1 INTRODUCTION

Climbing robot have various applications in industry and

hazardous environments: inspection oI high chimneys,
vertical and inclined pipes in nuclear power plants, wiring
on high voltage power transmission towers are some well
known examples oI such applications |1|. Moreover, there
are some new and other important tasks Ior these robots, Ior
instance, examining cables on a cable-stayed bridge.
Cables are one oI the most important stress components oI
the cable-stayed bridges. In the United States, several
cable-stayed bridges have shown signs oI cable damages
mostly induced by susceptibility oI stay cables to
wind-induced vibrations and corrosion |2|. This has caused
public concerns and initiated a series oI inspection and
security monitoring programs. The cables must be
examined periodically to ensure the reliability oI the bridge.
For a long time, the inspection was conducted artiIicially,
which is ineIIicient and dangerous. Under these
circumstances, this paper proposed a new wheel-based
cable climbing robot, Ior the cable inspection usages.
Climbing robots belong to a specialized Iield oI mobile
robots. The main Ieature is the mobility against gravity oI
the body. An important aspect to consider in the design oI
such robots is that they need to be lightweight and powerIul
enough to move up to support their own weight |3|.
ThereIore, the designer should consider not only the
locomotion method, as in conventional mobile robots, but
also the techniques sticking to the cable. There have
emerged several Iorms oI adhesion methods:
Magnetic mechanisms Ior climbing on Ierrous
surIaces via electromagnets or permanent magnets
|4|.

This work was supported by the National 863 Plan
(No.2006AA04Z234) and New Century Excellent Talent Plan oI MOE oI
China (No. NCET-04-0479)
Vacuum suction technologies Ior sticking the
robot on the walls |5| |6|.
Grippers or armed mechanisms that attach to the
structure such as beams, columns, pipes, and tubes
|7|.
According to the modes oI adhesion, climbing robot can be
classiIied into three main groups:
Wheel-driven machines climb vertical planes by
combining wheels Ior translation, and by magnets
or vacuum suction Ior surIace attachment |8|.
Legged climbing robots |9| consist usually oI Iour
or six legs, each oI them with magnets or a vacuum
pump Ior attachment but with limited
maneuverability
Locomotion based on arms with grippers or other
devices that provide the robot with skillIul
mobility |10|.
The climbing robot proposed in this paper utilized springs
to provide Iriction Iorces to support the whole robotic
systems. With the driving wheel actuated by a powerIul DC
motor, the robot can climb on a cable inclined at any angles,
including a vertical one.
2 STRUCTURE AND STATICS OF THE
ROBOT
The principle oI the climbing robot is shown in Fig.1. The
maximal static Iriction Iorce to hold the mechanism is
denoted by F .
For mechanism in Fig.1, we have:
G N F =
-static Iriction coeIIicient,
N -clasping Iorce,
G -the gravity oI the climbing mechanism and the
load.
In order to less the clasping Iorce , the Iriction coeIIicient oI
the wheel coating material to the cable should be as large as

A Wheel-Based Cable Climbing Robot With Descending Speed Restriction
Fengyu Xu , Xingsong Wang
School oI Mechanical Engineering, Southeast University, Nanjing, 211189, China
E-mail: xuIengyu598163.com

Abstract: A new wheel-based cable climbing robot which is able to climb up the vertical cylindrical cable on the
cable-stayed bridge was proposed. Unlike the previous one, the new designed robot proposed in this paper is composed
oI two equally spaced modules, which are joined by connecting bars to Iorm a closed hexagonal body to clasp on the
cable. The static Ieatures oI the robot were analyzed and the balanced torque oI the mechanism was given. For the present
design, it can climb up a cable with diameters varying Irom 65mm to 205mm with payloads below 3.5kg. A cable with a
diameter oI 139mm was selected as an example to calculate the parameters oI the robot. For the saIety landing in case oI
electricity broken-down, a gas damper with a slider-crank mechanism is introduced to exhaust the energy generated by
the gravity when the robot is slipping down along the cables. The landing system is tested experimentally and a
simpliIied mathematical model is analyzed. Several climbing experiments perIormed on real cables show the capability
oI the proposed robot.
Key Words: Cable-stayed bridge, climbing robot,slider-crank mechanism, gas damper
1570
978-1-4244-1734-6/08/$25.00 c 2008 IEEE
possible. The required driving Iorce to climb is denoted as
Q, and it is obvious that F Q
N G Q
1
+ =
1
- coeIIicient oI rolling Iriction.
So the required power oI DC motor P is:
J N G J Q P ) (
1
+ = =
J - the climbing velocity oI the mechanism,
- driving eIIiciency.
G
N
T
Q

