Robotic system with active steering capability for internal ... - CiteSeerX

Mechatronics 12 (2002) 713–736

Robotic system with active steering capability for internal inspection of urban gas pipelines H.R. Choi a

a,*

, S.M. Ryew

b

School of Mechanical Engineering, Sungkyunkwan University, 300 Chunchun-dong, Jangan-gu, Suwon, Kyonggi-do 440-746, South Korea b Department of Machine Design, Sungkyunkwan University, 300 Chunchun-dong, Jangan-gu, Suwon, Kyonggi-do 440-746, South Korea Received 14 April 2000; accepted 2 February 2001

Abstract In this paper, we present a robotic system for inpipe inspection of underground urban gas pipelines. This robot is developed on the purpose of being utilized as a mobile platform for visual and Non-Destructive Testing (NDT) of the pipeline networks. The robot is configured as an articulated structure-like a snake with a tether cable. Two active driving vehicles are located in front and rear of the system, respectively. Passive modules such as a control module and other optional modules are linked between the active vehicles. It has several characteristic features superior to the others such as flexible wheeled leg mechanisms, a steering mechanism with compliance control. Especially the steering mechanism called Double Active Universal Joint (DAUJ) intrinsically prevents rolling of the robot along the driving direction and enables to control its compliance. Those features provide the robot with excellent mobility inside the highly constrained space while negotiating the complicated configurations of the pipeline networks. We outline the construction of the robot and describe its characteristic features. Ó 2002 Elsevier Science Ltd. All rights reserved.

1. Introduction Recently natural gas supply becomes one of the fundamental public services and its impact on the urban infrastructure is getting larger. The urban gas pipelines, as they are buried under the ground, are prone to external corrosion usually derived by

*

Corresponding author. Tel.: +82-31-290-7449; fax: +82-31-290-7507. E-mail addresses: [email protected] (H.R. Choi), [email protected] (S.M. Ryew).

0957-4158/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 9 5 7 - 4 1 5 8 ( 0 1 ) 0 0 0 2 2 - 8

714

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

moisture and chemical agent in soil, which causes material losses of the pipe wall. Also, cracks in the welded region and the damages from third parties such as construction, electricity, sewage works are considered as one of the major reasons for pipeline failures. In the inspection of urban gas pipelines, there are a lot of needs for autonomous inspection equipment that can run through inside the pipelines. However, the inpipe inspection of gas pipelines in field conditions has several aspects in difficulty which are as follows: (1) The urban gas pipelines usually allow restricted access to the test location because they are buried under the ground. (2) Current urban gas pipelines take quite complicated configurations. Straight pipes and welded joints are the most popular ones, and there are also lots of elbows, branches, various valves and other special components. Different from plain surfaces, inside of the pipelines is a high constrained space with complicated configuration and thus it not easy to move inside while overcoming intrinsic obstacles in the pipelines as well. (3) Most of urban gas pipelines are composed of those with small inner diameter which makes it impossible to carry the inspection equipment inside the pipelines. In the last the urban gas pipelines have abnormal pipelines such as gouges and dents obstructing normal activities of maintenance due to the frequent damages from third parties. On the contrary various types of robotic systems recently developed though they begin to be used for servicing inpipe applications, do not cope with given situations successfully [1,2]. Up to date technology allows robots to be built small enough to go through pipelines with as small as 4 in. diameter although there exist much smaller ones [3] that is not adequate for general usages. Most of those systems have robots with drive wheels pressed passively against the wall of the pipe by springs and linkages. One typical approach is scissor-like structure with three wheels, one at the joint and the others at the end of the two limbs [4]. This robot called MOGRER, has already commercialized aimed at the application in the gas industry [5]. Also, Fujiwara et al. [6], and Taguchi and Kawarazaki [7] addressed similar approaches. Kawaguchi et al. [8] developed a mobile robot with magnetic wheels which has special features in the steering. As a system adopting a little different type of robots, Ilg et al addressed an autonomous sewer inspection system. This system uses an articulated mobile robot with three-degree of freedom (d.o.f.) active joints at each joints [9]. In some cases they utilize specially designed walking mechanisms [10] or hydrodynamic forces [11]. Recently, Hirose et al. presented a summary of their works concerning various inpipe inspection robots developed themselves up to now [12]. As the commercialized ones, Visual Inspection Technologies [13] produces several pipe–crawling inspection robots with visual inspection capability. To employ the robots in the inspection of the urban gas pipelines they need to satisfy following conditions. First, they should have as much flexibility as to change their body depending on the shape, size, and the configuration of the pipelines. In the second the active steering capability is prerequisite. In case of the urban gas pipelines there exist a lot of branches that is three-dimensional, not in the plane. Thus, it is not only necessary to steer the robot along the desired direction but also difficult to realize the steering. It is a good challenge with respect to the design of the robot. In the last, the robot should have sufficient tractive force to pull the tether cable, climb

