middle) the side view of evaluation of the steel structure
the robot. (Right) the front view of the robot. The robot equipped with

Fig. 5 Roller chain with

integrated Hall sensor and IR
distance sensors. The IR sensor
is served as a feedback of the
PID control to regulate the
reciprocating mechanism
automatically on different
contour surfaces. The detail of
PID control function is described
on Section 3.1 and Fig. 13
J Intell Robot Syst (2021) 102: 75 Page 5 of 21 75

Table 1 Robot parameters permanent magnets for adhesive force creation enable
the robot to adhere to steel surfaces without consuming
Length 163 mm
any power. The control architecture of the robot consists
Width 145 mm of low-level and high-level controllers. The low-level
Height 198 mm controller handles tasks including (i) converting velocity
Weight 3 kg and heading command from the high-level controller to
Drive 2 motorized roller-chains and 1 motorized transformation Pulse Width Modulation (PWM) data to drive motors,
and (ii) reading data from multiple sensors for navigation
purposes. The high-level controller is embedded in an
This paper presents a practical climbing robotic system on board computer to enable data processing and ground
to provide an efficient solution for steel bridge inspection. station communication. Both controllers fuse sensor data to
The robot can adapt to a wide range of different provide desired linear velocity and heading for the robot and
types of bridge surfaces (flat, curving, rough) and carry acquire data from advanced sensors. Furthermore, the high-
sufficient measurement devices including cameras and level controller sends data wirelessly to ground station for
Eddy current sensor. The proposed small tank-like robot processing and logging.
with reciprocating mechanism features various deformable The robot is equipped with various sensors for navigation
3D configurations, which can allow it to transition among as well as steel structure evaluation. There are two video
steel structure members for efficient inspection. The robot cameras: one for capturing images of inspected surfaces,
utilizing adhesive force generated by permanent magnets and the other one for guiding navigation of the robotic arm.
is able to adhere well on steel structures while moving. There are two 5 DOF robotic arms: one on the robot for
The roller-chain design allows the robot to overcome navigating the Eddy current sensor probe, and the other one
obstacles including nuts, bolts, convex and concave corners. for the operator/inspector, who can observe the camera’s
To demonstrate the robot’s working principle, it has been visual feedback then control this arm for manual operation
deployed for climbing on more than 20 steel bridges. Video purposes. The robot uses an IR sensor mounted on top of
of this deployment can be seen in this link: https://ara.cse. manipulator to give the feedback to operator by trigger a
unr.edu/?page id=11. tiny vibrating motor on Haptic device. The robot has two
The rest of the paper is organized as follows. Section 2 roller-chains, and each roller-chain is integrated with two
describes the overall robot design. Section 3 describes hall-effect sensors, which provide velocity feedback for PID
detailed mechanical design and magnetic force analysis speed controller. Figure 5 illustrates the sensor’s mounting
of the robot when moving on surfaces with different for each roller chain as we use two hall-effect sensors, which
inclinations. Section 4 shows various experiments to verify are mounted next to each other and close to robot’s roller.
the proposed robot design and validate the magnetic force Since the magnet block inside each roller-chain will
analysis. Finally, the conclusion and discussion of future move when the robot moves, we can extract the velocity
work are presented in Section 5. and traveling distance of each roller-chain after combining
the data from these two hall-effect sensors. Additionally,
an Inertial Measurement Unit (IMU) is used for the robot’s
2 Overall Design localization. Moreover, to avoid falling off, the robot has IR
(Infrared Radiation) sensors mounted in the robot to detect
The overall design concept of a tank-like climbing robot if a surface underneath exists.
is shown in Fig. 3 and the implementation of this design For fatigue crack detection on steel structure, the Eddy
is shown in Fig. 4. The roller-chains embedded with current sensor’s probe (Nortec 600) as shown in Fig. 6

Table 2 Motor parameters

Moving motors Transforming motor

Torque 1.2 N.m (2S Li-Po) 3.2 N.m

Speed 0.12 sec/ 60◦ (2S Li-Po) 0.15 sec/ 60◦ (2S Li-Po)
Length 40.13 mm 60.5 mm
Width 20.83 mm 30.4 mm
Height 39.62 mm 45.6 mm
Weight 71 g 156 g
Voltage 6-8.5V (2S Li-Po battery) 6-8.5V (2S Li-Po battery)
75 Page 6 of 21 J Intell Robot Syst (2021) 102: 75

Fig. 6 a kinematic structure of 5-DOF arm; b executive arm integrating on the robot; c controlling arm for user operation in manual mode

is integrated with the robot. A mini 5-DOF robotic arm approaches the steel surface or not. When distance from IR
(Fig. 7a) is designed to hold and move the Eddy sensor’s sensor to the surface meets a calibrated number (45mm), a
probe for data collection as demonstrated in Fig. 7b. signal will be sent back to ground station to trigger a tiny
Maximum moment on joint 1 in Fig. 7a is described vibrating motor on the simulated arm to enable Eddy data
as collection. Additionally, a pressure spring is placed between
the probe and robotic arm to improve sensor’s approaching.
This design helps the sensor’s probe to efficiently collect
Marm = Parm ∗ larm , (1)
Eddy current data on complex surfaces such as weld, rough
or curving.
where Parm = 0.3kg is the total weight of the manipulator
arm, larm = 6cm is the distance from center of manipulator
mass to joint 1, so Marm = 0.18N.m. Five motors used are 3 Mechanical Design and Analysis
0.25 N.m-torque-mini servos, which satisfy (1).
On ground station, the operator observes visual feedback A tank-like robot mechanism design is proposed to take
from the camera and then controls a simulated arm (same advantage of the flexibility in maneuvering. Two motors are
scale as the one integrated with the robot) by holding used to drive two roller-chains, and another motor is used to
and moving a probe model. The moving control signal is drive the transformation of the robot to approach different
wirelessly sent via Xbee module to control the arm on the contour surfaces. The robot’s parameters are shown in
robot. Five joint’s angles are obtained by potentiometers Table 1 while the motor’s parameters are listed in Table 2.
(Fig. 7c) then processed before sending required control A roller-chain is designed to carry 22 Neodymium
angles to servo motors. An IR sensor is mounted on the magnet blocks with poles on flat ends as presented in Fig. 4-
Eddy current sensor’s probe to determine whether the probe front view. With each motion, there will be a maximum of

