J Intell Robot Syst DOI 10.1007/s10846-013-9820-z
Development of Wall Climbing Robot System by Using Impeller Type Adhesion Mechanism Ig Mo Koo · Tran Duc Trong · Yoon Haeng Lee · Hyungpil Moon · Jachoon Koo · Sun Kyu Park · Hyouk Ryeol Choi
Received: 17 June 2012 / Accepted: 31 January 2013 © Springer Science+Business Media Dordrecht 2013
Abstract In this paper, we present a wall climbing robot system, called “LARVA”, developed for visual inspection of structures with flat surfaces. The robot has two differential driving wheels with a suspension and an adhesion mechanism. The adhesion mechanism is composed of an impeller and two–layered suction seals. It is designed to provide sufficient adhesion force and be controlled
so that the robot can move freely on various wall surfaces. The static and aerodynamic modeling of the adhesion mechanism is given and the analysis of the adhesion mechanism, air leakage, and inner flow are carried out to be useful for the design as well as the control. Finally, the performances of the robot are experimentally verified on several kinds of walls and its feasibility is validated. Keywords Impeller · Adhesion mechanism · Suspension mechanism · Wall climbing robot
Electronic supplementary material The online version of this article (doi:10.1007/s10846-013-9820-z) contains supplementary material, which is available to authorized users. This work (research) is financially supported by the Ministry of Knowledge Economy (MKE) and Korea Institute for Advancement in Technology (KIAT) through the Workforce Development Program in Strategic Technology.
1 Introduction Recently, automatic inspection technologies become keen interests in robotic applications. It is widely used in various social infrastructures
I. M. Koo · T. D. Trong · Y. H. Lee · H. Moon · J. Koo · H. R. Choi (B) School of Mechanical Engineering, Sungkyunkwan University, Chonch’on-dong, Jangan–gu, Suwon, Kyonggi–do, Korea e-mail:
[email protected]
Y. H. Lee e-mail:
[email protected]
I. M. Koo e-mail:
[email protected]
S. K. Park School of Civil Engineering, Sungkyunkwan University, Chonch’on-dong, Jangan–gu, Suwon, Kyonggi–do, Korea e-mail:
[email protected]
T. D. Trong e-mail:
[email protected]
H. Moon e-mail:
[email protected] J. Koo e-mail:
[email protected]
J Intell Robot Syst
such as buildings, bridges, nuclear power plants, and marine structures, which play very much important roles in our daily lives. Because of the enormous damages caused by their troubles, periodic maintenances to ensure their safety and integrity are necessary. In addition, since most of inspection tasks are performed in the environments hard to be accessed by human workers, we need automated inspection robots which can provide cost–effective and safe alternative to the maintenance works. Conventional wall climbing robots are classified into four groups depending on the adhesion methods such as magnetic, hand–hold, biologically inspired and vacuum suction pad. Apparently, magnetic and hand–hold ones can provide very strong adhesion forces but they are useful just in special environments with ferromagnetic surfaces and knobs [1, 2]. In contrast, a biologically inspired method is applicable to various kinds of surfaces and no additional means is necessary for creating the adhesion force [3]. However, comparatively, the payload of the biologically inspired method is very low. The vacuum suction method is very simple and easy to control but it may have problems depending on the surface conditions like cracked and rough surfaces [4–7]. In order to design a wall climbing robot, there are several issues to be considered. The first one is the size of the robot because it is limited by the geometry of man-made structures, for example, girders, crossbeams, poles, etc. In the second, wall climbing robots need to carry inspection equipment which usually weigh more than 4 Kg, such as portable ultrasonic inspection device(e.g. ECHOGRAPH 1090 BASIC, 2 Kg, KarlDeutsch Co. in Germany) and portable x–ray image scope(e.g. RayzorXpro, 3.5 Kg, VIDISCO Ltd.) among the nondestructive testing equipment. Because of the limitation of the robot size, the payload is quite large for the climbing robot. Third, the robot must be able to climb over vertical walls as well as maneuver on horizontal ceilings because most of man–made structures are composed of vertical and horizontal surfaces. From here onward, the horizontal surface means the horizontal climbing surface, in other words ceiling. Mechanics of these
