Proceedings of the 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems San Diego, CA, USA, Oct 29 - Nov 2, 2007
ThA6.1
Runtime Reconfiguration of a Modular Mobile Robot with Serial and Parallel Mechanisms H. X. Zhang, Member, IEEE, S.Y. Chen, Member, IEEE, W. Wang, J. W. Zhang, G. H, Zong
Abstract—This paper presents a novel field robot JL-I based on a reconfigurable concept for urban search and rescue applications. The robot consists of three identical modules; each module is an entire robotic system that can perform distributed activities. It features three-degrees-of-freedom (DOF) active joints actuated by serial and parallel mechanisms for changing shape and flexible docking mechanism. The docking mechanism enables adjacent modules to connect or disconnect flexibly and automatically. DOF analysis, working space analysis and the kinematics of the 3D active joint between connected modules are studied thoroughly. In the end a series of successful tests confirm the principles and the robot’s capabilities.
T
I. INTRODUCTION
he last few years have witnessed an increasing interest in research and application of mobile robotic technologies for civil rescue and search all over the world. The history of human development has always been a struggle with natural disasters such as earthquakes, storms and floods. Recently the number of disasters by accidents or terrorism has evidently been increasing too. Urban search and rescue is a domain that involves a great amount of manpower; and it is quite dangerous and laborious in a hostile environment [1]. The development of mobile robots offers an intelligent alternative solution to the above-mentioned problems. In Japan, some prototypes aiming to mitigate the damages and decrease the number of victims during accidents and disasters were achieved at the International Rescue This work is supported in part by the “Hi-Tech Research and Development Program of China” (No. 2006AA04Z241). H. X. Zhang is with the Institute of Technical Aspects of Multimodal Systems, Department of Computer Science, University of Hamburg, Vogt-Koelln-Strasse 30, 22527, Hamburg, Germany ( As the corresponding author, e-mail:
[email protected],
[email protected]). S. Y. Chen is with the College of Information Engineering, Zhejiang University of Technology, Hangzhou, China, currently as a guest researcher in the Dept of Informatics, University of Hamburg, Germany and supported with a fellowship from the Alexander von Humboldt Foundation. W. Wang is with Robotics Institute, School of Mechanical Engineering and Automation, Beijing University of Aeronautics and Astronautics, 37 Xueyuan Road, 100083, Beijing, China. J. W. Zhang is with the Institute of Technical Aspects of Multimodal Systems, Department of Computer Science, University of Hamburg, Hamburg, Germany. He is also the leader of Cognitive Technology Laboratory, Shenzhen Institute of Advanced Technology, Shenzhen, China. (e-mail:
[email protected]). G. H. Zong is with the Robotics Institute, School of Mechanical Engineering and Automation, Beijing University of Aeronautics and Astronautics, Beijing, China (e-mail:
[email protected]).
1-4244-0912-8/07/$25.00 ©2007 IEEE.