Fig. 7 (Left) Industrial Eddy

current sensor Nortec 600;
(Right) sensor probe for data
collection
J Intell Robot Syst (2021) 102: 75 Page 7 of 21 75

Fig. 8 Pull force over air gap of

10mm × 10mm magnet cylinder

8 magnet blocks contacting the flat steel surface. According mechanism has been added in order to transform the
to [54], if there is a gap between magnet blocks and robot to adapt with different contour surfaces as shown in
steel frame, the pull force is significantly affected. The Fig. 9. The specifications of the robot’s design is shown in
characteristics of the used magnets (15mm length ×10mm Fig. 10.
width ×5mm thickness magnet block) are described in Regarding magnetic force, denote that Fmi is magnetic
Table 3 and pull force over air gap is shown in Fig. 8. Roller- forces created by each magnet blocks as shown in Fig. 5.
chains are designed to enable the robot to overcome several At each moment there are a maximum 8 magnet blocks
real climbing scenarios including transitioning among in the chain contacting the steel surface. Hence, we have
surfaces with different inclination levels (0 − 90◦ change the
 magnetic force created by each robot’s roller-chain as:
in orientation) or getting rid of being stuck. A reciprocating Fmj (j = 1 : 8). Since the robot has two roller-chains,

Fig. 9 Reciprocating mechanism for robot transformation on different surface: flat, negative curve and positive curve
75 Page 8 of 21 J Intell Robot Syst (2021) 102: 75

Table 3 Pull Force (P F ) Over

Air Gap of Distance PullForce P F (Newton - N)
15mm × 10mm × 5mm
magnet block N35 (mm) Magnet to steel plate Between two steel plates Magnet to magnet

0 39.7 73.1 39.7

1 17.1 29.8 22
2 10.1 17 15.3
3 6.4 10.5 11.1
4 4.3 6.9 8.4
5 2.9 4.6 6.4
6 2 3.2 5
7 1.5 2.2 4
8 1.1 1.6 3.2
9 0.8 1.1 2.5
10 0.6 0.8 2.1


the total magnetic adhesive force, Fm , created by these two From Fig. 11: y = XZ = XY 2 − Y Z 2 , where Y Z =
roller-chains is: b1 − OT ⇔ Y Z = b1 − OX cos φ; (φ = α − γ );

Fm = 2 ∗ Fmj (j = 1 : 8). (2) 
→y= XY 2 − (b1 − a cos φ)2 . (5)
3.1 Robot Transformation Kinematics
From Eq. 4, (b1 = b − 27) Fig. 12 and Eq. 5 we get y =
Kinematics analysis of reciprocating mechanism is to f(x)):
calculate radius of steel cylinder (x), that the robot can 
climb on. Table 4 shows kinematic parameters while Fig. 9 cos α cos β(f tan α − x − e)
→y= XY 2 − ( − a cos φ − 27)2 .
presents kinematics of the robot, and Fig. 11 illustrates a cos α − 1
general architecture of the robot’s reciprocating mechanism. (6)
CF
From Fig. 12: x = I C − CR, in which, I C = =
cos β The feed screw mechanism has gear ratio is 1 : 19 with
f
OF − OC b− screw pitch is 0.8mm.
= cos α = b − f
; CR = From Eq. 6 and the designed gear ratio of 1:19, we can
cos β cos β cos β cos α cos β
e − BC = e − f tan α with α + β = 90◦ ; calculate radius x of the steel cylinder based on rotations
of the transformation motor. The robot is designed to work
on steel cylinders having a smallest radius from +5cm to
b b
→x= − − e + f tan α. (3) −25cm with 7.5cm feed screw movement.
cos β cos α cos β When the robot travels on different contour surfaces,
reciprocating mechanism is driven automatically by the
cos α cos β(f tan α − x − e) driving motor, which is based on the feedback from two IR
→b= . (4)
cos α − 1 distance sensors. A PID controller is applied to maintain
the distance between the two sensors and the steel surface
Table 4 Kinematics parameters equally as illustrated on Fig. 13. The controller helps keep
approaching area of permanent magnet with steel surface
b 72 mm
consistent.
b1 45 mm
f 11 mm 3.2 Sliding Failure Investigation
e 55 mm
a 33.7 mm To understand stability of the robot while climbing on steel
XY 32 mm structures, the sliding and turn-over failures as illustrated in
γ 12◦
Fig. 14a and b should be investigated.
J Intell Robot Syst (2021) 102: 75 Page 9 of 21 75

Fig. 10 Reciprocating
mechanism for robot
transformation and
specifications of the robot’s
design (unit in mm)

In general case, based on the proposed design, the robot When the robot climbs on top of an inclined surface
can climb on different shapes of structures (cylinder, cube (Fig. 15a), based on our previous work [30], we can obtain
or flat) with different inclination levels as shown in Fig. 15. P sin φ
the magnetic adhesive force: Fm = − P cos φ.
In this analysis, we focus on basic working condition- μ
flat surface to calculate the needed magnets and motor’s Hence, the sliding failure can be avoided if the magnetic
parameters, then do experiments on different surfaces force satisfies the following condition:
(curving, cube) in different conditions to find out optimal P sin φ
results. Fm > − P cos φ. (7)
μ
Let P be the robot’s total weight (P = mg, where m
is the robot’s mass, and g is the gravitational acceleration). When the robot climbs underneath an inclined surface
Let Fm be the magnetic adhesive force, N be the reaction P sin φ
(Fig. 15b), we obtain: Fm = + P cos φ. In this case,
force, μ be the frictional coefficient, Ff be frictional force, μ
and φ be the degree of inclination. Denote F as the total
force applied
 to the robot. Based on Newton’s second law
of motion, F = 0 when the robot stops.