two surfaces are very different, and thus, the robot needs to meet the requirements of both of them. Finally, the robot must be able to maneuver on uneven surfaces such as surfaces with bolts, bumps, grooves, or thresholds, which may cause the loss of adhesion force. Considering the aforementioned issues, we focus on a specific type of wall climbing robots using the vacuum suction method because the method can generate sufficient adhesion forces compared to its size and weight. In this work, we manufactured two wall climbing robots, called LARVA series, that is LARVA–I, and II. Figure 1 displays LARVA–II. Both of them are similar mechanically except minor structural differences, that is, the location of the center of gravity, size, weight, and etc. They use an impeller in the adhesion mechanism and two wheels as a locomotive driving system which carries huge payload as well as mobility in high speed. The AC power is supplied from the outside. The locomotion mechanism of the proposed robot is different from [8–10] in the sense that the two suspended driving wheels are placed outside of the vacuum chamber and front–side of the robot body. This configuration allows the driving wheels to make contact with locally concave surfaces such as bolts and nuts on the wall before the sealing pad. The mechanism prevents the robot body from being detached from the surface by
Fig. 1 Wall climbing robot LARVA–II
J Intell Robot Syst
allowing the driving wheels rolling over such uneven surfaces so that the vacuum sealing is maintained properly, thus adhesion is maintained. As a result, the proposed robot can avoid the collision with higher threshold and accommodate other bumps better than the previous works [8–10]. This paper is organized as follows. In Section 2, the mechanical analysis and the dynamic system modeling of the adhesion mechanism are introduced. Section 3 discusses the wall climbing robotic system. Including the hardware platform, the detailed description on every mechanism and the control system of the robot are given. In the next, for evaluating the performance of the robot, preliminary experiments are performed and the results are shown in Section 4. Finally, concluding remarks are given and future works are discussed in Section 5. Note: pressure type that mentioned in this paper is based gauge pressure. 2 Modeling and Analysis 2.1 Static Analysis of Adhesion Mechanism The adhesion mechanism consists of a vacuum chamber and an impeller that is used to remove the air from the chamber. This mechanism can create two kinds of forces: one is the thrust force generated by the outside flow. The other is the force produced by the pressure gradient between the ambient and the vacuum chamber. The force by the pressure gradient is much stronger than the thrust force in most cases. The environments that the robot is expected to move on are mainly classified into two basic types, that is the horizontal surfaces and the vertical ones. Major difference between two cases is the effect of the gravity force. In the first, we derive the static equilibrium equation in the horizontal surface, and then, the conditions required for the adhesion force are derived. As shown in Fig. 2, the equilibrium of forces in the robot moving on the horizontal surface can be expressed as follows. (1) Fz = PA − mrobot g = 0
Fig. 2 Static force analysis in horizontal plane
where Fz denotes the z–directional force. P and A are the pressure difference between the pressure inside the robot and atmospheric pressure and the area of vacuum chamber, respectively. g represents the gravity constant. The other variables in Fig. 2 are defined as –
– – – –
FVacuum : adhesive force by air pressure difference between the inside the chamber and atmospheric pressure FThrust : thrust force by the air flow between the impeller and that blade FTraction : tractive force by the wheel FGravity : gravity force FFriction : frictional force between the robot and the surface
If the weight of the robot is larger than the sum of the thrust force and adhesion one, the robot falls down. In this analysis, two important points needs to be noted. Firstly, the influence of pressure and gravity is the most significant factor for the robot to be attached to the surfaces. In fact, the thrust force FThrust is relatively smaller than the adhesive force FVacuum . Therefore, considering that the thrust force is negligible while the robot is attached to the horizontal surfaces, the pressure level for adhesion is derived as follows. P ≥ mrobot g/A
(2)
J Intell Robot Syst
The pressure P obtained from Eq. 2 gives the reference pressure to be needed while moving on the horizontal plane. In the vertical surface, we need to consider the friction force. As depicted in Fig. 3, the static equilibrium of the robot moving on the vertical surface is expressed as (3) Fz = FVacuum + FThrust = N
