Journal Search Engine
Search Advanced Search Adode Reader(link)
Download PDF Export Citaion korean bibliography PMC previewer translate in English
ISSN : 1975-6291(Print)
ISSN : 2287-3961(Online)
Journal of Korea Robotics Society Vol.7 No.2 pp.65-75

인간 동작 표현용 스프링 백본 구조 소프트 암의 구현

윤 현 수1, 최 재 연2, 오 세 민3, 이 병 주, 윤 호 섭4, 조 영 조4

Implementation of a Spring Backboned Soft Arm Emulating Human Gestures

Byung‐Ju Yi, Hyun‐Soo Yoon1, Jae Yeon Choi2, Se Min Oh3, Ho Sup Yoon4, Young‐Jo Cho4
Corresponding author : Department of Electronics & System Engineering, Hanyang University
1 Department of Electronic, Electrical, Control, and Instrumentation Engineering, Hanyang University, 2 Korea Institute of Robot and Convergence,
3 Department of Intelligent Robot Engineering, Hanyang University,
4 Electronics and Telecommunications Research Institute(ETRI)
Received : Aug. 12. 2011; Reviewed : Sep. 28. 2011; Accepted :Mar. 29. 2012


This study deals with the design of a spring backboned soft arm, which will be employedfor generation of human gesture as an effective means of Human‐Robot interaction. The specialfeatures of the proposed mechanism are the light weight and the flexibility of the whole mechanismby using a spring backbone. Thus, even in the case of collision with human, this device is able toabsorb the impact structurally. The kinematics and the design for the soft arm are introduced. Theperformance of this mechanism was shown through experiment emulating several human gesturesexpressing human emotion and some service contents. Finally, this soft arm was implemented as thewing mechanism of a penguin robot.

02 KRS-11-02 026-R1-(20120509-142003)-Final-115.pdf2.92MB

1. Introduction

 In the environment where human and robots coexist, robots should be designed and programmed in such a way that they are able to interact with human socially and physically in a safe fashion. The concept of HRI (human‐robot interaction) has been evolved according to this demand.[1]

 Robots being developed with the background of HRI collect sensors’ data, process the signals, recognize the information, and finally provide valuable services to human. As an example, the robot recognize the face of human by camera, detect the motion of human by a position sensor, and receive the intention of human by voice recognition. Combining those sensors’ information, the robot decides the contents of the service, the sequence of service, and finally come to service action.  Besides routine services such as many manipulation jobs and vocal service for human, facial expression or some gesture of robots help communication with human. Imitation of the facial expression[2,3] and gesture[4,5] of human are very important since it provides a familiar feeling when robots interact with human.

One of most critical issues of HRI is safety. Since the robot interacts with human physically or indirectly, a small position error or some programming error might cause crash between human and robot and resultantly damage the human body. Therefore, robot must have a soft arm structure that can absorb the shock or impact when contacting the human body. Design of a structurally soft robot hardware or precise control algorithm is required at the existence of such probable physical interaction. 

Investigation of soft arm has drawn much attention recently. Kawamura, et al.[6] developed a soft arm using pneumatic actuator. Bicchi, et al.[7] investigated the tactic of fast and soft arm. Choi, et al.[8] proposed a continuum mechanism applied to the design of soft endoscope. Park, et al.[9,10] proposed compliant arms that have variable compliance at the robot joint or link. Khatib, et al.[11] proposed a dual‐actuator based feedback control algorithm to realize the soft arm. Yuki, et al.[12] developed a soft manipulator with flexible joints using smart fluid and pneumatic cushion. Yoon, et al.[13] designed a safe arm with MR‐base passive compliant joints and viscous‐elastic covering.

 The objective of this study is to suggest a new soft arm for imitating human gesture, which is not only a means to enhance the effectiveness of HRI, but also enhances the safety by employing a soft arm structure. The special feature of this soft arm is the spring‐backboned structure, which allows structurally flexible characteristic. Even in the case of collision with human, this device is able to absorb the impact. The driving unit and the mechanism are separated. Thus, wide range of workspace can be ensured and a fast motion can be created because of small inertia of the system.

The section II deals with the kinematics of the soft arm. Its design will be described in the section III. Experimental implementation of diverse gestures is demonstrated in the section IV. Finally, we draw conclusion in the section V. 

2. Kinematics of the soft arm

2.1 Structure

 This study deals with the design of a soft arm with less number of actuator and emulate the motion imitating a couple of human gestures. Only three DC motors are used for this purpose. Fig. 1 shows the prototype of the soft arm. A spring is used as a backbone. The spring backbone is wrapped by acetyl cylinders at a regular interval, which are connected by wires through three holes pierced in the cylinders. As shown in the figure, the arm is slimmed down from the proximal joint to the distal joint in order to reduce the weight of the arm.