Fig. 11 Reciprocating mechanism: a = 33.7mm, XY = 32mm, and Fig. 12 Kinematics of the robot transforming mechanism: b = 72mm,
γ = 12◦ b1 = 45mm, f = 11mm, and e = 55mm
75 Page 10 of 21 J Intell Robot Syst (2021) 102: 75

Fig. 13 Examples of different

cases when the robot moves
from flat to curved steel surface:
(Left) positive curved surface;
(Right) negative curved surface.
A PID controller will be
activated to keep the distance of
two sensors D1 and D2 equally

the magnetic force should be In summary, the robot’s magnetic adhesive force should be
greater or equal to 2.5 of the robot’s weight. The value
P sin φ
Fm > + P cos φ. (8) of one permanent magnet, where n = 8 is the maximum
μ
number of magnet blocks contacting to the steel surface,
When the robot climbs on a vertical surface (φ = 90o ) satisfies the condition (12).
P
Fm > . (9) 2.5P
μ Fmj (j = 1 : n) ≥ . (12)
n
From Eqs. 7, 8 and 9, to avoid sliding failure in any cases,
the magnetic force should be 3.3 Turn-over Failure Investigation
 P sin φ P sin φ 
Fm > max − P cos φ; + P cos φ . (10) Let l be the distance between first and last magnet block
μ μ contacting to the surface, and h be the distance between the
Since 0 < φ ≤ 90 ⇒ cos φ ≥ 0, the overall condition center of mass to the surface (Fig. 14c). Moment at point
for avoiding sliding failure is O (the point that the first magnet block contacts the steel
surface) is calculated as follows:
P sin φ
Fm > + P cos φ. (11)
μ  P ∗h
M = P ∗ h − 2Fm1 ∗ l = 0 ⇔ Fm1 = .
Assume that the frictional coefficient μ between two 2l
roller-chains and steel surface is from [0.4−0.8], we see that
sin φ
To avoid turn-over failure, the magnetic force of the first
+ cos φ decreases when μ increases, or we have: contacting magnet block:
μ
0.4 ≤ μ ≤ 0.8; 0 < φ ≤ 90 P ∗h
 sin φ  Fm1 > . (13)
2l
⇒ max + cos φ = 2.5 ⇒ Fm ≥ 2.5P .
μ
h
From Eq. 13, to avoid the failure we can lower , which
l
means making the robot’s center of mass closer to the steel
surface. In the proposed design (Fig. 10), h = 4.6cm, and
the total robot height hr = h + 9.231cm = 19.751cm.
Therefore, to avoid both sliding and turn-over failures, the
robot’s magnetic force of each magnet block should satisfy:
 2.5P P ∗ h 
r
Fmj (j = 1 : n) > max ; . (14)
n 2l

Following the proposed design, P = 30N, n = 16 magnet

blocks (each roller-chain has maximum 8 magnet blocks
contacting the steel surface), l = 9.8cm, or Fmj (j = 1 :
 2.5 ∗ 3 3 ∗ 19.751 
Fig. 14 a Sliding failure; b Turn-over failure; c Moment calculation at 16) > max ; > 3(N). We conducted
point O 16 2 ∗ 9.8
J Intell Robot Syst (2021) 102: 75 Page 11 of 21 75

Fig. 15 a Climbing on top

inclined surface; b Climbing
underneath inclined surface

some tests as discussed in Section 4.1 to make sure the The minimum pull-force is 13.5N at 100mm diameter
proposed design satisfied this condition. steel cylinder, and this means that it is always greater than
3N, the required force, as showed in Eq. 14.
3.4 Magnetic Force Analysis on Curved Surfaces
3.5 Motor Torque Analysis
The previous analysis is only applied when robot moves on
a flat steel surface. However, there are also structures, which Apart from the magnetic force analysis, we have conducted
have curving surface, so the following analysis will help another analysis to determine the appropriate motor to drive
determine the impact of curving surface on magnetic force the robot. In order to make the robot move, the force
created by the robot. created by the motor should win the adhesive force of the
An experiment conducted on different steel cylinder last permanent magnet and the steel surface. As shown in
shapes with real condition to verify the performance of Fig. 16, denote M as the torque of one motor, Q is rotation
magnet blocks working on curving surfaces is shown on fulcrum, i is the arm from Fmj to Q. Assume that the total
Table 5. driving force of the robot is the sum of two motor’s forces,
the required moment is satisfied:
P sin φ
M > hr ∗ + Fmj (j = 1 : n) ∗ i.
2
Table 5 Average Pull force (P F ) over inside and outside steel
cylinders of 15 × 10 × 5mm magnet block N35 When the robot moves on a vertical surface, maximum
φ = 900
Diameter (mm) PullForce P F (Newton - N)
P
M > hr ∗ + Fmj (j = 1 : n) ∗ i. (15)
100 13.5 2
200 15.5 → M > 3.16(N.m) with hr = 19.751cm, P =
300 16.2 30N, g = 9.8, Fmn = 39.7N (obtained this number from
400 17.4 the magnet’s datasheet), i = 0.5cm. The selected moving
500 19.3 motor moment is 1.2N.m and gear ratio is 7 : 20, hence the
600 20.9 moment of a robot roller-chain is 3.43N.m, which satisfies
700 23.2 (15).
800 25.4 Regarding transformation mechanism as shown in
900 28 Fig. 17, when the robot moves from flat surface to curving
flat 32.5 surface or reversing, the transforming mechanism works
−900 24.4 to make sure that the roller-chains contact steel surfaces
−800 21.2 with best condition. The transformation motor through
−700 19.7 mechanical system creates moment (M) to release magnet
−600 18.3 blocks from steel surface. However, the calculation gear
−500 17.2 ratio of reciprocating mechanism is not simple due to its
non-linearities, as showed in Eq. 5. We choose the special
Experiment was conducted on 5mm thickness steel with 0.2mm paint case when friction is strongest to calculate the moment of
on surface the transforming motor (powering feed screw mechanism).
75 Page 12 of 21 J Intell Robot Syst (2021) 102: 75