F y = FTraction − FFriction − FGravity = 0
(4)
and Eq. 4 yields FTraction = FFriction + FGravity
(8)
It is noted that the friction force in Eq. 8 acts in the opposite direction of the robot motion. And we assume that the robot is moving up and the gravitational force is acting along −y direction. Substituting Eq. 7 into Eq. 8 leads us to the inequality condition for keeping adhesion as follows. hFGravity ≥ FGravity , (9) μ (FThrust + FVacuum ) − r
M = −hFGravity + r(FThrust + FVacuum ) = 0 (5)
where we have FFriction = μN
(6)
and mrobot is the weight of the robot, and h is the distance between the center of mass and the surface. r represents the radius of the robot body and N denotes the normal force along the surface. The sign of the gravity force FGravity is determined depending upon the direction of moving. From Eq. 5, we can have hFGravity = r(FThrust + FVacuum )
(7)
where μ is the friction coefficient between the robot and the surface. In this paper, we use Coulomb friction model to compute the required thrust and adhesion force. To calculate the critical value of the friction force, we search for the point where the total moment is zero. If the friction force is less than the weight of the robot, it will slip down. In addition, if the normal force FThrust + FVacuum is less than zero, the robot will fall down. Based on this analysis, we can know the relation between thrust and adhesion force in the design process so that the wall climbing robot keeps adhesion stably. That is, the robot’s center of gravity, weight, and the radius of vacuum area (it determines the size of vacuum chamber and thus that of the robot) need to be considered in the design stage and this analysis helps us determine the design parameters. 2.2 Analysis and dynamic modeling of aerodynamic system 2.2.1 Air Leakage Analysis
Fig. 3 Static analysis in vertical plane
When the adhesion mechanism moves on the wall surface, air leakage occurs due to the gap between the mechanism and the wall surface. The air leakage and the inside pressure have the relation of the inverse proportional according to the gap distance. It produces the biggest influence on the adhesive force of the system and thus, the fluid leakage analysis is essential for the design of the adhesion mechanism.
J Intell Robot Syst Fig. 4 The ideal model of fluid system
(a) Flow direction
In this subsection, we use Navier-Stokes equation to analyze the air leakage flow under the assumption that the air flow moves along the fixed direction as shown in Fig. 4a. By introducing the cylindrical coordinate the continuity equation for the flow around with the sealing pad can be written by ∂uz 1 ∂ 1 ∂uθ (rur ) + + =0 r ∂r r ∂θ ∂z
(10)
Furthermore, the equation can be rearranged by using uθ = uz = 0 such as ∂ur ur + =0 r ∂r
(11)
Substituting into the Navier-Stokes equations, we obtain vr2 1 ∂ P μ∂ 2 vr + − = − ρ ∂r ρ∂z2 r
(12)
In Eq. 12, the internal pressure Pi is calculated by using the boundary conditions of non–slip condition and zero flow velocity with the radius of the sealing pad r as follows. Pi = Po −
6μQ ln(r2 /r1 ) ρπ h3
(13)
Therefore, by using Eq. 13 the air leakage flux (Q) of the adhesion mechanism system can be derived as follows [11]. Q=
π h3 ρ(Pi − Po ) 6μ ln(r1 /r2 )
(14)
where, – Q - leakage flow rate – h - distance between the sealing and the surface
(b) Leakage flow model
– ρ - average pressure such as (Pi -Po )/2 – r1 - outer radius of the sealing pad – r2 - inner radius of the sealing pad
2.2.2 Inner Flow Analysis When the air moves from the wide tube to the narrow one, the energy is dissipated due to the internal friction resistance of the pipe, which influences the adhesion mechanism. Because dissipating energy in the adhesion mechanism means decrease in the adhesive power, the analysis for internal friction resistance should be performed. Assuming that z1 = z2 from Fig. 5, the Bernoulli Equation can be expressed as ρ u22 − u21 + ζ u22 P1 − P2 = 2
Fig. 5 Inner flow model
(15)
A B u1 u2 p1 Q p2 A2 A1
J Intell Robot Syst
Here, for the same flow rate of between A and B from Fig. 5, each term representing the resistance values can be expressed as R1 =
Q Q ζQ , R2 = , Rζ = 2A21 2A22 2A22
(16)
The coefficient of air compressibility in the model is derived as Eq. 20, where C, β, and V mean the capacitance of the air, the bulk modulus of the air and the volume of the vacuum chamber, respectively. V , β
By using Eqs. 15 and 16 the following equation can be calculated:
C=
P1 − P2 = ρ Q(R2 − R1 + Rζ )
With these components, the air system is modeled as shown in Fig. 6. Equation 21 shows the relation of the pressure and the flow in 0–junctions and 1–junction.