Fig. 1. Structure of Soft Arm

 Applying tensions to wires as shown in Fig. 2(b), the general motion of the soft arm can be represented as superposition of compression along the axial direction and bending in the global X-Z plane. More specifically, Fig. 2(c) shows the deflection of one spring module caused by wire tensions. Note that because of the symmetry of the spring module, the length of the center line is the average of the lengths of the two sides.

Fig. 2. Side view of the soft arm structure

2.2 Kinematics for a planar model

To find the end‐position of the soft arm relative to the global reference coordinates, first, a planar model is considered. Fig. 2 shows the configuration of the planar model being bent in the global X-Z plane where β1 denotes the absolute rotation angle of the soft arm about the axis. As in Fig. 2(a), one spring and one link are defined as one node. Particularly, the dotted lines in Fig. 2(c) represent the locations of the virtual wires which are assumed to be located both on the inner surface and outer surface of the soft arm in the global X-Z plane. It is noted that the end positions of the last node can be computed by summing up the lengths of wires that are assumed to be embedded along the center line of every node. 

To compute the lengths of two wires for the given output angle (β1), the positions (Xpe and Zpe) of the last node relative to the node reference frame XOZ are obtained first. Then, the lengths of two wires are calculated.

Since the soft arm is designed symmetric as in Fig. 1, deflection will occur uniformly throughout every module. Thus, analysis of one module could lead the kinematics for the whole mechanism. For our convenience, the module is numbered from the distal node to proximal node. Fig. 3 shows a module i of the planar model where lengths of the center lines (i.e. P and ) are assumed known.

Fig. 3. A node of the planar model

Since the relative rotation angle of every module with respect to its neighboring module toward the wall is the same, the absolute rotational angle of each module about the axis (β1, for i = 1, …, N) can be found as  

 where N is the number of nodes. Thus,  θa can be expressed as

For the known value of θa, a is obtained from the geometry of Fig. 3 by using the cosine’s law as below 

where Δl is the amount of compression of the spring backbone as shown in Fig. 2(b). 

 Also, βi+1 can be found from Fig. 3 as


It is also noted that the proximal node is constrained to the wall such that βN+1 = 0°. 

 Finally, the position of the distal point of the node i relative to the node i+1 is obtained as

Summing up Eqs. (5) and (6) for all nodes, the positions of the end‐point in the reference coordinate can be obtained as 


2.3 Length of structure for the planar model

Now, θv denotes the angle change of the backbone spring module as shown in Fig. 2(c). The angle θv of every spring module has the following relation with β1 as below:

 To bend the structure, the relationship between β1 and the lengths of the pulling wires should be obtained. In the planar model, two wires are assumed to be embedded on the inner surface and on the outer surface of the backbone structure. In Fig. 2, Lc, Ls, and Ll, denote the length of the center line, the inner and outer boundaries of the planar soft arm model, respectively.

When the structure bends, Ls denotes the shortest length of the inner wire and Ll is the longest length of the outer wire as in Fig. 2(c). For the constant r and , k is obtained as 

where r is the radius of the cylinder. Then, the lengths of Ls, Lc and Ll are obtained as 

2.4 Kinematics for the spatial model

In the previous planar model, only one output angle β1 was required to specify the position of the end‐point of the robot or to find lengths of two wires which are assumed to be embedded on the inner and the outer surfaces of the soft arm. However, in the spatial model, one more output rotation parameter(γ) should be employed. The spatial model could be found by rotating the planar model (Fig. 2) by an azimuth angle γ about the global  axis as shown in Fig. 4 where one such configuration of the spatial model is illustrated. Two output variables β1 and γ represent the tilting angle of the soft arm and the orientation of the tilting angle, respectively. Note that this model only uses azimuth and tilt angles (β1, γ) in the tilt and torsion angle representation as shown below

Fig. 4. The bent shape of the spatial model

where σ represents the torsion angle and is assumed to be 0 in our model. 

 As shown in Fig. 4, when the end position of the soft arm in the planar model is known as Ppe(Xpe,0,Zpe), the global end positions Pe(Xe,Ye,Ze)in the spatial model can be found as

2.5 Length of wire for the spatial model

The position of the soft arm structure is represented by the given angles of β1 and γ. Particularly, because the rotation angle γ only transforms coordinates, it doesn’t affect the length change of the structure. Thus, the length of the inner boundary (Ls) and the length of the outer boundary (Ll) in the spatial model are equal to ones in the planar model.