Fig. 16 a Robot moves on top of

inclined surface; b Robot moves
on bottom of inclined surface

The moment has to satisfy: 4 Robot Deployment

M > Ff ∗ e ⇔ M > e ∗ (k.N) ⇔ M > e ∗ k(P + Fm ).
To evaluate design and performance of the robot, experi-
ments for evaluating the magnetic force created by roller-
⇒ M > e ∗ k(P + nFmj (j = 1 : n)). (16) chains have been conducted. The ability of climbing and
failure avoidance were tested. During the test, a Lipo bat-
tery (2 cells) 7.4V 900 milliampere-hour (mAh) is used to
From Eq. 5 and kinematics parameter of reciprocating power the robot for about 1 hour of working. One laptop
mechanism, it is straight forward to calculate the gear ratio which can connect to a wireless LAN is used as a ground
in this case. The ratio α : y is approximately 1◦ : 1mm. station. The robot’s mass m = 3kg, and if we assume that
Besides, the ratio of a : e = 1 : 1.7, and the gear ratio of the gravitational acceleration g = 10m/s 2 , the total weight
the feed screw is 0.8mm : 360◦ . As result, a total gear ratio of the robot is approximately P = mg = 30N. Since
of whole system is 26.5 : 1. From Eq. 16 and assumption 15mm × 10mm × 5mm magnet blocks are used, the total
that system efficiency is 80%, the required moment of motor magnetic force is calculated as
is > 1.375N.m. The selected transformation motor with 
3.2N .m is satisfied. Fmj (j = 1 : 8) = 39.7(N)
(17)
When the robot travels on curving surfaces, reciprocating Fm = 2 ∗ 8 ∗ Fmj = 635.2(N)
mechanism is driven automatically by driving the motor
based on IR sensor signals. A PID controller is also applied which satisfies magnetic force condition as presented in
to maintain the distance between the IR sensor and the Eq. 14.
surface. This is to keep approaching area of permanent
magnets with the steel surface consistent.

Fig. 17 Transformation mechanism with 7 joints to create climbing

flexibility of the robot Fig. 18 Experimental setup for magnetic force measurement
J Intell Robot Syst (2021) 102: 75 Page 13 of 21 75

Fig. 19 Adhesive force

measurement in dynamic
condition on non-coated flat
surface (robot moves from right
to left images)

4.1 Adhesive Force Measurement Multiple tests have been conducted to measure the pull
force when the robot’s body is placed on different surfaces.
4.1.1 Static Condition The first test is on a flat non-coated steel surface, the second
one is on a flat coated steel surface while the third test is
In order to measure the adhesive force created by permanent on curve coated steel surfaces (positive and negative sides)
magnet, we have setup an environment as shown in Fig. 18. with diameters (D) ranging from 100mm to 900mm. All the
The robot’s body - whose mass is m = 3kg - is placed on top tests are executed three times, and the results are presented
of a flat steel surface while it is connected to a scale through in Figs. 21 and 22.
an inelastic wire. We create a pull force onto the scale trying
According to Eq. 14, the magnetic force for each
to lift the robot off the surface. At the time the robot is about
magnet block should be Fmj (j = 1 : n) > 3(N)
to be off the surface, the force applied to the scale is equal
to the sum of robot’s weight and the magnetic pull force. to avoid sliding and turn-over failure. Since at each
Denote Fpull as the force we applied onto the scale, M is the time the robot has 16 magnet blocks, which physically
value shown on the scale while P is the weight of robot’s contact the steel surface, the total required adhesive force
body, and Fmag is the magnetic force. With g = 10m/s 2 , should be greater than 3 × 16 = 48(N). We can see
P = mg = 30N and Fpull = Mg = 10M, we can calculate that in Figs. 21 and 22 the minimum magnetic force is
magnetic adhesive force as follows 115(N), which is much greater than the required one,
48(N) with safety factor is around 2.5. Therefore the robot
Fpull = P + Fm ⇒ Fm = Fpull − P
adheres well on both coated/non-coated flat and curving
⇒ Fmag = 10M − 3(N). (18) surfaces.