(17)
by naming the (R2 − R1 + Rζ ) = R12 and using Eq. 16 we can find the R12 , internal friction resistance, to be equal to: R12
(1 + ζ )Q Q = − 2 2A2 2A21
(18)
where ζ is the internal friction resistance coefficient. Therefore, we can conclude that the smaller the internal friction resistance R12 becomes, the smaller the flow difference between the area A1 and A2 is generated. By applying this analysis, we can develop the adhesion mechanism with high efficiency. 2.2.3 Modeling of Fluid System In most cases, the force by pressure gradient is much larger than the thrust force. Thus, dynamic modeling of the fluid system is required to find the relation between the pressure difference and its generator. In the first, we can simplify the complex fluid system as shown in Fig. 4. By introducing the bond graph method each mechanical component of the fluid system is translated with the bond graph component. In the first, R O means the resistance caused by the leakage flow. Thus, R O is written by p 6μ r1 (19) RO = = ln ρQ ρπ h3 r2
P˙1 = − P˙2 =
(20)
1 P1 − C2 R12
1 1 + P2 C2 Rr C2 R12
1 + P(t) C2 Rr
(21)
where – C1 - capacitance of the chamber – C2 - capacitance of the narrow circle – P1 - pressure at the chamber – P2 - pressure after the throat – R12 - resistance caused by the throat – Rr - resistance caused by the impeller – P(t) - evacuating pressure by the vacuum motor. In Eq. 21, P1 is the most important factor that dominantly affects the overall adhesion force. By
R12
C1 e f
where – r1 - outer radius of the sealing pad – r2 - inner radius of the sealing pad – h - distance between the sealing and the surface
1 1 1 + P1 + P2 C1 R0 C2 R12 C1 R12
2,
Ro
e f 1,
1
0 e P1
3,
C2
e f 4,
2
f
1 3
e f 6,
4
e f 5,
Rr
5
0 e P2
e f 8,
6
7,
f
1 7
Fig. 6 Bond graph of the adhesion system
8
e f 9,
P(t) 9
J Intell Robot Syst 0
Adhesion pressure (kpa)
- 0.1 - 0.2 - 0.3 - 0.4 - 0.5 - 0.6 - 0.7 - 0.8
Fig. 8 Wall climbing robot LARVA–I
- 0.9 0
1
2
3
4
5 6 Time (s)
7
8
9
10
Fig. 7 Step response of fluid dynamic model
transforming the function using Cramer’s rule, we obtain P1 (s) A´ = ´ + C´ P(s) s2 + Bs 1 Rr + R12 + R O A´ = C´ = C1 C2 Rr R12 C1 C2 R O Rr R12 C1 R O R12 + C1 R O Rr + C2 R O Rr + C2 Rr R12 B´ = C1 C2 R O Rr R12 (22) Equation 22 shows that the adhesion mechanism behaves as a second order system. The results of the simulation for the step input is depicted in Fig. 7. In this figure, 1 to 6 s interval shows the transient response and, after that time (settling time) the robot reaches to the steady state and adheres stably on the surface. In the next section, the control algorithm will be proposed based on this modeling.