Fig 1: Principle oI Climbing.
Unlike our previously designed climbing robot which was a
trisected robot |12|, there are several improvements
introduced here.
The Iirst improvement is that the climbing robot proposed
here (shown in Fig.2), is composed oI two equally spaced
modules Iacing each other. Only one oI the two modules is
powered and the other is passive. Each module possesses
two wheel limbs at the two ends. Only the upper wheel oI
the powered module is actuated with a DC motor. The
passive module is applied to increase the system stability
and provide the supporting Iorce.
The two modules were joined by eight connected bars to
Iorm a whole closed hexagonal body to clasp the cable. By
changing the joined holes oI the connecting bars, the robot
can climb cables with varies diameters. All the wheels
(including the driving wheel) were machined in V-shape
with open angle =150 to increase the contacting area.
This will also reduce the abrasion, prevent the driving
wheel deIlecting oII the cable, and avoid getting stuck.
As shown in Iig.3, the wheel limbs oI the passive module
are connected with an extension spring to provide the robot
with enough clasping Iorce. The limbs hinged with the body
oI the passive modules also enable the robot climbing over
small obstacles by the elongating oI the springs. In order to
reduce the drag Iorce and to ensure only one wheel crawl
the obstacle at one time, the passive wheel was oIIset 'd
Irom the driving wheel. (As showed in Fig.5-a)
Through a bevel gear reducer, the motor drives the wheel
Iixed on the body oI the powered module via a magnetic
clutch. In case oI power shortage or electrical Iailure, in
order to ensure the robot saIely slipping down due to its
gravity, there is also a saIety landing mechanism attached to
the driving wheel oI the robot through a one-way clutch.
This is the second improvement. The principle oI the
landing mechanism will be discussed in detail in section 4.
When the robot climbs up, the magnetic clutch locked to let
the driving power acting on the driving wheel. Meantime,
the one-way clutch is Iree and the landing mechanism does
not work.
An encoder is applied to measure the distance the robot
climbed. And there is moustache sensor to detect the top
end oI the cable. When it arrives at the top end, the driving
motor will be powered oII and the robot will slip downward
by its gravity. At this time the saIety landing mechanism
will work to ensure the robot saIely landing without
consuming any energy.

Fig 2 Structure oI the climbing robot.

Fig 3 Passive module oI the robot.

Fig 4 the structure oI the driving vehicle.


2008 Chinese Control and Decision Conference (CCDC 2008) 1571
3 STATIC FEATURE ANALYSIS
Fig.5-b is the top view oI the climbing mechanism. The
upper wheels are called as wheel1 and wheel2, while the
lower wheels are called wheel3 and wheel4 respectively.
Suppose
1 i
N and
2 i
N ( i is the number oI the wheel) are the
Iorces the cable acting on the two contacting points oI
wheel i ,
i
N is the resultant Iorce oI
1 i
N and
2 i
N , then:
2 sin 2 sin
2 1

i i i
N N N + = (1)
The direction oI the resultant Iorce is perpendicular to the
axes oI the wheel.
1 fi
F and
2 fi
F are the rolling Irictional
resistance that the cable acts on the two points oI wheel i .
As the wheels are identical, according to symmetrical
characteristic oI the passive wheels and the symmetrically
arrangement around the cable, we can have:
2 1 fi fi
F F =
2 1 i i
N N = (2)