4.1.2 Dynamic Condition 4.2 Robot Field Deployment

We also measure the adhesive force of the robot in moving The outdoor experiments and robot deployments are
state as shown Fig. 19 (robot moves from right to left). Each conducted on more than 20 steel bridges. Due to limited
measurement is conducted three times as shown in Fig. 20. space, only some typical climbing examples are shown in
In average, the adhesive force in dynamic condition is lower Figs. 23–27. The steel structures have different thicknesses
than approximately 10% comparing to static state. of paint coated on steel surfaces. Some paint-coated steel

Fig. 20 (Left) testing result on flat surfaces, (right) testing result on coated curving surfaces (cylinder tubes)
75 Page 14 of 21 J Intell Robot Syst (2021) 102: 75

Fig. 21 Magnetic force measurements on coated and non-coated flat

steel surface

Fig. 22 Magnetic force measurements on curving steel surface with diameters (D) ranging from 100mm to 900mm

Table 6 Statistical result

Structrural parameters Decription

Thinnest steel surface 1 (mm)

Smallest steel cylinder diameter 100 (mm)
Thickest coated paint 4 (mm)
Highest nut or bolt area 12 (mm)

Fig. 23 Adhesive and climbing test on a thick paint-coated steel structure: flat surface
J Intell Robot Syst (2021) 102: 75 Page 15 of 21 75

Fig. 24 Adhesive and climbing test on a thin paint-coated steel structure: cylinder shape D=100mm

surfaces are very rusty, some are not clean, and some others can not pass these areas. Robot deployments can be seen
are still in fine condition (minor-rusty). Statistical results in the submitted video and the ARA lab link: https://ara.
with description of specifications that the robot can work cse.unr.edu/?page id=11 or youtube link: https://youtu.be/
with is presented in Table 6. 1Wl9Trd3EoM.
During the experiments, the climbing capability tests are The robot is controlled to move and stop at every certain
done on steel bridges and steel structures, and on with distance (e.g. 12cm) to capture images of steel surfaces and
coated or unclean surfaces as seen on Figs. 23, 24, 25, do closer investigation on potential crack areas, then send
26 and 27. The robot is able to adhere very well on these to the ground station. To enhance steel surface inspection,
steel structures while climbing with maximum speed is 200 acquired images are then stitched together to produce an
mm/s during 1 hour. Even for the case of curving surface overall image of steel surfaces as shown in Fig. 28. The
(Figs. 24 and 25), the robot can still adhere tightly to the image stitching is followed by our previously developed
steel structures while performing the climb. For rusty steel algorithm [41].
surfaces, it also shows strong climbing capability (Fig. 27). In addition to the visual data collection, the robot also
However, in some very tough conditions, such as bird drops collects Eddy current data as shown in Fig. 29. The survey
or high nuts, the robot might loose adhesive force and is conducted on a 170 x 70 mm area, which contains some

Fig. 25 Adhesive and climbing test on a thin paint-coated steel structure: cylinder shape D=250mm
75 Page 16 of 21 J Intell Robot Syst (2021) 102: 75

Fig. 26 Adhesive and climbing test on a thick paint-coated steel bridge: flat surface

crack areas as shown in Fig. 30a. The robotic arm holds the shapes to perform inspection and evaluation. The robot
Eddy current sensor’s probe and performs a zig-zag path design is implemented and validated on climbing on more
consisting of four lines. Figure 30b-e shows raw signals than twenty different steel bridges. During the tests, the
collected in a 13-seconds time frame for each line. We robot is able to firmly adhere on steel structures with
can see from these plots that at the crack areas, the signal various inclination levels. Rigorous analysis of magnetic
amplitude is higher. For the ease of illustration, a color adhesive force has been performed to confirm that the robot
coded map is built as shown in Fig. 31, and we can see is able to adhere to both flat and curving steel surfaces
that the Eddy current condition/fatigue crack map has defect in various conditions (coated, non-coated and/or rusty).
areas (red/yellow color) correlating very well with the crack Moreover, motor torque analysis to make sure the robot can
areas on the image. win the magnetic adhesive force to move on steel surface
is presented. Various experiments have been conducted
including magnetic force measurement, indoor and outdoor
5 Conclusion and Future Work climbing tests in order to validate the force analysis as well
as the climbing capability of the robot. The results show
This paper presents a new development of a tank-like robot, that when the magnetic adhesive force requirement is met,
which is capable of climbing on different steel structure the robot is able to move and transition safely between

Fig. 27 Adhesive and climbing test on a rusty paint-coated steel bridge: flat structure with bolts/nuts
J Intell Robot Syst (2021) 102: 75 Page 17 of 21 75

Fig. 28 Images stitching result:

(Top) 7 individual images taken
by the robot climbing on a
bridge in Fig. 27; (Middle)
Stitching image result from
those 7 individual images;
(Bottom) Closer look (zoom-in)
at some areas, which has serious
rusty condition with holes on the
surface

steel surfaces without any failures. Several sensors are Further works can be focusing localization on using
integrated to support the robot’s navigation as well as data odometry, IMU and visual data, and implementation of
collection. The rigorous magnetic force analysis can serve map construction methods as well as visual deep learning-
as a framework to calculate and design different types of based crack detection algorithms. The robot can be further
steel inspection robots in the future. Inspection data (Eddy equipped with other NDE sensors (e.g., thermal sensors) for
current and visual) is collected and transferred to ground more in-depth inspection of steel structures. Additionally,
station for visualization and processing. the robot can be further improved to move in circumfer-

Fig. 29 Crack surveying

process. Visual data is acquired
by camera 1 (top-right image).
Operator controls Eddy current
sensor’s probe remotely by a
haptic system (control arm:
bottom-left image) with closer
view from camera 2
(bottom-right image). Eddy
current sensor measures depth
of crack (top-left image), more
detail in Fig. 30
75 Page 18 of 21 J Intell Robot Syst (2021) 102: 75