System Institute (IRS) [2]. Another good example is the use of robots for the search and detection operation in the collapsed World Trade Center in September 2001 [3]. In Europe, security robotics becomes an important part of the 7th Framework Programme [4]. Locomotion capability is the lowest basic functionality of the robot system for urban search and rescue application. The robot should have a flexible mobility in rugged terrain to get to every point in the working space. In order to complete a task in an unstructured environment, the ability to cross high obstacles and span large gaps is indispensable. Therefore the robot should have the capability of adopting different configurations to match various tasks and suit complex environments. This paper presents a novel modular reconfigurable mobile robot named JL-I, which to date consists of three identical modules. Actually JL-I is a general platform which can also be used for industrial inspection and conducting surveillance, military reconnaissance and civil exploration. Modular approach meets the requirements of flexibility, functionality and extensibility for future uses. In order to implement locomotion capabilities, independent identical units should work alone and collectively in different environments according to the tasks. For example, usually searching and rescuing is based on the cooperation in a team. The given targets will be assigned separately. The single unit should work independently and communicate with the others to perform distributed activities. While several units connecting together form a new train-like system to cross big obstacles or span large gaps in order to cover the working environment. The emphasis for discussion of this paper is on the locomotion capabilities and reconfiguration realization. DOF analysis, working space analysis and the kinematics of the 3D active joint between any two connected modules are studied thoroughly. A series of successful on-site tests are given to confirm the principles proposed and the robot’s capability. II. PROTOTYPE DESIGN A. Previous work on runtime reconfiguration The common characteristics of reconfigurable robotic systems lie in two points. Firstly, the robotics system comprises several similar modules which are independent units with full locomotion functions. Secondly, specially designed joints connect individual units to form a more
2999
flexible prototype. It is noted that the first prototype [5] with powered wheels was designed by Hirose and Morishima in 1990, which consists of several vertical cylindrical segments. Another robot with wheels on each module to provide the driving force was developed by Klaassen for the inspection of sewage pipes [6]. A serpentine robot from Takayama and Hirose consists of three segments. Each segment is driven by a pair of tracks, but all tracks are powered simultaneously by a single motor located in the centre segment [7]. The special ability of adapting to irregular terrain is passive and provided by springs. KOHGA [8] has recently been developed by IRS in Japan. It consists of eight serially interconnected individual units with two tracks except the first and last modules. Another group of reconfigurable robots features passive modules [9] [10]. It can only move after the modules are assembled by active joints. In [11], 1D, 2D and 3D chain robots are classified according to their topology. As an example, PolyBot is able to optimize its parts to fit the specific task. Generally, this kind of reconfigurable robots is relatively simple so that the locomotion capability is not as efficient as the above-mentioned kind with powered tracks. However, for urban rescue and search, the fact that the known reconfigurable robots can only assume few configurations due to relatively simple connecting and pose-adjusting mechanisms is a ubiquitous deficiency [12] [13]. Since 1999 our group has been focusing on the design and development of mobile robots for urban search and rescue purposes. A smart mobile robot was proposed as a flexible mobile platform carrying a CCD camera and other sensors [14]. A more flexible structure with two linked-track vehicles was proposed [15]. The structure can be reconfigured so that the robot can move between surfaces standing at an angle of 0 - 90 degrees due to the pitching DOF actuated by the joint to increase the flexibility. The project presented in this paper has the aim of developing an automatic field robot to meet the requirements of high flexibility, robustness. B. Design of a robot prototype A novel modular reconfigurable mobile robot named JL-I with various moving modes was proposed [16]. To date, the system consists of three connected, identical modules for crossing grooves, steps, obstacles and traveling in complex environments. Each module is an entire robot system that can perform distributed activities. Three-DOF active spherical joints between two modules and the docking mechanism enable the adjacent modules to adopt optimized configurations to negotiate difficult terrain or to split into three small units to perform tasks simultaneously. Fig. 1 shows the system working alone and collectively in different environments. One module is about 43 cm long, 25 cm wide and 15 cm high. The mechanical structure of the module comprises two powered tracks, a serial mechanism, a parallel mechanism, and a docking mechanism (Fig. 2). Two DC motors drive the tracks providing skid-steering ability to realize omnidirectional movement.
Fig. 1 JI-I working alone and simultaneously.
Fig. 2 The mechanical structure in design of the module.
The docking mechanism consists of a cone-shaped connector at the front and a matching coupler at the back of the module. The coupler is composed of two sliders propelled by a motor-driven screw. Two sliders form a matching funnel which guides the connector to mate with the cavity. Then the slides will move toward each other until these two surfaces mate with other two surfaces on the cone-shape connector so that two mating planes between the sliders and the cone-shaped connector constrain the movement, thus locking the two modules. According to the required task, JL-I can change its posture by pitching around the Y axis, yawing around the X axis and rotating around the Z axis in order to adopt optimized configurations to negotiate difficult terrain. JL-I features serial and parallel mechanisms to form a 3DOF active spherical joint. The serial mechanism can rotate 360° around the Z axis. The parallel mechanism can pitch around the Y axis and yaw according to the X axis. Each leg of this parallel joint consists of a driving platform, a universal joint, a screw, a synchronous belt system, a DC motor and a base platform.