+ = + =
=

2 1 2 1
) 1 (
i i i fi fi fi
if fif
N N F F F
i N F
(3)
fi
F denotes the resultant Iorce oI
1 fi
F and
2 fi
F .
Direction oI
fi
F is along the cable.
All the Iorces acting on the robot system are shown in Fig.5.
Assume that the two modules are same in clamping the
cable and are same in supporting oI the loads and their own
gravities. In the direction oI y axis, we have the Iollowing
equilibrium equations.
In the plane oI x: ,

= + = 0
4 2 3 1
N N N N X (4)

=
3 2 4 1 f f f f
F F F Mg F Z (5)
0 |
) ( | 15 cos 2
) ( (
) ( ) ( ) (
41 31 21 11
41 31 21
1
4
1
2 1
4
1
) 1 1
4
1
2 1
4
1
1 1 1 1
= + + +
+ + +
= +
+ + =


= =
= =
r F r F r F r F
d L N L N d N
mgl F M F M
N M N M mgl F M
f f f f
i
fi o
i
fi o
i
i o
i
i o o

(6)
Where is the output torque oI the driving motor.
P is the gravity center oI the mechanism;
r is the radius oI the cable.
d is the eccentric distance Irom passive wheel to the
driving wheel.
The Iorces diagram oI wheel 2 as Fig.6-(a) shows:

= =
+ + =
15 cos 2 sin
cos
21 2 2
2 2 2
N N F
F g m F F
f w t

(7)
Where denotes the angle between limb and the body;
g m
wi
is the gravity Iorce oI the wheel i ;
t
F is the pulling Iorce oI the spring;
i
F is the supporting Iorce limb acting on the passive
wheel. From (1) (3) (7):
) 15 (cos 2
) (
1
2
22 21

tg
tg g m F
N N
w t

+
= = (8)
1
N

d
2 f
F
2
N
:
2 wheel
1 wheel
1
O
1 f
F
Mg
P
1
l
L
3 wheel
4 wheel
x
O
3
N
d
4
N
3 f
F
4 f
F

(a)
x
v
12
N
2 wheel
22
N
1 wheel
21
N
21
N

(b)
Fig 5: view oI the climbing mechanism.
2
N
22
N
21
N
t
F
g m
w2
150
o

2
F
2 f
F

(a)
t
F
150
o

4
N

42
N
41
N
4
F
2 f
F
g m
w4

(b)
1 f
F
12
N
11
N
1
N
g m
w1
1
F
2
O

150

(c)
Fig 6: Force diagram oI the wheels.
1
is the coeIIicient oI rolling Iriction.
The Iorce diagram oI wheel4 shown in Fig. 6-(b),