Fig. 30 a Probe path; b line 4

data; c line 3 data; d line 2 data;
e line 1 data
J Intell Robot Syst (2021) 102: 75 Page 19 of 21 75

Fig. 31 a Image of an inspected area with cracks on it; b Fatigue crack map from Eddy current sensor showing defect areas with red/yellow color,
corresponding well with the crack areas on the image

ential directions on small diameter cylinder-like objects. 2. Stark, T.D., Benekohal, R., Fahnestock, L.A., LaFave, J.M., He,
Beyond steel bridge inspection, this robot can be applied J., Wittenkeller, C.: I-5 skagit river bridge collapse review. J.
Perform. Constr. Facil. 30(6), 04016061 (2016)
for inspection of oil tank, pipeline, and other civil facilities.
3. McCrea, A., Chamberlain, D., Navon, R.: Automated inspection
A collaborative algorithm can be developed to employ and restoration of steel bridges – a critical review of methods and
multiple robots for faster inspecting a large steel bridge or enabling technologies. Autom. Constr. 11(4), 351–373 (2002)
structure. 4. La, H.M., Dinh, T.H., Pham, N.H., Ha, Q.P., Pham, A.Q.:
Automated robotic monitoring and inspection of steel structures
and bridges. Robotica 37(5), 947–967 (2019)
5. Fischer, W., Tâche, F., Siegwart, R.: Magnetic wall climbing
Acknowledgements This work is supported by the U.S. National robot for thin surfaces with specific obstacles. In: Laugier, C.,
Science Foundation (NSF) under grants NSF-CAREER: 1846513 and Siegwart, R. (eds.) Field and Service Robotics: Results of the 6th
NSF-PFI-TT: 1919127, and the U.S. Department of Transportation, International Conference, pp. 551–561. Springer, Berlin (2008)
Office of the Assistant Secretary for Research and Technology 6. Fischer, W., Caprari, G., Siegwart, R., Moser, R.: Locomotion
(USDOT/OST-R) under Grant No. 69A3551747126 through INSPIRE system for a mobile robot on magnetic wheels with both axial
University Transportation Center, and the Vingroup Innovation and circumferential mobility and with only an 8-mm height for
Foundation (VINIF) in project code VINIF.2020.NCUD.DA094. The generator inspection with the rotor still installed. IEEE Trans. Ind.
views, opinions, findings and conclusions reflected in this publication Electron. 58(12), 5296–5303 (2011)
are solely those of the authors and do not represent the official policy 7. Ratsamee, P., Kriengkomol, P., Arai, T., Kamiyama, K., Mae, Y.,
or position of the NSF, the USDOT/OST-R and any other entities. Kiyokawa, K., Mashita, T., Uranishi, Y., Takemura, H.: A hybrid
flying and walking robot for steel bridge inspection. In: IEEE
Inter. Symp. on Safety, Security, and Res. Robo., pp. 62–67 (2016)
8. Wang, H., Yamamoto, A.: Analyses and solutions for the buckling
References of thin and flexible electrostatic inchworm climbing robots. IEEE
Trans. Robot. 33(4), 889–900 (2017)
1. FHWA: U.S Department of transportation highway administra- 9. Shen, W., Gu, J., Shen, Y.: Permanent magnetic system design for
tion, national bridge inventory data. http://www.fhwa.dot.gov/ the wall-climbing robot. In: IEEE International Conference Mecha-
bridge/nbi.cfm (2019) tronics and Automation, 2005, vol. 4, pp. 2078–2083 (2005)
75 Page 20 of 21 J Intell Robot Syst (2021) 102: 75