Now, the relationship between the two boundary lengths and the lengths of three wires (l1, l2 and l3) is obtained from the bending geometry of Fig. 5 and the cross‐sectional area of the spring backbone structure given in Fig. 6. 

Fig. 5. Bending geometry

Fig. 6. Cross‐sectional area of the spring backbone structure

In Fig. 5 and Fig. 6, the lines connecting three points, p1, p2, and p3 denotes the bending plane. From the cross‐sectional geometry, θl1, θl2, and θl3 which represents the location of three wires, respectively, are expressed in terms of γ as follows 

It is noted that the amount of deformation along the axial direction of all elements on the dotted line perpendicular to the bending plane as shown in Fig. 6 is equal. Thus, the lengths of three wires can be computed as follows. In Fig. 2(c), the lengths of Ls and Ll can be expressed in terms of their unit lengths ( and ) such as 

Then,  and  are related as  

 Noting that in the bending plane of Fig. 6, deformations of all points on each dotted line (e.g., M1,M2,M3) are the same, the unit length liu of the wire li is related to the shortest inner length as


Thus, from (19) and (20), the unit length liu of the wire li can be expressed as 

The relationship between liu and the length li of the i‐th wire can be expressed as 

where P is the length of one cylinder module. By using (17), (18), (21), and (22), the length li of the i‐th wire is obtained as 


When the deflection due to the compression of the spring along the axial direction, denoted as Δl, is taken into account, the length of each of three wires can be found as 


Thus, Δli denoting the total length change of each of three wires li can be found as 


which is the amount of the required length change of wires for the given output variables (β1, γ, Δl). 

3. Design of the soft arm

3.1 Design of a Soft Arm

Figure 7 shows the design concept of the proposed soft arm, which consists of an arm part and a driving part. The arm is designed by assembling a spring and hollow plastic cylinders. Three wires penetrate into the holes located at the cross sectional area of the cylinders.

Fig. 7. The design concept of the soft arm

The driving unit adapts a linear actuator as the basic unit, which consists of a TM screw (M8; triangular gear with pitch 1.25mm) and a DC motor with encoder. The stroke of the screw is 35 mm. Three of this driving unit are assembled and placed 120 degrees apart. A pulley is employed to guide the pulling wire to the actuator and a photo sensor is used for home positioning. The arm comes to action when pulling the three wires simultaneously. 

3.1.1 Arm structure

The soft arm is designed to have a soft structure when contacting to environment or human. The spring mechanism plays such a role. Also, in order to prevent bucking of the whole mechanism, an extension spring or a flexible tube line is inserted at the center of the arm. 

Figure 8 shows the detail of the soft arm. It is designed by assembling several springs and plastic cylinders in a successive manner. As the length of the arm becomes longer, the whole mechanism deflects due to its own weight. Thus, we employ a light‐weight cylinder made of acetyl whose specific weight is about 1.4. Also we made several holes at the cross section of the cylinder to reduce the weight as shown in Fig. 9. 

Fig. 8. Structure of the arm

Fig. 9. Cross section of cylinder

Furthermore, the weight could be reduced by slimming down the arm from the proximal node to the distal node as shown in Fig. 8. Based on this design, Table 1. shows the final weight of each cylinder. 

Table 1. Weight of each cylinder

‐ Spring

The joint connecting two adjacent plastic modules is made by a coil spring so that the arm is able to bend when pulling the three wires. Table 2 shows the specification of springs. The diameter of each spring is decided properly according to the size of each node. The spring constants of all springs are the same. 

Table 2. Specification of springs

‐ Driving wire

The driving wire plays the role of driving the whole mechanism as well as connecting the plastic cylinders and the springs. The diameter of the stainless wire is 0.8mm and its maximum tensile load is about 0.588~0.667 kN (≒60~68 kgf). The three wires are placed 120˚ apart so that the soft arm can be steered in all directions. 

3.1.2 Analysis of the pulling load of wire

 In order to analyze the pulling load of wires, Fig. 10 depicts the moment created by the weight of the plastic cylinders.