3000
III. REQUIRED LOCOMOTION CAPABILITIES Due to the uncertainty of the practical environment, it is very important for a robot to be able to carry out various complicated locomotion processes for performing urban search and rescue tasks. The robot designed in this paper is capable of almost all necessary actions that can be required in real situations, e.g. crossing obstacles such as steps and roadblocks, self-recovery. A. Crossing a Step Fig. 3 shows the process of crossing a step from a side view. The step is almost twice as high as the robot. Firstly the robot is in its home state, and the sensor detects the step in the movement direction. Then the first module pitches up around the Y axis while the robot moves forward. The approaching movement does not stop until the first module touches the step. After that the first module pitches down to attach to the top of the step. The robot moves forward continuously until the first two modules are attached to the step. In the end the last module pitches up around the Y axis while the robot moves forward so that JL-I is now in its home state again.
Fig. 3 The process to move onto a high step.
B. Self-recovery It is possible for the robot to implement 90°recovering movement by adopting the proper configuration sequence as shown in Fig. 4. The robot lies on its side. The first module and the last module yaw up around the X axes of the active joints. Then the first module and the last module rotate 90° around the Z axes. After that, they pitch down around the Y axes of the active joints until they attach to the ground in order to lift the middle module. The middle module rotates around the Z axis until it is parallel to the ground. In the end, the module pitches down around the Y axes of the active joints until all three modules attach to the ground together. The robot is now in its home state again, and the process of 90° Self-recovery is over. It is also possible for the robot to tip over and implement the 180° recovery movement. The principle is similar (Fig. 5).
Fig. 4 90° recovering movement.
Fig. 5 180° recovering movement.
IV. KINEMATICS ANALYSIS OF THE LOCOMOTIONS A. The DOF of the active joint To demonstrate the reconfiguring capability, the kinematics analysis of two connected modules should be studied. Fig. 6 shows the kinematics model of the joint between two modules. Where OXYZ is the world coordinate fixed at the plane QEF which represents the front unmovable module during the reconfiguration. The origin is located at the universal joint O, the Z-axis coincides with the axis of the serial mechanism and the X-axis points to the middle point of line AB. Another reference coordinate O’X’Y’Z’ is fixed at triangular prism OABPCD which represents the back moveable module. The O’X’Y’Z’ is coincident with the OXYZ when the spherical joint is in its home state. The required orientation for the reference frame O’X’Y’Z’ on the back module is achieved by a rotation of θz, a pitching angle θy and a yawing angle θx according to the relative axes. From the mechanical point of view, actually the pitching and yawing motions are implemented by the outstretching and returning movement of the L1, L2 of the parallel mechanism, and the rotation of θz is actuated by the serial mechanism. The freedom of the reconfiguring movement is three and can be described with the generalized coordinate θ (1). The joint variants of the movement are named q, described as (2). θ= [θx, θy, θz]T (1) (2) q= [L1, L2, θz]T
Fig. 6The kinematics mode of the active spherical joint.
3001
There are altogether 8 joints, described as g, out of which 3 joints are active and actuated by respective DC motors. Three Hooker joints are at points O, A, and B; two linear movement joints are at links AC and BD; one rotating joint is along the axis Z1Z2 and two spherical joints are at C and D. According to (3), the DOF can be concluded as (3). Where m means the DOF; n means the movable links of the active joint. There are seven links totally. Where fi is the DOF of the relative joints. It is noted that the DOF of a rotating joint and a linear movement joint is one; the DOF of the Hooker joint is two while the spherical joint has three DOF. Fig. 7 The working space of the pitching movement.