LARVA’s power is supplied via AC 220 V tethered power cable and all the other components are embedded in the system. The outline of the overall system is described in Fig. 9. To move on vertical and horizontal surfaces, the robot needs the three BLDC motors, two of them for driving the wheels and the other for operating the vacuum system. Strong adhesion force can be produced by maintaining high vacuum level. However, excessive vacuum can cause the robot to be stuck because the wheel friction can not overcome the frictional force by the vacuum pad. Thus, a negative pressure sensor is used to measures the vacuum level, and the state of the vacuum is feedback to the system controller via analog signal. Also, for wall–inspection and obstacle detection, a
3 Wall Climbing Robot: LARVA 3.1 Outline of LARVA LARVA I illustrated in Fig. 8, has the total weight of 3.3 Kg with the width of 30 cm and height of 11 cm. The maximum pressure that can be created in the chamber is about 5 kPa which approximate to 300 N roughly. It can move on the wall at the maximum speed of 12 cm/s.
Fig. 9 Schematic view of whole system configuration
J Intell Robot Syst
vision camera is mounted. To get feedbacks of the robot’s posture, a gyroscope is used in the system. These sensors help the system to move freely on wall and ceiling and to avoid obstacles. The robot is controlled by three MCUs (Micro Controller Unit) and a SBC (Single Board Computer) using RTOS (Real Time Operation System) Linux system. The 8 bit MCUs control the motors and the feedback sensors. Two of them handle two driving motors, and theirs encoders. The other drives the vacuum motor, and gets the feedback data from the vacuum sensor and the gyroscope with SBC, controls the whole system. CAN (Control Area Network) is used for communication between these controllers. By wireless communication, SBC exchanges data and command with the host. We have developed two LARVAs’, that is LARVA–I, and LARVA–II. Two prototypes have minor differences in the mechanism and LARVA–II is a little bit heavier than the LARVA–I. The specifications of LARVA series are compared in Table 1. The different points between LARVA–I and II are impeller motor specification, CoM position, and vacuum chamber’s capacity, as shown in Table 1. The position of the wheel for LARVA–II is located the backside than LARVA–I’s wheel position. Namely, it means the CoM position of the LARVAR–II is more near to center of robot than LAVAR–I. In addition, the maximum torque of the motor for LAVAR–II was increased than LAVAR–I, and total weight was also reduced. It provided a help to improve the propulsive and adhesion force of the robot about adhesion surface.
locomotion system. Mechanical parts of LARVA are explained, respectively in this subsection. 3.2.1 Adhesion Mechanism The adhesion mechanism is the impellent force generator which consists of a main motor with an impeller and an evacuation cover to take the air out as shown in the Fig. 10. The motor drives gears which transfer its spinning torque to the impeller. In this case, the driving system has the transmission ratio of 1:2. The high speed rotation of the impeller makes the air evacuated from the inside of the vacuum chamber completely. If the vacuum chamber is sealed adequately, the robot can create the negative pressure, or partial vacuum of the chamber. With the cover in Fig. 10, the air flows is guided to the outside of robot, actually helping to increase the adhesion force by thrusting the suction forward. Figure 11 shows the inside pressure of the real impellent force generator mechanism by using 60 % of the maximum output of the impeller driving motor (at 8420 rpm). In this figure, there is a sudden drop of pressure between 30 to 40 s. The reason for this drop is the deformation of the impellent force generator mechanism. The material of the mechanism is made of acetal (engineering plastic), and it is very thin in thickness as about
BLDC Motor Evacuation Air Flow
3.2 Mechanism The wall climbing robot is mainly composed of three mechanical parts that is, an impellent force generator, a sealing suction for adhesion and a
Cover Impeller
Table 1 Series of LARVA specification Dimension Weight COM Static Payload
LARVA-I
LARVA-II
30 cm(φ)×11 cm(H) 3.3 kg Center – 5cm 7 kg
30 cm(φ)×12 cm(H) 3.2 kg Center 10 kg
Inhalation Air Flow Fig. 10 Exploded view of the impellent force generator mechanism
J Intell Robot Syst Fig. 11 Inside pressure data of the real impellent force generator mechanism
Pressure data @ 8420 motor rpm
18 16
Pressure ( X 100 Pa)
14 12 10 8 6 4 2 0 0
10
20
30
40
50
60