10. Mazumdar, A., Asada, H.H.: Mag-foot: A steel bridge inspection 28. Kamdar, S.: Design and manufacturing of a mecanum sheel
robot. In: 2009. IROS 2009. IEEE/RSJ International Conference for the magnetic climbing robot. Master Thesis, Embry-Riddle
on Intelligent Robots and Systems, pp. 1691–1696 (2009) Aeronautical University (2015)
11. Wang, R., Kawamura, Y.: A magnetic climbing robot for steel 29. Ward, P., Manamperi, P., Brooks, P.R., Mann, P., Kaluarachchi,
bridge inspection. In: 2014 11th World Congress on Intelligent W., Matkovic, L., Paul, G., Yang, C.H., Quin, P., Pagano, D., Liu,
Control and Automation (WCICA), pp. 3303–3308 (2014) D., Waldron, K., Dissanayake, G.: Climbing robot for steel bridge
12. Tche, F., Fischer, W., Caprari, G., Siegwart, R., Moser, R., inspection: Design challenges. In: Austroads Publications Online,
Mondada, F.: Magnebike: A magnetic wheeled robot with high ARRB Group (2015)
mobility for inspecting complex-shaped structures. J. Field Robot. 30. Pham, N.H., La, H.M.: Design and implementation of an
26(5), 453–476 (2009) autonomous robot for steel bridge inspection. In: 54th Allerton
13. Pack, R.T., Christopher, J., Kawamura, K.: A rubbertuator-based Conf. on Comm., Con., and Comp., pp. 556–562 (2016)
structure-climbing inspection robot. In: 1997. Proceedings., 1997 31. Wang, R., Kawamura, Y.: Development of climbing robot for steel
IEEE International Conference on Robotics and Automation, vol. bridge inspection. Indust. Robot: Int. J. 43(4), 429–447 (2016)
3, pp. 1869–1874 (1997) 32. Pham, N.H., La, H.M., Ha, Q.P., Dang, S.N., Vo, A.H., Dinh,
14. Abderrahim, M., Balaguer, C., Gimenez, A., Pastor, J.M., Padron, Q.H.: Visual and 3d mapping for steel bridge inspection using a
V.M.: Roma: a climbing robot for inspection operations. In: 1999. climbing robot. In: The 33rd Intern. Symposium on Automation
Proceedings. 1999 IEEE International Conference on Robotics and Robotics in Construction and Mining (ISARC), pp. 1–8
and Automation, vol. 3, pp. 2303–2308 (1999) (2016)
15. Leibbrandt, A., Caprari, G., Angst, U., Siegwart, R.Y., Flatt, R.J., 33. Takada, Y., Ito, S., Imajo, N.: Development of a bridge inspection
Elsener, B.: Climbing robot for corrosion monitoring of reinforced robot capable of traveling on splicing parts. Inventions 2 (2017)
concrete structures. In: Applied Robotics for the Power Industry 34. Versatrax: Versatrax 150T M . http://inuktun.com/en/products/
(CARPI), the 2nd Intern. Conf. on, pp. 10–15 (2012) (2019)
16. Leon-Rodriguez, H., Hussain, S., Sattar, T.: A compact wall- 35. Sato, G., Mae, Y., Kojima, M., Arai, T.: Transfer motion on planar
climbing and surface adaptation robot for non-destructive testing. structure of limb-mechanism robot in anti-gravity environment.
In: 2012 12th International Conference on Control, Automation In: Proceedings of 2018 ISFA International Symposium on
and Systems (ICCAS), pp. 404–409 (2012) Flexible Automation, pp. 467–472 (2018)
17. San-Millan, A.: Design of a teleoperated wall climbing robot for
36. Van Nguyen, L., Gibb, S., Pham, H.X., La, H.M.: A mobile
oil tank inspection. In: 2015 23th Mediterranean Conference on
robot for automated civil infrastructure inspection and evaluation.
Control and Automation (MED), pp. 255–261 (2015)
In: 2018 IEEE International Symposium on Safety, Security, and
18. Shen, W., Gu, J., Shen, Y.: Proposed wall climbing robot with
Rescue Robotics (SSRR), pp. 1–6 (2018)
permanent magnetic tracks for inspecting oil tanks. In: IEEE
37. La, H.M., Gucunski, N., Dana, K., Kee, S.-H.: Development of an
International Conference Mechatronics and Automation, 2005,
autonomous bridge deck inspection robotic system. J. Field Robot.
vol. 4, pp. 2072–2077 (2005)
34(8), 1489–1504 (2017)
19. Tavakoli, M., Viegas, C., Marques, L., Pires, J.N., deAlmeida,
38. Gibb, S., Le, T., La, H.M., Schmid, R., Berendsen, T.: A multi-
A.T.: Omniclimbers: Omni-directional magnetic wheeled climb-
functional inspection robot for civil infrastructure evaluation and
ing robots for inspection of ferromagnetic structures. Robot.
maintenance. In: 2017 IEEE/RSJ International Conference on
Auton. Syst. 61(9), 997–1007 (2013)
Intelligent Robots and Systems (IROS), pp. 2672–2677 (2017)
20. Eich, M., Vgele, T.: Design and control of a lightweight magnetic
39. Le, T., Gibb, S., Pham, N., La, H.M., Falk, L., Berendsen,
climbing robot for vessel inspection. In: The 19th Mediterranean
T.: Autonomous robotic system using non-destructive evaluation
Conf. on Control Automation, pp. 1200–1205 (2011)
methods for bridge deck inspection. In: 2017 IEEE International
21. Caprari, G., Breitenmoser, A., Fischer, W., Hurzeler, C., Tache,
Conference on Robotics and Automation (ICRA), pp. 3672–3677
F., Siegwart, R., Nguyen, O., Moser, R., Schoeneich, P., Mondada,
(2017)
F.: Highly compact robots for inspection of power plants. J. Field
40. La, H.M., Lim, R.S., Basily, B.B., Gucunski, N., Yi, J., Maher,
Robot. 29(1), 47–68 (2012)
A., Romero, F.A., Parvardeh, H.: Mechatronic systems design
22. Cho, K.H., Kim, H.M., Jin, Y.H., Liu, F., Moon, H., Koo, J.C.,
for an autonomous robotic system for high-efficiency bridge
Choi, H.R.: Inspection robot for hanger cable of suspension
deck inspection and evaluation. IEEE/ASME Trans. Mechatronics
bridge: Mechanism design and analysis. IEEE/ASME Trans.
18(6), 1655–1664 (2013)
Mechatron. 18(6), 1665–1674 (2013)
41. La, H.M., Gucunski, N., Kee, S.-H., Nguyen, L.V.: Data analysis
23. Zhu, D., Guo, J., Cho, C., Wang, Y., Lee, K.: Wireless
and visualization for the bridge deck inspection and evaluation
mobile sensor network for the system identification of a space
robotic system. Vis. Eng. 3(1), 1–16 (2015)
frame bridge. IEEE/ASME Trans. Mechatron. 17(3), 499–507
42. La, H.M., Gucunski, N., Kee, S.H., Nguyen, L.V.: Visual and
(2012)
acoustic data analysis for the bridge deck inspection robotic
24. Lee, G., Wu, G., Kim, J., Seo, T.: High-payload climbing and
system. In: The 31st International Symposium on Automation and
transitioning by compliant locomotion with magnetic adhesion.
Robotics in Construction and Mining (ISARC), pp. 50–57 (2014)
Robot. Auton. Syst. 60(10), 1308–1316 (2012)
43. Lim, R.S., La, H.M., Shan, Z., Sheng, W.: Developing a crack
25. Seo, T., Sitti, M.: Tank-like module-based climbing robot using
inspection robot for bridge maintenance. In: 2011 IEEE Intern.
passive compliant joints. IEEE/ASME Trans. Mechatron. 18(1),
Conf. on Robotics and Automation (ICRA), pp. 6288–6293 (2011)
397–408 (2013)
44. Lim, R.S., La, H.M., Sheng, W.: A robotic crack inspection
26. Wang, R., Kawamura, Y.: A magnetic climbing robot for steel
and mapping system for bridge deck maintenance. IEEE Trans.
bridge inspection. In: Proceeding of the 11th World Congress on
Autom. Sci. Eng. 11(2), 367–378 (2014)
Intelligent Control and Automation, pp. 3303–3308 (2014)
45. Gibb, S., La, H.M., Le, T., Nguyen, L., Schmid, R., Pham, H.:
27. Guo, J., Liu, W., Lee, K.M.: Design of flexonic mobile node
Nondestructive evaluation sensor fusion with autonomous robotic
using 3d compliant beam for smooth manipulation and structural
system for civil infrastructure inspection. J. Field Robot. 0(0)
obstacle avoidance. In: 2014 IEEE International Conference on
(2018)
Robotics and Automation (ICRA), pp. 5127–5132 (2014)
J Intell Robot Syst (2021) 102: 75 Page 21 of 21 75