m = 6 ( n − g − 1) +
g
∑
i =1
fi = 3
(3)
B. Working space analysis The next question is the working space analysis in the world coordinate in order to implement all locomotion capabilities since there are three general DOFs. As discussed in the former section, the angle θz is required to have 360° rotation around the Z axis. It is noted that θz is an independent DOF actuated by the serial mechanism and it normally occurs after the pitching movement or yawing movement. So we only need to focus on the working spaces of θx and θy since both pitching and yawing movements are dependent on the extending or contacting cooperation of L1 and L2 in the parallel mechanism. For a general movement, the JL-I robot only needs to pitch up at a tiny angle in order to cross a low step or some small obstacles; it should also only yaw at a small angle to turn left or right during the movement. As a result, the working spaces of θx and θy can be described as (4). ⎧θ x ∈ [0°, ± θ x max ] ⎨ ⎩θ y ∈ [0°, ± θ y max ]
Synthesizing from (4) and (5) and taking the example parameters of the robot implemented in practice, i.e. y is 75 mm and l is 35 mm, finally the general working space of the pitching movement can be concluded to be
θ y ∈ [0°, ± 50.0°]
(6)
Now the minimum position should be calculated as the next step in order to actuate one module to rotate 90-180° around the Z axes of the active joints without any collision, as shown in Fig.5 (b). In Fig.7, c1 in red is the minimum pitching position in order to implement the DOF of θz. This minimum angle limitation is nevertheless very important for the practical implementation of the robot’s reconfiguration. The largest rotating section is illustrated in Fig. 8 in red, which is the same red rectangle in Fig. 7. According to the geometric analysis in Fig. 7, we have the following (7), (8).
(4)
In practice the maximum positions should be calculated considering the mechanical constraints and collision avoidance. Fig. 7 shows the analytic draft of the working space for the pitching movement. Here three rectangles represent the modules to simplify the discussion in the lateral view. The left module is fixed and the right one is moveable. Situation a1 in black is the same home state as that shown in Fig. 5(a); b1 in blue is the maximum pitching position taking the mechanical constraints into account in order to avoid collision. From Fig. 7, the maximum pitching angle θymax can be found as in (5). Where y is the half vertical height when the module stands on flat ground; l is the length of the equivalent connecting joint of JL-I; x is the half width of the largest rotating section. y (5) θ y max = 180 − 2arctg l
Fig. 8 The maximum rotating section during pitching movement.
θ y min = arccos
y l +x 2
2
− arccos
⎧⎪ x = Height 2 + Width 2 ⎨ ⎪⎩ y = Height / 2
x
(7)
l + x2 2
(8)
Given the height and width of the module, putting (8) into (7) we can get the minimum working space of the pitching DOF for robotic runtime reconfiguration in order to avoid collusions. Finally the working space for θy is as in (9). θ y ∈ [± arccos
y l +x 2
2
− arccos
x
y , ± 180 − 2arctg ] (9) l l +x 2
2
Similarly for the working space of the yawing movement, θx can be also described like θy (9) while y is the half width of 3002
the module at the moment. According to the prototype structure, the pitching and yawing working spaces are obtained. As in our implementation, when the module width is 250 mm, the height is 150 mm, the general working space is (10); while the restricted working space for avoiding collision is (11). ⎧θ x ∈ [0°, ± 32.0°] ⎨ ⎩θ y ∈ [0°, ± 50.0°] ⎧θ x ∈ [ ±8.0°, ± 32.0°] ⎨ ⎩θ y ∈ [ ±24.0°, ± 50.0°]
Normally a direct kinematics solution is hard to find because of multiple possible solutions due to the characteristics of extension movements of the parallel mechanism. A numerical method instead is used to resolve the problem. Even if the direct relationship is not clear, the required working spaces of joint movements can be plotted as in Fig. 9.