Time (s)
2.5 mm. Thus, when the high inner vacuum pressure is generated, this mechanism experiences a little bending, which causes the air leakage. But, this phenomenon appears when tested only with the force generator mechanism, not in the case of the whole system assembled, and it is cleared out in the steady state of inner vacuum level. 3.2.2 Sealing Mechanism The sealing mechanism needs to keep the negative pressure inside the chamber by preventing the air leakage. In this work, two sealing layers are applied, that is the straight inner layer and the
flexible bending layer as shown in Fig. 12. The straight inner layer surrounds the center of suction to keep an adequate sealing. It is more efficient than the other types like bristle, or sponge etc. Since it is only efficient at flat surface, the supplementary layer is added, named as the flexible bending layer. The flexible bending layer prevents the unexpected air leakage of suction by keeping contact with irregular surfaces. As shown in Fig. 13, the contact is naturally maintained because of the pressure difference between the inside and outside of the sealing layer. Furthermore, the bent shape of the flexible bending layer prevents the layer itself from being rolled up by the friction force. The double layered sealing mechanism generally reduces the air
Bending Layer
Fig. 12 Double layered sealing mechanism
Fig. 13 Cross sectional view of the flexible bending layer
J Intell Robot Syst Fig. 14 Simulation of sealing mechanism
(a) Single layered sealing pad
(b) Double layered sealing pad
leakage of the chamber more than the single layered one and thus, the negative pressure by the impellent force generator can be maintained better while the robot moves on the various surfaces. Figure 14 displays the results of analysis for the sealing efficiency of single and double layered sealing mechanisms by using the software called Cosmos Flowwork. The height between sealing and wall is set by 0.1 mm. The other initial conditions are the same as the adhesion mechanism of real robot. The left color bar represents the pressure level of the chamber. The blue color means the lowest level and the red denotes the highest one. From the result of this simulation, it can be noted that the double layered sealing pad has better sealing efficiency, and thus, it can preserve the higher pressure difference than that of the single layered one by keeping the air from leaking out of the chamber.
drive wheels through miter gears, are arranged in parallel. The direction of robot is controlled by differential driving.
3.2.3 Locomotion Mechanism As shown in Fig. 15, the locomotion mechanism is composed of motors, wheels, gear boxes, and shock absorbers and frames. Two motors, which
Fig. 15 Locomotion mechanism
J Intell Robot Syst Fig. 16 Block diagram of adhesion force control Input +
Impller System
fmotor (s)
fimpeller (s)
PID Controller
-
Disturbance
BLDC Motor & Amplifier
Interger Data
Vacuum Chamber
fchamber (s)
OUTPUT (kpa)
fsensor (s) Pressure snesor
In addition, the suspension device is also added to the locomotion mechanism. The suspension mechanism makes the wheel’s height follow the roughness of the wall surface or the curve of surface such as bridge poles. The contact between surface and the robot, especially sealing layer and wheels, is maintained with this mechanism. Thus, the robot keeps its adhesion force and the thrust
Fig. 17 System response graph of pressure control
of the wheel. Also, the robot uses a ball castor for balancing itself and the pitch of ball castor is adjusted by springs as shown in Fig. 15. 3.3 Control Strategy of Adhesion Mechanism In wall–climbing robots, maintaining a proper adhesion force is the most important, but excessive
1 Measured pressure
Adhesion Pressure (Kpa)
0.5
Input signal
0 -0.5 -1 -1.5 -2 -2.5 -3 -3.5 -4
4
5
6
7
8
9
10
11
12
13 14
15 16
17 18
19
20
21 22
23
24
25
80
85
90
Times (s)
(a) The response of adhesion mechanism on step input
Adhesion pressure (kpa)
0
-1
-2 -3
Disturbances -4
-5
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
Time (s)
(b) System recovery on sudden disturbances
75
J Intell Robot Syst Fig. 18 Experiments of adhesion force with static payload: LARVA–I
(a) Horizontal surface