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

715

the vertical pipelines and go over obstacles. The robots announced up to date, however, do not seem to show enough capability to overcome those problems. Especially it is hard to find the system to have the capability of selective navigation in branch though it is prerequisite in urban gas pipelines. The robot to be presented in this paper has been designed based on those requirements and is able to conquer the problems with its flexible structure and an active steering mechanism. The robot is configured as an articulated structure-like a snake with a tether cable. Two active driving vehicles are located in front and rear of the system, respectively which provide sufficient trust. Passive modules such as a control module and other optional modules are linked between the active vehicles. The new steering mechanism employed by the proposed robot is called double active universal joint (DAUJ), which provides omni-directional steering capability enabling to negotiate with complicated configurations of pipeline networks. In fact the robot has been designed as a mobile platform for visual and Non-Destructive Testing (NDT) of the pipeline networks. In this report, however we address the robot focusing on its mechanism. Characteristic features of the robot are described and its construction is outlined. This paper is organized as follows: In Section 2 and 3, design requirements and the proposed system are overviewed, respectively. Section 4 describes the issues of mechanism including the driving vehicle, wheeled leg mechanism, and steering mechanism. In section 5 we explain the controller architecture with the graphical user-interface, respectively and section 6 describes the performance evaluation of the proposed system. Then, finally we will conclude with summary in Section 7.

2. Requirements of design Often, the shape and size of the robot are the most critical factors in determining maneuverability, which depend on the pipeline configuration. The urban gas pipelines basically consists of straight pipelines running horizontally and vertically. There are also elbows, branches, reducer, valves with unexpected mechanical damages such as dents, gouges, removed metals caused by third-parties, which are not reflected in the layout drawing and demands highly flexible design of robots. The presented robot is for the inpipe inspection of the urban gas pipelines with the diameter of 8 in. that is the most popular size in our country, but the methodologies adopted in this work is not limited to. Based on the considerations of the pipeline configuration, the requirements of design can be derived as follows: (1) active steering capability in branches, (2) surmounting right angle elbow with curvature of smaller than inverse of 1.5 times of its nominal diameter, (3) driving through pipelines with a diameter of 160–240 mm (nominally 8 in.
20%), (4) sufficient traction forces (which is assumed to be more than 40 kg of vertical load excluding self-weight), (5) minimum 500 m of running distance/per launching and faster speed than 3 m/s in horizontal pipelines. Items from (1) to (3) provide the fundamental requirements for the mobility that the robot

716

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

should have and prerequisite to negotiate through wide range of configurations of pipelines. Also, (4) and (5) describes the supplementary capabilities, which is necessary to carry and perform useful tasks by using appropriate inspection tools such as CCD cameras, NDT unit. Major design issues of the robotic system correspond to how to enhance mobility inside pipelines. The design of the system mainly depends on the present state of the art in technology as well as the requirements of the system. The configuration of pipelines restricts the whole size of the robot and the current technology determines the possibility of implementation because actuator, drive electronics, embedded controller, power supply, sensor, and communication tools would have to be placed in an extremely small space. From the present technological point of view, therefore, only a very large robot in size is possible. One reasonable solution to this problem is the use of articulated structure such as snake-like or multi-joint robots though the control of the robot gets more difficult [14]. Fig. 1 describes a possible configuration of the robot that is composed of functionally partitioned modules such as driving

Fig. 1. Schematic of proposed system.

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

717

modules, control modules, sensor modules etc., and the proposed system has been designed accordingly.