= =
+ + =
15 cos 2 sin
cos
41 4 4
4 4 4
N N F
g m F F F
w f t

(9)
From (1) (3) (9):
) 15 (cos 2
) (
1
4
42 41

tg
tg g m F
N N
w t
+
+
= = (10)
The diagram oI the driving wheel shown in Fig.6-(c)
1572 2008 Chinese Control and Decision Conference (CCDC 2008)
1 1
r F
f
= (11)
1
F is the Iorce that the robot body acts on the driving
wheel;
1
r is the diameter oI the driving wheel.
From (2) (3) (5):
1 41 1 31 1 21 11 1
2 2 2 2 N N N Mg F F
f f
+ + + = = (12)
According to (7) (11) (12):
A
C N B N r r l Mg
N
2
2 2 ) 15 cos (
41 21 1 1
31
+
= (13)
Where, ) ( 15 cos ) 15 cos (
1 1
r L r r A + + =
) ( 15 cos ) 15 cos (
1 1 1
r d r r B + + + =
) 15 cos ( ) ( 15 cos
1 1 1
+ + = r r r d L C
ThereIore, when the robot climbing at a uniIorm speed, the
output torque oI the DC motor is:
1 1 41 1 51 1 21
) 2 2 2 (
1 1
r N N N Mg r F
f
+ + + = = (14)
Take a 139 mm cable as an example, the material oI the
driving wheel is aluminum coating with hard rubber and the
passive wheels are MC nylon; mm d 20 = , mm l 17
1
= , other
parameters shown in table1.
From (4), (8), (10), (12), (13), we have:
N N 25 . 99
21
= , N N 7 . 92
41
= , N N 35 . 75
31
= ,
N N 6 . 116
11
= , N F
f
73 . 118
1
= .
The maximal static Iriction is N N 56 . 186 2
2 11
=
Setting
1 2 11
2
f
F N , thus the mechanism satisIies the
climbing condition.
According to (14), we can get Nm 13 . 4 =
Maxon DC motor (P90W) was chosen as the driving
motor. Experiments showed that the robot could climb the
cable reliably.
The prototype cable climbing robot is shown in Fig.2. It can
climb a cable smoothly at any slant angle, including vertical
ones, which is the most diIIicult case Ior climbing.
Table1 Symbols and values
Symbol Quantity Value
G1 Gravity oI the robot 5.7kg
G2 Payload 3.5kg
r1 EIIective radius oI the driving wheel 34.82mm
L1 Distance between wheel1 and wheel3 315mm
K StiIIness coeIIicient oI the spring 3.61N/mm
Ft Pulling Iorce 321N
D Diameter oI the cable 139mm
mw Passive wheels` weight 0.138kg
1 CoeIIicient oI rolling Iriction 0.05

Angle between limb and body 30
0

2 static Iriction coeIIicient 0.8
4 ANALYSIS OF THE LANDING
MECHANISM
In this section, the second big improvement oI the landing
speed constraint mechanism is introduced. In order to make
the robot land in a controllable velocity (the value is not too
large) in case oI power shortage or electrical Iailure, the
new designed robot have a Ialling speed control mechanism
consisting oI a pneumatic cylinder and a slider-crank
mechanism whose crank Iixed with the driving shaIt
through the one-way clutch (as shown in Iig2). While the
robot is slipping down, the rotation motion oI the driving
wheel is transmitted to the piston`s reciprocating motion in
the pneumatic cylinder. The gas in the cylinder inhaled or
ejected alternately Irom the hole Iorming a gas damper to
consume the kinetic energy oI the robot. The size oI the
hole can be adjusted to obtain diIIerent damper ratios so as
to control landing speed oI the robot.
The principle oI the landing mechanism is shown in Iig7. A
slider-crank mechanism is a singlelooped mechanism,
which has a constrained condition as Iollows:
) sin arcsin(
2
=
l
r
(15)
The rotary inertia oI the connecting rod is very small, so the
connecting rod can be considered as a two-Iorce member
neglecting its reciprocating and rotating,

cos
) (
2 0
m p p A
Fe

=
For the crank, the angular velocity and the acceleration are:

+ =
=

)) sin( (
2
1

r Fe J
dt
d
w
(16)

is the acceleration oI the piston.

The displacement oI the piston is:

=
+ + + =
l
r
r r L B x
2
2 2
)
4
2 cos
4
(cos

(17)
The speed and acceleration oI piston are as Iollows:

+ = =
+ = =
) 2 sin (cos
) 2 sin
2
(sin
2
2
2

w r
dt
dJ
dt
d
r
dt
dx
J
(18)
When slipping down the cables, the robot accelerates at the
beginning, and Iluctuates aIter the velocity achieves a Iixed
value. So the robot is in quasi-uniIorm motion, the
approximate torque acting on the crankshaIt is:
Mg r
2 2
=

(a)

(b)
Fig 7: Landing mechanism
1crank; 2connecting rod; 3linear bearing; 4 pneumatic cylinder;
5 piston connecting rod; 6 piston; 7 nozzle.
The physical process involved in the emission and suction
oI gas is very complex. To simpliIying the analysis, some
2008 Chinese Control and Decision Conference (CCDC 2008) 1573
assumptions are made: (1) the gas Ilows adiabatically in the
cylinder and is isentropic at the release point; (2) the model
oI Ilow is considered as a one-dimension model.
The value oI the release rate in the hole depends on whether
gas Ilow is sonic or subsonic. This will be decided by the
critical pressure ratio |11|:
528 . 0 )
1
2
(
1
=
+
=
k
k
k
CPR
When 528 . 0 CPR , the gas release in the hole is a sonic
Ilow. The release rate passing the oriIice is described by the
Iollowing equation:
1
1
1
)
1
2
(