adhesion force makes the robot stuck. Thus, to move the robot on the wall and ceilings without falling down, a suitable control algorithm for the vacuum system is required. By using dynamic system analysis, the adhesion mechanism is modeled as a SISO (Single Input Single Output). The input variable is the desired vacuum value of system, and disturbance is the factor corresponding to the drop of the vacuum level, that is, the irregularity of wall surface, crack on the surface, unexpected external forces etc. Considering these variables, a simple PID controller is developed as shown in Fig. 16. As the control strategy we applied a simple PID controller. The reference state is given as the pressure level of the vacuum chamber, which is measured in realtime by using a pressure sensor. The impeller velocity is controlled to achieve the goal pressure level. The functions of the each block in Fig 15 are explained as follows: input is the inside goal pressure of the vacuum chamber; PID controller and BLDC motor & amplifier is a sub–controller(motor driver), and it performs the control of the motor rotational speed based on the input value; impeller system is the impellent force generator, and it generates a vacuum pressure by the rotation of the motor; vacuum chamber is the pressure of vacuum chamber inside. Finally, sensor(ft) is a pressure sensor to measure the vacuum pressure inside the chamber. To test its performances, several experiments have been performed. The robot was set on the vertical surface without payload and the adhesion performance was tested. Figure 17a shows the sys-
(b) Vertical surface
tem response on the step input signal. It is noted that the adhesion mechanism can be analyzed as the 2nd–order system with time delay and the overshoot. This is well matched with Eq. 22, and its simulation results of Fig. 7. Furthermore, Fig. 17b shows that the control system is capable to overcome the disturbances. The vacuum level is dropped to 1 kPa, but that is not critical drop of vacuum to lose the adhesion force.
4 Experiments Several experiments were conducted to evaluate the performances of the robot. Before driving the robot, we tested the performance of the adhesion method on vertical and horizontal surfaces as show in Fig. 18. The robot was attached to the wall
Fig. 19 Experiment of curved surface adhesion: LARVA–I
J Intell Robot Syst
1
2
3
4
5
6
1
2
3
4
5
6
Fig. 20 Experiment of horizontal surface driving: LARVA–I
Fig. 21 Experiment on the vertical surface
J Intell Robot Syst
Fig. 22 Experiment of uneven (8 mm bump) horizontal surface
when the robot went over bumps. Bouncing occurred twice and the negative pressure increased instantly. This phenomena can be explained as the effect of sealing layer, especially the flexible bending layer. For a moment, the body of the robot was raised up while the robot attempts to overcome the obstacle and the flexible layer got deformed due to the effect of pressure gap. As the result, the leakage was not so much, but the volume of chamber got increase temporarily. The
-2 -2.5 -3
Pressure (kpa)
with 7 kg payload at 5 KPa of vacuum pressure in static condition. As shown in Fig. 19, the robot was successfully attached to a curved surface by producing sufficient adhesion force. The scenes of experiments on the horizontal flat surfaces are displayed in Fig. 20. Figure 21 illustrates the result of another test on a vertical surface. The robot climbed up the wall, changed the direction of motion and rotated. Also, the robot successfully ran over various surfaces with different surface characteristics, such as glass, concrete, and etc. Figure 22 displays the performance of locomotion mechanism on an obstacle with 8 mm height. Without the suspension mechanism, 8 mm of bouncing would have caused the failure of adhesion because of air leakage. However, the proposed mechanism enabled the robot to easily overcome the issue as shown in Fig. 22. We can see the sealing layer keeps the contact with the surface. As mentioned, the height of the obstacle to be accommodated highly depends on the capability of suspension. Figure 23 shows the pressure change inside the sealing mechanism measured during experiments. Two peaks indicate the sudden changes of pressure inside the chamber
-3.5 -4
Bouning again
Bouning
-4.5 -5 -5.5 -6 2
3
4
5
6
7
8
9
Time (s) Fig. 23 Pressure during the experiment in Fig. 22
10
11
J Intell Robot Syst Fig. 24 Pressure and motor speed during the experiments in Fig. 25
pressure dropped at that moment and the change of inside pressure due to the obstacle could be compensated. Finally, the robot was driven in the column of the real bridge. The surface of the bridge was concrete and thus, very much irregular with a lot of gaps, cracks and holes. Figure 24 displays the inside pressure measured and the scenes of experiments are given in Fig. 25. According to the experimental results, the proposed robot could run over the concrete wall successfully.