(10) (11)
In order to simplify the mechanical structure, we can design the working space of θx, θy to be the same, e.g. both within -50 to 50 degrees, which not only reduces the implementation cost but also slightly increases redundancy for practical operation. C. Inverse kinematics analysis The purpose of the kinematics analysis is to deduce the relationship between q and θ in (1) and (2). From Fig. 6, there are two mechanical constraints as shown below. 1) Three triangles QEF, OAB, PCD are equal and parallel; 2) QF is perpendicular and equal to QE. A, B, C, and D have the OXYZ coordinates in (12). Where L is the initial length of L1 and L2 without any extending or contracting changes and K is a half width of the parallel mechanism. Practically L is 133 mm and K is 20 mm. A= [K, -K, 0]T; B= [K, K, 0]T; C= [K, -K, L]T; D= [K, K, L]T
(12)
After a reconfiguring movement, A, B, C, and D are changed to new positions, denoted A1, B1, C1, and D1 expressed as (13), (14). B1] = Rot(Z )[A B] D1 ] = Rot (Y ) Rot ( X ) Rot ( Z )[C
[ A1 [C1
D]
(13) (14)
From the constraints of the mechanical structure, we know that the lengths of the link L1 and L2 are equal to the distance between C1A1 and D1B1, respectively. Combining this knowledge with (13) and (14), we get (15). [L1
L2 ] = [C1
D1 ] − [ A1
⎡ K (cθ Y cθ Z + sθ X sθ Y sθ Z ) − K ( − sθ Z cθ Y + sθ X sθ Y cθ Z ) + Lcθ x sθ y − Kcθ Z − Ksθ Z ⎢ B1 ] = ⎢ K (cθ X sθ Z ) − K (cθ X cθ Z ) − Lsθ x − Ksθ Z + Kcθ Z ⎢ K ( − sθ Y cθ Z + sθ X cθ Y sθ Z ) − K ( sθ Y sθ Z + sθ X cθ Y cθ Z ) + Lcθ x cθ y ⎣
K (cθ y cθ z + sθ x sθ y sθ z ) + K (− sθ z cθ y + sθ x sθ y cθ z ) + Lcθ x sθ y − Kcθ z + Ksθ z ⎤ ⎥ K (cθ x sθ z ) + K (cθ xcθ z ) − Lsθ x − Ksθ z − Kcθ z ⎥ K (−sθ y cθ z + sθ x cθ y sθ z ) + K (sθ y sθ z + sθ xcθ y cθ z ) + Lcθ x cθ y ⎦⎥
(15)
Fig. 9 The 3D plots of L1 and L2 according to θx and θy within 180 degrees.
However, as restricted by (10), the practical working space of L1 and L2 should be within some disperse zones respectively to echoes the conclusion. For implementing the serial and parallel mechanism, an equivalent form of (16) derived from (15) is gotten when θz is set to zero since at the moment we only focus on the parallel mechanism. ⎧d1 = L1 ⎪ ⎨ ⎪⎩d 2 = L1
2 2
− L2
2
= 4 Ksθ x ( Ks θ y − L )
+
2
= 8 K 2 + 2 L2 − 4 KLs θ y cθ x − 4 K 2 cθ y − 4 K 2cθ x
L2
(16) Where ||L1|| is the Frobenius norm (i.e. the segment length) of vector L1 and ||L2|| is that of L2. The Jacobian matrix J of this system is (17). ⎡ d ( d1 ) ⎢ dθ x J =⎢ ⎢ d (d 2 ) ⎢ dθ x ⎣
d ( d1 ) ⎤ dθ y ⎥ ⎡ 4 Kcθ x ( Ksθ y − L) ⎥= d (d 2 ) ⎥ ⎢⎢4 KLsθ x sθ y + 4 K 2 sθ x ⎣ dθ y ⎥⎦
⎤ 4 K 2 sθ x cθ y ⎥ − 4 KLcθ x cθ y + 4 K 2 sθ y ⎥⎦
(17) With the Jacobian matrix, equation (16) can be exactly solved to find θx and θy using the Levenberg-Marquardt algorithm, when L1 and L2 are given. The 3D plots for θx and θy are shown in Fig. 10. Therefore, the transformation
3003
between q and θ are solved so far. As noted, it is complicated and time consuming to calculate the outputs of the serial and parallel mechanism in order to actuate the module to an expected reconfiguration according to (15) or (16). In practice a numerical method is often better to realize real-time kinematics. Now Fig. 10 provides an “actuation surface” which tells us all possible positions of L1 and L2 and the corresponding spatial angles θx and θy. Thus it offers an easy and quick solution for practical robotic control. According to the amount of output by the joints, the locomotion implementation will be anticipated immediately in the world coordinate, thus reducing the computation time and improving the working efficiency.