3. System overview In this section we will briefly mention how the problems in the urban gas pipelines are solved with the proposed robot. Details of the robot are explained in the following sections. As illustrated in Fig. 1, the proposed robot consists of articulated bodies including driving vehicles, control module, tether cable and ground station. The instrumentation module for NDT can be attached on the robot optionally. Fig. 2 depicts the proposed robot inside a pipeline simulating facility (which has been newly constructed to evaluated the performance of the proposed robotic system and will be described in the section 6). Basically the robot is designed to have enough traction forces to climb the vertical pipelines or pull the tether cables, which are provided by two driving vehicles in front and rear of the robot. Each vehicle has flexible wheeled leg mechanisms pressing against the wall, respectively, and the friction between the wheel and inside wall of the pipelines helps generate driving forces. During the forward navigation the driving vehicle in front of the robot generate traction forces and the vehicle in the rear side gives pushing forces, and vice versa. It cannot be expected in usual mobile type robots [8,9] though there are some with magnetic wheels [8]. A driving vehicle consists of two vehicle segments and a steering mechanism between the segments. The steering mechanism called DAUJ enables three-dimensional steering of the robot in the pipelines. It makes it possible to choose the direction in the branches actively like a two-d.o.f. joint in an articulated manipulator. The other passive modules such as a control module and NDT modules are just linked via usual universal joints. As an additional feature of DAUJ it enables to control its compliance. Due to the wheeled leg mechanism as well as the compliance of DAUJ the whole body of the robot behaves like a flexible and spring-like one which avoids the excessive reflective forces from the wall during movements and provides excellent mobility. Also, DAUJ structurally does not

Fig. 2. Inpipe robot.

718

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

allow rolling of the segment along the driving direction. Otherwise the vehicle may be twisted or experience spiral motion while the driving wheels are pressing the wall. It may damage the robot seriously as well as cause the breakage of the tether cable eventually. The inpipe robot communicates with the ground station by a specially designed tether cable. The tether cable is composed of power lines, optical fibers for video signal and the transmission of digital data.

4. Driving vehicle The driving vehicle, as shown in Fig. 3, is composed of two segments, that is a front segment and a rear segment. Two segments are linked with a DAUJ. Both the front and rear segment have three wheeled leg mechanisms and those are arranged 12° apart circumferentially on the main body of the segments. The three wheeled leg mechanisms are simultaneously folded and unfolded by the action of the coil springs wound along the outer side of the main body that exert pressing force on the wall by way of the wheeled leg mechanism. In the proposed robot, the traction force mainly is generated by the wheels on the legs, actuated by a DC motor located in the main body of the rear segment. The power of the motor is evenly transmitted via worm gear reduction mechanism and timing belts in the legs. The front segment does not have active power and just has a CCD camera and several sensors useful for navigation. In this section we introduce characteristic features and design considerations of the driving vehicle.

Fig. 3. Driving vehicle.

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

719

4.1. Geometric constraints in pipelines Typical configurations of pipelines are level pipes, vertical pipes, elbows, branches, reducer, and valves. Most inpipe robots easily negotiate level pipe runs and can climb and descend pipes with inclinations of 30° or less. Vertical pipe runs are more difficult to achieve, and few robots today perform this feat. Those pipeline configurations gives geometric limitations and a robot should be designed to satisfy the limitations to traverse through pipelines successfully. Among these, the curvature of pipeline is the most important information because the robot is caught when it is too long and tightening up when it is too thick. In fact the government regulates the curvatures of pipelines should be kept less than inverse of 1.5 times of its diameter. Thus, each segment of the robot just needs to satisfy this condition in order to traverse pipelines. Now, let us consider a right angle elbow and derive constraint equations for the design of a module. A segment in a pipeline may be simply modeled as a cylindrical segment shown in Fig. 4. Then, the robots can be considered as linked cylindrical segments and we can derive relations among the diameter of pipelines, curvature, and the size of the segment. As illustrated in Fig. 4, the worst location of the segment is where it is located inclined 45°. In this situation, we can think about two cases: (a) the width of segment w is relatively smaller than the height h and both ends of the segment are located on the straight parts of the pipelines. (b) Both ends of the segment are located on the curved part of the pipelines. Depending on the situations, we can derive the constraint equations to determine the size of the segment. In case of (a) the range of w can be derived as follows. 0 < w 6 fðR þ D=2Þ sin 45°  ðR  D=2Þg;

Fig. 4. Segment in elbow.