+
+
=
k
k
k ZRT
M
k AP Q (19)
When 528 . 0 CPR , the gas release is subsonic, and the
release rate is:
| ) ( ) |(
1
2
1
1
0
2
1
0
1
k
k
k
p
p
p
p
k
k
ZRT
M
AP Q
+

= (20)
The speciIic volume is:
J
m
= =

1

So the volume Ilow is:

+

=
+

+
| ) ( ) |(
1
2
)
1
2
(
1
1
0
2
1
0
1
1
1
1
k
k
k
or
k
k
or
p
p
p
p
k
k
ZRT
M
P A
k ZRT
M
k P A
C


(21)
Where
or
A is the area oI hole(m2), C is the empirical
discharge coeIIicient(Ior subsonic gas oI Reynolds number
larger than 30,000, C0.61; Ior other situations C1, Q is
mass Ilux (kg m-2s-1)), Q is release rate(kg s-1), K is
heat capacity ratio, M is molecular weight(kg/k mol),
1
P is
pressure oI the upstream(Pa), R is the constant oI gas(Pa
m3 Mol-1 K-1), T is the temperature oI gas (K), is
density oI gas(kg m-3), J is inertia oI the crank.
The pressure in the cylinder is inIluenced by two Iactors:
(1) the variety oI the cylinder`s volume caused by the
movement oI the piston; (2) the gas Ilowing through the
hole caused by the pressure diIIerence inside and outside oI
the cylinder.
K K
t
J J
C
P
J J
J
P P ) ( ) (
0 0
0
+