Fig. 10. Moment due to the weight of cylinders

Summation of moments due to all nodes is denoted as 

 According to the weight and the distance information of each node from Table 2 and 3, the total moment that should be supported at location of the shoulder is calculated as

Table 3. Distance from the shoulder to each node


This moment should be resisted by the pulling load of the wire (Fig. 11). When the moment due to the weight of the arm is equal to the resisting moment due to the pulling wire, the soft arm maintains its shape without deflection due to its weight. Now, the moment due to pulling wire is denoted as

Fig. 11. Moment due to the pulling wire

where fw and r, respectively, denote the pulling load and the location of the wire from the center line of the spring. Resultantly, the pulling load of the wire preventing the deflection of the arm is calculated as 

‐ Backbone

 In addition to the deflection of the soft arm due to the weights of all nodes, there exists an internal deflection (i.e., a buckling phenomenon) of the soft arm due to the mechanical coupling among modules. In order to prevent such internal deflection, a spring backbone is inserted into the center position of the spring. Fig. 1 shows that the whole structure does not have such deformation due to the additional backbone spring.

3.1.3 Actuator sizing
‐ Torque

 The torque required for the system operation can be derived from the screw mechanics depicted in Fig. 12.

Fig. 12. Screw mechanics

In Fig. 12, W, P, μ, λ, T, and de denote the axial direction weight (kgf), the radial force of screw (kgf), the friction factor of the surface of screw, the lead angle, the required torque (kgf․mm), and the efficient diameter of screw, respectively.

 Neglecting the dynamics of the system, the force equilibrium equation of the screw is given by

 Reformulation in terms of P gives

Since μ = tan ρ , we have 

Eq. (34) is simplified, by using tan , as

Thus, for the given W, the moment T of the screw is calculated as 

The pulling load (W) of the wire to prevent the deflection of the soft arm was 0.67 kg. Then, the minimum torque to drive the soft arm for a given specification (W : 0.67 kg, μ : 0.15, P : 1.25 mm, de : 6.8 mm (M8)) is calculated as 0.35kg․mm. However, considering the dynamic inertial load of the soft arm, energy loss due to friction, and the motor efficiency, the actual actuator size should be decided greater than this result. 

‐ Speed

Another important factor in the decision of actuator is the speed. For the given maximum operational speed of the soft arm, the maximum rpm of the motor is decided. 

The design procedure is as follows. First of all, estimate the fastest motion of the arm such as the surprise motion. Assuming that the time for the fastest motion is set as 1 second, the driving unit should be able to move the whole stroke 35mm of the screw within 1 second, too. Since the pitch of the screw is 1.25mm, the RPS (Revolutions per Second) of the motor can be calculated as 

Then, the RPM (Revolutions per Minute) of the DC motor is converted as 1,680 RPM. Same as the motor torque, the RPM should be decided with consideration of the safety factor. 

‐ Motor Specification

Based on the above analysis on the motor torque and the speed, a DC motor (Model : IG‐32PGM, D&G Motors) was selected. Considering the motor efficiency, the safety factors in terms of the motor torque and the speed are calculated, respectively, as 

A large safety factor is given to the torque size since the system should compensate for the dynamic inertial load due to the motion of the soft arm. 

4. Experimental Works

4.1 Bending Motion of the Soft Arm

Figure 13 shows the results of simulation and experiment when the soft arm is bent to 45°and 90° on the X‐Z and the Y‐Z plane, respectively. Table 4 shows the comparison of the simulation result and the experimental result. The error is caused by non‐uniformly deflection of each module and deflection due to gravity. The error can be minimized by kinematic calibration of the mechanism. However, the purpose of the proposed mechanism is an emulating of human gestures, not a precise positioning. Thus, using the control algorithm of the soft arm, we perform experiments in which the soft arm emulates several human gestures expressing human emotion and some service contents. 

Fig. 13. Results of simulation and experiment

Table 4. Comparison of simulation and experimental data

4.2 Emulation of Human Gestures

Figure 14 shows the picture of a dual soft arm. Using this device, we would like to emulate several human gestures expressing human emotion and some service contents. The driving unit and the mechanism are separated. Thus, a wide range of workspace can be ensured and a very fast motion can be obtained because of the small inertia of the arm. 

Fig. 14. Dual soft arm

4.2.1 HRI by gestures expressing emotion

Figure 15, 16, 17, 18 denote the gesture of the dual soft arm expressing “angry, “delight”, “sorrow”, and “surprise” emotion, respectively. 

Fig. 15. Gesture expressing “angry”

Fig. 16. Gesture expressing “delight”

Fig. 17. Gesture expressing “sorrow”

Fig. 18. Gesture expressing “surprise”

4.2.2 HRI by gestures expressing several service actions

Figure 19 and Fig. 20 show the gesture of the dual soft arm denoting “shaking hands” and “direction guide service”, respectively. 