ð1Þ

720

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

where D is the diameter of the pipelines and R denotes the radius of curvature. Thus, the maximum length of the segment is given by pffiffiffi h ¼ 2 2fD=2 þ R  ðR  D=2 þ wÞ cos 45°g: ð2Þ In the case of (b), both ends of the segment is located at the curved part of the pipelines. Thus, the range of the width w is written by fðR þ D=2Þ sin 45°  ðR  D=2Þg < w < D and the maximum length of the segment becomes qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 h ¼ 2 ðD=2 þ RÞ  ðR  D=2 þ W Þ :

ð3Þ

ð4Þ

Eqs. (2) and (4) provide the basic constraint equations. Those can be applied to the case of branches and the sizes of the segments of the proposed robots has been determined by employing those equations. 4.2. Wheeled leg mechanism One of the most important issues in the design of a driving vehicle is how to obtain the traction force enough to pull instrumentation as well as the vehicle itself. Especially in vertical pipelines, it is desirable to keep adequate wall pressing forces in order to ensure sufficient traction forces. Excessive forces may dissipate power and be in danger of damaging the robot. On the contrary insufficient forces may cause the robot to fall down. On the condition that the wheel does not slip on the pipeline surfaces, the traction force is proportional to the friction coefficient and the pressing force between the wheel and the pipeline surface, and the friction coefficient depends on the material of wheel and the surface condition of pipelines. In addition, the link mechanism of the vehicle should minimize the variation of traction force caused by variation of pipeline diameters. Therefore, a leg mechanism has to meet the following three requirements. At first, it should be possible to push against the pipeline wall with adequate pressing forces. In the second, the pressing force should not show significant change during navigation in order to provide stable traction force and flexible locomotion. At last, the mechanism should be simple and small in size to occupy minimal space inside the pipelines. As shown in Fig. 3, the driving vehicle of the proposed mechanism has three wheeled legs circumferentially spaced 120° apart on the main shaft of the vehicle. Fig. 5 illustrates the kinematic diagram of the wheeled leg mechanism. The mechanism employs a pantograph mechanism with a sliding base that permits the natural folding and unfolding of the leg. Here, l is the length of link, h means the folding angle of the link measured by the rotary potentiometer, K denotes the spring constant, h represents the distance of the center of the wheel from the base. Fw denotes the wall pressing force, Ax and Ay are the forces acting on the link by the spring, x is the displacement of the sliding base. In the proposed mechanism when the wheels are pressed they just contract or expand along the radial direction. It is a very advantageous feature because undesirable distortion forces are not exerted on the robot when the robot goes over obstacles.

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

721

Fig. 5. Wheeled leg mechanism.

Using Fig. 5 we can derive several basic equations necessary for optimizing the wall pressing forces. First the relation between h and x can be obtained as pffiffiffiffiffiffiffiffiffiffiffiffiffiffi h ¼ 2x tan h ¼ 2 l2  x2 : ð5Þ When the link rotates by h, the radial force Ax and the axial force Ay acting on the spring is written by Ax ¼ 2Fw = tan h;

Ay ¼ 0:

By using Eqs. (5) and (6), Eq. (7) will be derived. pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi l2  h2 =4 2Fw x : Ax ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 4Fw h l2  x 2

ð6Þ

ð7Þ

Now, let us differentiate Eq. (7) and derive spring constant K at the operating point xd (8 in.) which satisfies Ax ¼ Kðx  x0 Þ;

ð8Þ

where x0 denotes the initial displacement. Then, we have 2Fw l2 K ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 2 ; l2  xd l  x2d x0 ¼

x3d : l2

ð9Þ ð10Þ

722

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

Eq. (9) represents linearized spring constant and Eq. (10) denotes the initial length of the spring. Both are the basic equations for computing the wall pressing forces. By adjusting K and x0 properly, the wall pressing force with minimum variation can be obtained. 4.3. Steering mechanism In the proposed system, the steering is realized with a new mechanism called DAUJ [14–16]. Fig. 6 describes the mechanism in detail. As illustrated in Fig. 6, DAUJ is a two-d.o.f. joint whose each d.o.f. moves actively or passively with coupled motions of two motor-and-clutch pairs. DAUJ has a gear-bearing-gear system, and an inner and an outer universal joints to prevent each segment from rolling. Without the universal joints, the gear heads are free to move with respect to each other upon a bearing which lies on a plane tilted /° from the perpendicular planes of each gear axis of rotation. As described in Fig. 7 depicting the kinematic structure of DAUJ, the proposed mechanism is able to yaw and pitch

Fig. 6. Details of steering mechanism.