+
= (22)
t
P is the instantaneous pressure aIter the movement oI
the piston and the gas Ilowed; P is the instantaneous
pressure;
0
J is the original volume; the value oI J , the
variety oI the volume caused by the piston, is positive when
the volume is increasing, otherwise J is negative; C is
the volume Ilow mentioned above.
The Quasi-steady-state (QSS) simulation methodology was
used to simulate the landing mechanism. The whole
concept is to replace the transient dynamic state by its QSS
equilibrium state. Divide the whole process into many
small time segments, denoting as t , each oI which is a
simulation step. One can regard this process as a
steady-state when t is small enough. The change oI
pressure and the movement oI the piston occur
simultaneously. The air input oI the cylinder is calculated
according to the steady discharge. This is called
quasi-steady-state simulation methodology which is
applied widely in engineering. With the method, according
to the state at the beginning oI t , one can obtain the
piston`s motion state and the pressure in the cylinder at the
end oI t , which can be the parameters used in the next t .
The smaller t is, the greater precision obtained. By
adjusting t , the accuracy oI simulation will increase. This
is called varying-step simulation.
Varies tests oI movement were perIormed on horizontal,
inclined and vertical cables. The robot was Iirst tested on an
iron tube inclined about 75 , Iollowed by trials perIormed
on a more uneven surIace (Iive meters cable used on
SuTong Bridge) to veriIy the reliability oI the contact under
conditions similar to those on bridges. All the tests
perIormed gave good results; in no case did the wheels lose
contact with the cable, and aIter cut oII the power the robot
can return to the ground saIely.
Fig.8 is simulation curve oI the landing velocity and the
pressure in the cylinder when the robot is slipping down a
vertical cable in the laboratory. The abscissa is the angle oI
the crank. The diameter oI the testing cable is 139 mm, the
diameter oI the holes is 3mm. When reached 0.7m/s, the
velocity began to Iluctuate at the velocity oI 1m/s. The
operation oI the robot is shown in Fig.9-(a) and Fig.9-(b).
5 CONCLUSION AND FUTURE WORK
In this paper, we proposed an innovated wheel-based cable
climbing robot that can climb cables with diameters varying
Irom 65mm to 205mm. Its advantages are the small
physical volume, simple structure, and high running speed.
Furthermore, it can move on the cable reliably with about
3.5kg payloads to meet the demands oI cable detecting
tasks. Furthermore, we Iirstly proposed a landing
mechanism using a gas damper. Both oI the simulation
results and the spot test results indicate that the mechanism
can guarantee the return oI robot to the ground saIely.
Further work in vortex-induced cable vibration oI cable and
study on the leakage method are needed to consummate the
whole system`s inspecting precision. Furthermore, the
inIluence oI the robot to the cable surIace is very important,
especially to the long one.
REFERENCES
|1| D.Bevly, Steven D. and C. Mavroidis, 'A SimpliIied
Cartesian Computed Torque Controller Ior Highly Geared
Systems and Its Application to an Experimental Climbing
Robot, Journal oI Dynamic Systems, Measurement and
Control, Trans. ASME, 122(1), 2000, pp. 27-32.
|2| Mehrabi, Armin B., Telang, Niket M., Ghara, Hossein,
Fossier, Paul, 'Health monitoring oI cable-stayed bridges - A
case study, Proceedings oI the 2004 Structures Congress -
Building on the Past: Securing the Future, May 22-26, 2004,
Nashville, TN, United State, pp. 59-66.
|3| Zaidi Mohd,Tan Beng Soon, A.B.Abdullah, 'Development
oI a Low Cost Modular Pole Climbing Robot, TENCON
2000 Proceedings, 2000, pp.196-200.
1574 2008 Chinese Control and Decision Conference (CCDC 2008)
|4| Wang Yan, Liu Shuliang, Xu Dianguo etc., 'Development
and application oI Wall-Climbing Robots, Proc. oI the 1999
IEEE International ConIerence on Robotics and Automation,
Detrout, Michigan, 1999,pp. 1207-1212.
|5| Briones.L, Bustamante.P, Serna.M. A, 'Wall-climbing robot
Ior inspection in nuclear power plants, Proc. oI the 1994
IEEE International ConIerence on Robotics and Automation,
1994, pp.1409-1414.
|6| G.La Rosa, M.Messina, G.Muscato, 'A low-cost lightweight
climbing robot Ior the inspection oI vertical surIaces,
Mechatronics 12(1), 2002, pp.71-96.
|7| T.White, N.Hewer, N.Luk, 'The design and operational
perIormance oI a climbing robot used Ior weld inspection in
hazardous environments,Proc.1998 IEEE International
ConIerence on Control Applications, Trieste, Italy, 1998,
pp.451-455.
|8| S.Weimin, G.Jason, S.Yanjun, 'Permanent Magnetic
System Design Ior the Wall-climbing Robot, Proc. oI the
IEEE International ConIerence on Mechatronics and
Automation, Niagara Falls, Canada, 2005, pp.2078-2083.
|9| Bing L.Luk, David S.Cooke, Stuart Galt, 'Intelligent legged
climbing service robot Ior remote maintenance applications
in hazardous environments, Robotics and Autonomous
Systems, 53(2005), 2005, pp.142-152.
|10| C.Balaguer, J.Pastor, A. Gimenez, V. pardon, and M.
Abderrahim, 'ROMA: MultiIunctional autonomous
selI-supported climbing robot Ior inspection applications,
proc. Int. ConI. Intelligent Automonous Vehicles, Madrid,
Spain,1998.
|11| Dong Yuhua, Gao Huilin, Zhou Jing`en, Feng Yaorong,
'Evaluation oI gas release rate through holes in pipelines,
Journal oI Loss Prevention in the Process Industries, 15(
2002), pp.423-428.
|12| Xingsong Wang, Fengyu Xu, Conceptual Design and Initial
Experiments on Cable Inspection Robotic System, IEEE Int.
ConI. on Mechatronics and Automation, Harbin, China, Aug.
6-8, 2007.


(a) the slipping velocity

(b) the pressure in the cylinder
Fig 8 Landing velocity and the cylinder pressure.

a

(b)
Fig 9: the operation oI the robot.

2008 Chinese Control and Decision Conference (CCDC 2008) 1575