Fig. 19. Shaking hands

Fig. 20. Direction guide service

4.2.3 Soft characteristic of the soft arm

The special feature of this soft arm is the spring‐backboned structure, which allows structurally flexible characteristic. Even in the case of collision with human, this device is able to absorb the impact. Fig. 21 shows that the soft arm deflects without giving damage to the human when it collides with the human body. 

Fig. 21. Flexible soft arm

4.3 System Implementation

The developed soft arm is implemented as the arm structure of a character service robot. The soft arm is used as the wing of the penguin service robot. This robot will be employed for the purpose of story‐telling to young children. Fig. 22 shows gestures expressing the four kinds of emotion and several service actions of the penguin service robot. 

Fig. 22. Gestures expressing emotions and several service actions of the penguin service robot

5. Conclusions

The contribution of this study is the suggestion of a new soft arm with a spring backbone. The driving unit and the mechanism are separated. Thus, a wide range of workspace can be ensured and a fast motion can be created because of the small inertia of the arm. Even in the case of collision with human, this device is able to absorb the impact due to the spring backboned structure. This soft arm can be employed as a means of HRI by emulating gestures expressing several kinds of emotion. This soft arm mechanism was successfully implemented as the wing structure of a penguin service robot.  


1.J.L. Burke, R.R. Murphy, and E. Rogers, V.J. Lumelsky, and J. Scholtz, "Final Report for the DARPA/NSF Interdisciplinary Study on Human–Robot Interaction," IEEE Transactions on Systems, vol.34,  no.2, pp.103–112, May, 2004.
2.M. Blow, K. Dautenhahn, A. Appleby, C.L. Nehaniv, and D.C. Lee, "Perception of Robot Smiles and Dimensions for Human‐Robot Interaction Design," in Proc. 15th IEEE International Symposium on Robot and Human Interactive Communication, pp. 469–474, Sept., 2006.
3.M. Hackel, S. Schwope, J. Fritsch, B. Wrede, and G. Sagerer, "A Humanoid Robot Platform Suitable for Studying Embodied Interaction," in Proc. IEEE/RSJ Intelligent Robots and Systems, pp. 2443–2448, Aug., 2005.
4.H.‐D. Yang, A.‐Y. Park, and S.‐W. Lee, "Gesture Spotting and Recognition for Human–Robot Interaction," IEEE Transactions on Robotics, vol. 23, no. 2, pp. 256–270, April, 2007.
5.T.P. Spexard, M. Hanheide, and G. Sagerer, "Human‐Oriented Interaction with an Anthropomorphic Robot," IEEE Transactions on Robotics, vol. 23, no. 5, pp. 852–862, Oct., 2007.
6.S. Kawamura, Y. Hayakawa, M. Tamai, and T. Shimizu, "A design of motion‐support robots for human arms using hexahedron rubber actuators," in Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, pp. 1520‐1526, 1997.
7.D.‐G. Choi, B.‐J. Yi, and W.K. Kim, "Design of a Spring Backbone Micro Endoscope," in Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, pp.1815‐1821, 2007.
8.A. Bicchi, G. Tonietti, "Fast and "soft‐arm" tactics [robot arm design]," IEEE Robotics & Automation Magazine, vol. 11, pp. 22‐33, June, 2004.
9.J.‐J. Park, B.‐S. Kim, J.‐B. Song, and H.‐S. Kim, "Safe Link Mechanism based on Passive Compliance for Safe Human‐Robot Collision," in Proc. IEEE Int. Conf. on Robotics and Automation, pp. 1152‐1157, 2007.
10.J.‐J. Park, J.‐B. Song, and H.‐S. Kim, "Safe Joint Mechanism based on Passive Compliance for Collision Safety," in Proc. IEEE Int. Conf. on Advanced Robotics, pp. 473‐478, 2007
11.M. Zinn, O. Khatib, B. Roth, and J.K. Salisbury, "Playing it safe [human‐friendly robots]," IEEE Robotics & Automation Magazine, vol. 11, pp. 12‐21, June 2004.
12.Y. Akamatsu, T. Nakamura, and Y. Kusaka, "Development of a soft manipulator with flexible joints using smart fluid and Pneumatics cushion for collision with human," in Proc. IEEE/ASME Int. Conf. on Advanced Intelligent Mechatronics, pp. 1‐6, 2007.
13.S.‐S. Yoon, S. Kang, S.‐J. Kim, Y.‐H. Kim, M. Kim, and C.‐W. Lee, "Safe arm with MR‐based passive compliant joints and viscous‐elastic covering for service robot applications," in Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, pp. 2191‐2196, 2003.
오늘하루 팝업창 안보기 닫기