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

723

Fig. 7. Kinematic structure.


2/°, and rolling is prohibited by the inner and the outer universal joints. Thus, it prevents the segment from rotating along the axial direction and the trails of pressing wheels on the wall of the pipeline can be preserved continuously. Otherwise the vehicle may be in danger of twisting during steering. Also, it has the advantage that the electrical cables such as power or signal lines are free from twisting and thus, wire harness can be quite simplified. Detailed kinematic analysis of DAUJ can be referred to [16].

5. System control The hardware of the robot consists of an embedded controller, motor drivers, optical communication modules, an operator module and an optional instrumentation module (UT module). As described in Fig. 8, the control system can be largely divided into two parts, such as a robot part and an operator part. In the robot part, there are two CCD cameras in front and rear side of the robot. The front camera is for monitoring the forward navigation and the rear one for the backward one. The robot motion is controlled by an embedded controller with an Intel 80196 microprocessor (we have an 80196 microprocessor in UT module for the realtime signal processing but its description is not included in this paper). The main controller supervises the robot control and communicate with the ground station. It sends the results of inspection from instrumentation modules to the ground station. The inpipe inspection robot has six motors (three for each driving vehicle) whose

724

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

Fig. 8. Controller structure.

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

725

Fig. 9. Elbow.

driving circuit is integrated in the embedded controller. Also, supporting electronics for optical signal converters and sensors, are packaged in the control module of the robot. Control programs have been developed with C and assembly languages. Several levels of interrupts are employed in the code and its realtime control is realized. In

Fig. 10. Pass by branch.

726

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

Fig. 11. Steering in branch.

Fig. 12. Control Algorithm.

the autonomous mode the front and rear driving vehicles are controlled under a velocity controller. We measured the velocity of the vehicle by an encoder in the motor and differentiating its readings. There is no other feedback action on the

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

727

Fig. 13. Graphical user interface.

robot. Simultaneously the system keeps the clutches of DAUJ’s ‘‘OFF’’ state. By releasing the clutches of DAUJ it can act just like a usual universal joint which helps prevent the excessive constraint forces in the elbows or non-straight pipelines. Therefore in the autonomous modes the whole robot will be just passively chained multiple segments. Fig. 9 describes the typical situation that the autonomous mode works. In the elbow or straight pipelines just velocity control is more than enough and thus the robot guided by the wall goes through the pipelines. On the contrary when the robot comes up with a branch things are quite different. Its control is turned over to the operator and the operator teleoperates the robot with a joystick. The typical situations we encounter in the branch are illustrated in Figs. 10 and 11. In Fig. 10 the robot keeps its original direction. Fig. 11 denotes the case that the robot steers to the pipeline different from the current one. In case of Fig. 10 longer segments are desired to prevent falling down into the branch. Thus we turn on the clutch and accordingly the length of the segments in the vehicle is doubled enough to go over the branch. In case of steering in the branch shown in Fig. 11 the robot is steered manually with a joystick while monitoring CCD images from the front vehicle. As illustrated in Fig. 11 the vehicle should experience large constraint forces due to the contact between the body and the wall, though the vehicle has flexibility in the wheeled leg mechanism. In the proposed robot this problem

728

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

Fig. 14. Testbed.

is nicely overcome by controlling the stiffness of DAUJ. In the teleoperation mode the operator drives the robot while adjusting the stiffness of DAUJ’s through the graphical user interface. The compliance control of DAUJ has been realized by a simple PI controller cascaded with a PI controller for adjusting the engagement ratio Tc through Pulse Width Modulation (PWM) control of the clutch. The PI controller for the clutch is as follows. Tc ¼ Ks Cs absðes Þ þ C0 ;

ð11Þ

where Cs and C0 denote arbitrary calibration factors, absð Þ represents the absolute value of its argument and he means the angle errors. Ks is the stiffness that determines the omni-directional stiffness of DAUJ. Compliance controllability are prerequisite in steering, otherwise the robot will get serious damages. Fig. 12 briefly covers up the control algorithm. Fig. 13 illustrates Graphical User Interface in the ground station for controlling the robot. GUI uses Windows98 as the platform and coded with Visual C++. It provides information about the motion, CCD images of front and rear sides, op-