5 Conclusions In this paper, we presented a wall climbing robot system, which can move in vertical or horizontal flat and rough surfaces. The robot mechanism was designed and it was analyzed by building up a dynamic model. Two LARVA prototypes, that is LARVA I, and II were manufactured and their performances were validated experimentally in the lab environments as well as the real one.
Fig. 25 Experiments on the real bridge: LARVA–II
(a) Horizontal surface
(b) Vertical surface
J Intell Robot Syst
In fact, if we have excessive suction pressure, which makes the wheels slip, it is not easy to go straight even though the wheel velocities are controlled equally. This is an important issue to follow the reference trajectories, because we have to control the suction pressure at the optimum level to avoid the slip of wheels, while modulate the wheel velocities by using another sensor such as the IMU sensor etc. Also, it is influenced by the surface conditions, and we are working on it as the future works. References 1. Longo, D., Muscato, G.: SCID—a non-actuated robot for walls exploration. In: in Procceding of IEEE/ASME International Conference on Advanced Intelligent Mechatronics, vol. 2, pp. 874–879 (2001) 2. Bretl, T.: Multi-step motion planning: application to free–climbing robots. PhD’s thesis, University of Stanford, CA (2005) 3. Kim, S., Spenko, M., Trujillo, S., Heyneman, B., Mattoli, V., Cutkosky, M.R.: Whole body adhesion: hierarchical, directional and distributed control of adhesive forces for a climbing robot. In: Procceding of IEEE International Conference on Robotics and Automation, pp. 1268–1273 (2007) 4. Song, Y.K., Lee, C.M., Koo, I.M., Tran, D.T., Moon, H., Choi, H.R.: Development of wall climbing robotic
5.
6.
7.
8.
9.
10.
11.
system for inspection purpose. In: Procceding on IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 1990–1995 (2008) Ota, Y., Kuga, T., Yoneda, K.: Deformation compensation for continuous force control of a wall climbing quadruped with reduced-DOF. In: Proccding of IEEE International Conference on Robotics and Automation, pp. 468–474 (2006) Nagakubo, A., Hirose, S.: Walking and running of the Quadruped wall–climbing robot. In: Procceding of IEEE International Conference on Robotics and Automation, vol. 2, pp. 1005–1012 (1994) Lal, R., Tummala, Mukherjee, R., Ning, X., Aslam, D., Dulimarta, H., Jizhong, X., Minor, M., Dang, G.: Climbing the walls [Robots]. IEEE Robot. Autom. Mag. 9(4), 10–19 (2002) Longo, D., Muscato, G.: The Alicia/sup 3/ climbing robot: a three-module robot for automatic wall inspection. IEEE Robot. Autom. Mag. 13(1), 42–50 (2006) Xiao, J., Sadegh, A., Elliott, M., Calle, A., Persad, A., Chiu, H.M.: Design of mobile robots with wall climbing capability. In: Procceding of IEEE/ASME International Conference on Advanced Intelligent Mechatronics, pp. 438–443 (2005) Wu, S., Li, M., Xiao, S., Li, Y.: A wireless distributed wall climbing robotic system for reconnaissance purpose. In: Procceding of IEEE International Conference on Mechatronics and Automation, pp. 1308–1312 (2006) Zhi–yuan, Q., Yan–zheng, Z., Zhuang, F., Ya, W.: Fluid model of sliding suction cup of wall–climbing robots. Int. J. Adv. Robot. Syst. 3(3), 275–284 (2006)