(a)
(b) (c) Fig. 12 Snapshots of the 90° self-recovery
VI. CONCLUSION This paper proposes a novel reconfigurable robot named JL-I for urban search and rescue application. The active spherical joints formed by serial and parallel mechanisms endow the robot with the ability of changing shapes in three dimensions. The kinematics model of reconfiguration between two modules is given. The relationship of the world coordinate and the reference joint coordinate is concluded. Furthermore, the movements can be anticipated according to the joints’ driving outputs. REFERENCES [1] [2] [3] [4] [5] [6]
[7]
[8]
[9]
Fig. 10 The 3D plots of θx and θy within 0 to 50 degrees.
[10]
V. ON-SITE EXPERIMENTS Recently some on-site experiments are implemented. Fig.11 demonstrates the process of crossing a step from a side view. As mentioned before, it is possible for JL-I to implement a 90-180° recovering movement by adopting the proper configuration sequence, as shown in Fig. 12. Experimental tests have shown that the JL-I can implement a series of various reconfigurations and have implied the mechanical feasibility, the rationality of the analysis.
[11]
[12] [13]
[14] [15] (a)
(b) Fig. 11 Snapshots of crossing a step.
(c)
[16]
3004
F. Matsuno, S. Tadokoro, “Rescue Robots and Systems in Japan”, Proceedings of Robio2004, Shenyang, China, 22-26 Aug. pp.12-20. Http://www.rescuesystem.org J. Casper, “Human-Robot Interactions during the Robot-assisted Urban Search and Rescue Response at the World Trade Center”, MS Thesis, Computer Science and Engineering, USF, April, 2002. Ftp://ftp.cordis.europa.eu/pub/ist/docs/so/advanced-robotics/ie-so-mar ch06-ag-3.pdf S. Hirose, A. Morishima, “Design and control of a mobile robot with an articulated body”, The International Journal of Robotics Research, Vol. 9 No. 2, pp. 99-113, 1990 B. Klaassen, and K.L. Paap, “GMD-SNAKE2: a snake-like robot driven by wheels and a method for motion control”, Proceedings of IEEE International Conference on Robotics and Automation, Detroit, MI, 10-15 May, pp. 3014-3019,1999. T. Takayama, S. Hirose, “Development of Souryu-I connected crawler vehicle for inspection of narrow and winding space”, Proceeding of the 26th Annual Conference of the IEEE Industrial Electronics Society, Vol.1, pp.143-148, 2000. T.Kamegawa, T. Yamasaki, H. Igarashi and F. Matsuno, “Development of the Snake-like Rescue Robot “KOHGA””, Proceedings of the 2004 IEEE International Conference on Robotics and Automation, pp. 5081-5086, 2004. M. Yim, Y. Zhang and E. Mao, “Distributed control for 3D shape Metamorphosis”, Autonomous Robots, Vol.10, No.1, pp.41-56, 2001. S. Murata, E. Yoshida, A. Kamimura, H. Kurokawa, K. Tomita, S. Kokaji, “M-TRAN:Self-Reconfigurable Module Robotic System”, IEEE/ASME Transactions on Mechatronics, Vol.7, No.4, pp.431-441, 2002. J. Gonzalez-Gomez, H. Zhang, E. Boemo, J. Zhang, “Locomotion Capabilities of a Modular Robot with Eight Pitch-Yaw-Connecting Modules” , Proceeding of CLAWAR 2006, Brussels, Belgium, September 12-14, 2006. H.B. Brown, et al, “Millibot Trains for Enhanced Mobility”, IEEE/ASME Transactions on Mechantronics, Vol.7, No.4, pp.452-461, 2002. M. Park, W. Chung, M. Kim, “Control of a mobile robot with passive multiple trailers”, Proceedings of the 2004 IEEE International Conference on Robotics and Automation, New Orleans, LA, United States, pp.4369-4374, Apr.-May 2004. Http://tams-www.informatik.uni-hamburg.de/people/hzhang/projects/i ndex.php?content=Gecko W. Wang, G. Zong, “Analysis on The Mechanics Stability for a New Kind of Robot”, Journal of Robot, Vol.21, No.7, pp.642-648, 1999. H. Zhang, W. Wang, Z. Deng, G. Zong, “A Novel Reconfigurable Robot for Urban Search and Rescue”, International Journal of Advanced Robotic Systems, Vol.3 No.4, 2006, pp.359-366.