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

729

Fig. 15. Swiveling of proposed mechanism.

eration menus such as speed, steering angle, etc., which can be commanded in realtime by the operator, and virtual map describing the configuration of the traversed pipelines. In fact, the proposed robot has additional sensors to detect the direction of gravitation. With these sensor readings and the measured steering angles, the configuration of the pipelines can be estimated. By the results of the estimation we can construct a 3D graphical model representing the configuration of pipelines called Virtual Map. Thus, elbows, straight pipelines and other characteristic features such as diameter variation, inclination or mechanical damages etc., can be easily figured out by just traversing the robot inside the pipelines.

6. Performance evaluation and discussions In this section, we mention experimental results to evaluate the performance of the proposed robotic system. Most of experiments have been carried out in a testbed for 8 in. urban gas pipelines (total length 26 m with horizontal and vertical pipelines, several elbows, a branch, a full-bore valve) shown in Fig. 14. The testbed has been constructed for developing and testing of various inpipe inspection systems and the

730

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

Fig. 16. Tracking of rectangular trajectories.

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

731

Fig. 17. Experiment of compliance control.

arrow lines denote the path that the proposed robot traversed. The small box in Fig. 14 enlarges the branch included in the testbed where the steering capability of the robot was evaluated. In the first task, we demonstrated the roll-free swiveling of the steering mechanism. The front segment of the driving vehicle was commanded to follow the various values of spherical coordinate a while the rear segment of the driving vehicle was fixed as shown in Fig. 15. The arrow represents the path that the head of the front segment followed and thus, we can see the orientation of the front segment does not change proving the roll-free motion. It is a very advantageous feature for an inpipe inspection robot with wall pressing wheels. Unless the roll-free steering is not assured, the wheels contacted with the wall should abruptly change their contact points during steering, which drives the robot body twisted with excessive torques. It is a quite dangerous situation and as long as we use the orthogonal pair for usual robotic joints, it is impossible to generate the motion displayed in Fig. 15. Also, in the second task the performance of the position tracking was tested. In this experiments the joint was commanded to follow the rectangle curve and its motion was monitored. Fig. 16 represent the 3D trajectories, pitch and yaw angles.

732

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

Fig. 18. Results of compliance control.

The time to take one revolution was about 14 s with the maximum speed of the actuator. In this case we can observe that the position error was negligible and accurate position tracking was accomplished. In the third task, we verified its capability of the omni-directional compliance control. The proposed mechanism can actively modulate the mechanical compliance by controlling the motor command and the engagement period of the magnetic clutch with PWM method. As shown in Fig. 17 the wheeled leg mechanisms were stripped off from the main body of the front segment and the main body was forced to contact with the force sensor (JR3). Then, the position command was increased so as to make the main body press the face of the force sensor and the pressing force was measured with it. As illustrated in Fig. 17 the compliance controlled DAUJ is able to be modeled as a virtual spring with stiffness Ks and thus, the reflection force Fs is proportional to the position error and stiffness Ks . Similar experiments have been performed by varying the mechanical stiffness of the joint such as 294, 441, and 588 N/m, respectively. Fig. 18 shows the experimental results. It can be obviously observed that the measured forces increase proportionally with the displacements, whose ratios can be approximated as the commanded stiffness Ks . The experiments

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

733

Fig. 19. Steering in branch.

show that the proposed system can realize the mechanical compliance only by modulating the engagement of magnetic clutches. In the fifth task, we demonstrated the steering operation in a branch and an elbow. As explained, DAUJ operates in two modes, autonomous and teleoperated modes. When it comes up with branches the teleoperated mode should be chosen. As shown in Fig. 19, while the robot changed the direction of navigation in the branch, DAUJ behaved like a stiffness controllable two-d.o.f. joint. Thus, though it had contact with the rigid wall of the pipeline, it is bent like a spring and achieved the natural steering motion. On the contrary, when it is required to pass by a branch as shown in Fig. 20, DAUJ acts like a rigid joint by making the clutches completely ‘‘ON’’ state and prevents the robot from falling into the bifurcated pipeline. In the autonomous mode, the clutches are in ‘‘OFF’’ states and the torques from the steering motors are cut off. As shown in Fig. 21, when it was running through the elbows or pipelines without bifurcation, DAUJ acted like a usual universal joint. In this case the active steering is not required because the geometric configuration of pipelines guides the robot.