46. Sutter, B., Lelevé, A., Pham, M.T., Gouin, O., Jupille, N., Kuhn, Son Thanh Nguyen (IEEE M’2019) received his B.S. degree in
M., Lulé, P., Michaud, P., Rémy, P.: A semi-autonomous mobile Mechatronic Engineering from Military Technical Academy, Hanoi,
robot for bridge inspection. Autom. Constr. 91, 111–119 (2018) Vietnam in 2011, and M.S. degree in Control Engineering and
47. La, H.M., Gucunski, N., Kee, S.-H., Yi, J., Senlet, T., Nguyen, Automation from Ho Chi Minh University of Transport, Ho Chi
L.: Autonomous robotic system for bridge deck data collection Minh City, Vietnam in 2015. He is currently working toward his PhD
and analysis. In: 2014 IEEE/RSJ International Conference on degree at the Advanced Robotics and Automation (ARA) Laboratory,
Intelligent Robots and Systems, pp. 1950–1955 (2014) Department of Computer Science and Engineering, University of
48. Gillins, D.T., Parrish, C., Gillins, M.N., Simpson, C.: Eyes Nevada, Reno, NV89557, USA. His research interest includes robotic
in the sky: Bridge inspections with unmanned aerial vehicles. design, robotic control, and automated civil infrastructure inspection.
https://www.oregon.gov/ODOT/Programs/ResearchDocuments/
SPR787 Eyes in the Sky.pdf (2018)
49. Eschmann, C.W., Kuo, C.-M., Kuo, C., Boller, C.: Unmanned Hung Manh La (IEEE M’2009, ASCE M’2020) received his B.S.
aircraft systems for remote building inspection and monitoring. and M.S. degrees in Electrical Engineering from Thai Nguyen
In: The 6th European Workshop on Structural Health Monitoring University of Technology, Thai Nguyen, Vietnam, in 2001 and
(2012) 2003, respectively, and his Ph.D. degree in Electrical and Computer
50. Hallermann, N., Morgenthal, G.: Visual inspection strategies for Engineering from Oklahoma State University, Stillwater, OK, USA,
large bridges using unmanned aerial vehicles (uav). In: Bridge in 2011. Dr. La is the Director of the Advanced Robotics and
maintenance, safety, management and life extension, pp. 661–667 Automation (ARA) Lab, and Associate Professor of the Department
(2014) of Computer Science and Engineering, University of Nevada, Reno,
51. Zink, J., Lovelace, B.: Unmanned aerial vehicle bridge inspection NV, USA. From 2011 to 2014, he was a Post Doctoral research fellow
demonstration project final report. http://www.dot.state.mn.us/ and then a Research Faculty Member at the Center for Advanced
research/TS/2015/201540.pdf (2015) Infrastructure and Transportation, Rutgers University, Piscataway, NJ,
52. Dorafshan, S., Maguire, M.J.: Bridge inspection: human perfor- USA. He has authored over 120 papers published in major journals,
mance, unmanned aerial systems and automation. Civil Struct book chapters and international conference proceedings. His current
Health Monit (2018) (2018) research interests include robotic systems and mobile sensor networks.
53. Intel-Drone. Intel Flies Drones for Bridge Inspection, Decem- Dr. La is the recipient of the 2019 NSF CAREER award, and
ber 07, 2018. Available at https://www.engineering.com/BIM/ the 2014 ASCE Charles Pankow Award for the Robotics Assisted
ArticleID/18113/Intel-Flies-Drones-for-Bridge-Inspection.aspx Bridge Inspection Tool (RABIT), eight best paper awards/finalists,
(2018) and a best presentation award in international conferences (e.g.,
54. K&J Magnetics, I.: Original magnet calculator. https://www. ICRA, IROS, ACC). Dr. La is currently an Associate Editor of
kjmagnetics.com/ (2019) the International Conference on Robotics and Automation (ICRA),
the IEEE Transactions on Human-Machine Systems, the Journal of
Publisher’s Note Springer Nature remains neutral with regard to Drones, the Frontiers in Robotics & AI, and the International Journal
jurisdictional claims in published maps and institutional affiliations. of Robotic Engineering, respectively. Dr. La has served as an Associate
Editor of the International Journal of Automation and Control, and
a Guest Editor for International Journal of Robust and Nonlinear
Control.