734

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

Fig. 20. Pass by branch.

7. Conclusion In this paper we presented a robotic system for the inspection of underground urban gas pipelines. This system shows outstanding mobility and several characteristic features, which make it possible to apply the proposed system in pipelines with complicate geometries. As the system has been developed as a mobile platform to traverse inside of the pipelines, it can carry various NDT sensors. We are now developing several NDT modules such as Ultrasonic Testing (UT) modules, Magnetic Flux Leakage (MFL) modules and tools for carrying simple tasks out in the pipelines. In the near future, field tests will be conducted with the system and the system is to be modified according to the results of field evaluation. In this stage the robot has the disadvantage that it becomes a manually operated one when it meets a branch because there is no sensor to out the direction of the bifurcation. Thus how to automatically check out the driving direction will be a major point of improvement to the system in the future. We are now planning to use the CCD camera and image processing technique. For those means a new embedded controller based on the off-the-shelf hardware with PC-104 bus is under development and the more improved system is expected to be reported in the next version.

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

735

Fig. 21. Elbow.

Acknowledgements The authors are grateful for the support provided by a grant from the Korea Science & Engineering Foundation (KOSEF) and the Safety and Structural Integrity Research Center at the Sung Kyun Kwan University.

References [1] Web site of Pipetronix Inc., http://www.pipetronix.com. [2] Okamoto Jr. J, Adamowski JC, Tsuzuki SG, Buiochi F, Camerini SC. Autonomous system for oil pipelines inspection. Mechatronics 1999;9:731–43. [3] Suzumori K, Hori K, Miyagawa T. A direct-drive pneumatic stepping motor for robots: designs for pipe-inspection microrobots and for human-care robots. In: Conference Proceedings IEEE International Conference on Robotics and Automation, 1999. p. 3047–52. [4] Okada T, Kanade TA. Three-wheeled self-adjusting vehicle in a pipe FERRET. Int J Robotics Res 1987;6(4):60–75.

736

H.R. Choi, S.M. Ryew / Mechatronics 12 (2002) 713–736

[5] Okada T, Sanemori T. MOGRER: A vehicle study and realization for in-pipe inspection tasks. IEEE J Robotics Autom 1987;3(6):573–82. [6] Fujiwara S, Kanehara R, Okada T, Sanemori T. Development of an articulated multi-vehicle robot for monitoring and testing in pipe. J Robotics Soc Jpn 1994;12(2):318–27. [7] Taguchi K, Kawarazaki N. Development of in-pipe inspection robot for small radius pipelines. J Robotics Soc Jpn 1997;15(2):230–5. [8] Kawguchi Y, Yochida I, Kurumatani H, Kikuta T. Development of an in-pipe inspection robot for iron pipes. J Robotics Soc Jpn 1996;14(1):137–43. [9] Ilg W, Berns K, Cordes S, Eberl M, Dillmann R. A wheeled multijoint robot for autonomous sewer inspection. In: Conference Proceedings IEEE International Conference on Intelligent Robots and Systems, 1997. p. 1687–92. [10] Neubauer W. A spider-like robot that climbs vertically in ducts or pipes. In: Conference Proceedings IEEE International Conference on Intelligent Robots and Systems, 1994. p. 1178–85. [11] Moraleda J, Ollero A, Orte M. A robotic system for internal inspection of water pipelines. IEEE Robotics Autom Mag 1999;6(3):30–41. [12] Hirose S, Ohno H, Mitsui T, Suyama K. Design of in-pipe inspection Vehicles for £25; £50; £150 pipes. In: Conference Proceedings IEEE International Conference on Robotics and Automation, 1999. p. 2309–14. [13] Pellegrino BA, Sigrist WR, Roman TH. Pipe-crawling inspection robots: an overview. Material Evaluation 1992;50(5):591–6. [14] Hirose S. Biologically Inspired Robot. Oxford: Oxford Science Publication; 1993. [15] Ikeda H, Takanashi N. Joint assembly movable like a human arm. US Patent 4 683 406, 1987. [16] Choi HR, Ryew SM. Anthropomorphic joint mechanism with two degrees of freedom. In: Conference Proceedings IEEE International Conference on Robotics and Automation, 2000. p. 1525–30.