A Tendon-Driven, Preloaded, Pneumatically Actuated, Soft RoboticGripper with a Telescopic Palm

Jiawei Meng, Lucas Gerez, Jayden Chapman, and Minas Liarokapis

Abstract— In this paper, we present a highly adaptive, 3-fingered, soft robotic gripper with a reasonably low weight (657g) that is capable of grasping various everyday life objects. Theemployed soft, pneumatically actuated robotic fingers are ableto grasp objects of different shapes without damaging them,and the extendable, soft, telescopic palm is able to absorb thehigh impact forces of the grasped objects, protecting the on-board sensor and providing adaptation to the object shape.The gripper is tendon-driven and it employs a quick-releasemechanism that allows grasping of objects at high speeds. Thedeformability of the soft robotic fingers and the telescopic palmis evaluated by tracking their motions with an appropriatemotion capture system. The proposed gripper facilitates theexecution of smooth pre-contact motions. The high workingvolume of each finger allows for grasping of objects with avariety of shapes and sizes. The behaviour of the gripper andits performance are experimentally validated with graspingexperiments that involve a plethora of everyday life objects.

I. INTRODUCTION

Currently, numerous studies concentrate on designing softrobot hands with pneumatic mechanisms [1]–[4], or adaptiverobot hands with tendon-driven mechanisms [5], [6] fordifferent kinds of every day grasping tasks. Soft and adaptive(underactuated and/or compliant) grippers are of low-costand low-complexity, and they provide simplified solutionsfor the execution of robust grasping tasks with everyday lifeobjects by employing a minimal number of actuators. Bycombining best practices from both fields, we are able tomaintain the benefits of both sides while optimizing the totalnumber of the pneumatic actuators and motors on the pro-posed gripper. More specifically, we focus on a new, hybridtype of robot hand which is lightweight, easy-to-control, andrapidly actuated by integrating pneumatic and tendon-drivenmechanisms. We also investigate the incorporation of softtelescopic mechanisms in the palm module and how theyincrease conformability to the object shape.

Compared with rigid-bodied robot hands, soft-bodiedrobot hands are safer when interacting with everyday objectsin unstructured and dynamic environments. Additionally,the lack of rigid elements helps soft-bodied robot handsto conform to objects with relatively larger contact areas,improving grasping performance [7], [8]. Whilst progress hasbeen made within the field of soft robot hands in laboratoriesacross the world, relatively little work exists on combiningsoft components such as pneumatic fingers and palms withtendon-driven mechanisms.

J. Meng, L. Gerez, J. Chapman, and M. Liarokapis are with the NewDexterity, The University of Auckland, New Zealand. Email: {jmen218,lger871, jcha855}@aucklanduni.ac.nz, m.liarokapis@auckland.ac.nz

Fig. 1. A tendon-driven, pneumatically actuated, soft robotic gripperwith telescopic palms designed by the New Dexterity research team. Thethree soft fingers are preloaded and ready to grasp objects, while the twotelescopic mechanisms of the palm are inflated to allow better conformabilityto the grasped objects.

In this paper, we present a highly adaptive, 3-fingeredrobotic gripper that is capable of grasping various everydayobjects (e.g., cardboard boxes, plastic cups, and footballs).The proposed deformable soft finger is able to grasp objectsof different sizes and shapes without damaging them. Theextendable soft palm is able to absorb the impact forcesof oncoming objects (protecting the on-board sensor andproviding better grasping), and the tendon-driven systemthat is equipped with a quick-release mechanism allows thegripper to grasp objects with a reasonably high speed.

II. RELATED WORK

Robotic end-effectors evolved from simple parallel jawgrippers to complex robot grippers with multiple degreesof freedom (DOF) that are capable of reproducing a largevariety of grasps. Over the last decades, a new class ofadaptive robot hands has emerged, facilitating the executionof robust and stable grasps in unstructured and unknownenvironments. Such an advantage is due to a combinationof underactuation and inherited compliance that allows thesegrippers to conform to the object geometry [9]. Underactu-ated designs are usually simple to operate and control, andconsume less power as the number of motors is reduced.Thus, adaptive hands are affordable, lightweight, and oflow complexity, being an attractive option [10]. In [11], the

Fig. 2. The interaction between the gripper and quick-release mechanism:In state "1", the gripper is closed; in state "2", the gripper is opened.

authors proposed several open-source adaptive hand designsthat can be easily fabricated and replicated by the commu-nity. In [12], the authors optimized the design parametersof an adaptive tendon-driven robot hand based on geneticalgorithms by taking into account a series of geometricconstraints. In [13], the authors proposed a tendon-drivenfinger for an anthropomorphic robot hand, which mimics andreconstructs the anatomical model of a human finger.

However, adaptive robotic hands also have their limita-tions and drawbacks. More precisely, they exhibit limiteddexterity since they employ a limited number of actuatorsto control multiple DOFs independently. Also, since it isdifficult to estimate and control the contact forces appliedalong the robotic finger, they are not the most suitablesolution for grasping and manipulating delicate objects. Morerecently, following the path of intrinsic compliance in roboticdevices, soft grippers based on completely soft pneumaticactuators (SPAs) have received increased interest from theresearch community due to their flexibility, customizability,and environmental adaptability [14]. The popularity of soft,pneumatic actuators has increased over the last years afterthe dissemination of open source projects, such as theOpen Robotics Toolkit (a library of soft robotic designsfor researchers and enthusiasts) [15]. Soft robotic actuatorscan be used in many different environments to grasp awide range of objects. In [16], the authors describe anunderwater robotic gripper that employs soft fingers to graspand manipulate fragile species on the deep reef. In [17], theauthors use soft actuators composed of elastomeric materialswith integrated channels to produce bending motions that canconform with the human finger motion in the design of ahand rehabilitation glove. In [18], the authors propose a softrobotic gripper with a compliant structure for delicate sur-gical manipulation. Similar to tendon-driven adaptive robothands, SPAs also have several limitations and disadvantages.Pneumatic actuators are usually bulky, difficult to control andmanufacture, and suffer from leakages. For these reasons,we propose a hybrid approach combining characteristics oftendon-driven adaptive grippers and soft pneumatic actuatorsto design a highly adaptive, 3-fingered robotic gripper witha quick-release mechanism.

Fig. 3. Flowchart of the way that the proposed gripper operates in order tograsp an object. "D" is the distance between the object and the IR sensor.

III. DESIGNS AND FABRICATION

In this section, we present the design of the tendon-driven,preloaded, pneumatically actuated, soft robotic gripper. Wealso present the molding process of the soft parts of thegripper and we analyze how they respond to pressure.

A. Hybrid Soft Gripper

The overall structure of the proposed gripper in this papercan be divided into three main parts: the soft fingers, the soft,telescopic palm, and the gripper base, as shown in Fig. 1. Thegripper base contains a micro-controller (Robotis, OpenCM-9.04) to control a quick-release mechanism and monitorthe real-time data acquisition from the IR sensor, which isembedded on the top surface of the gripper base. Moreover,the tendons attached to the soft fingers are connected withthe wheel of the quick-release mechanism, as shown in Fig.2. Based on this design, the quick-release mechanism withthe wheel are able to open the soft gripper. It is worth notingthat the soft fingers and the soft palm are preloaded by twoseparate pressure supplies. The operation principle of theproposed gripper is shown in Fig. 3.

B. Quick-release Mechanism

The quick-release mechanism is the main component ofthe tendon-driven system and is based on a previous designof an aerial grasper [19]. The interaction between the softgripper and the quick-release mechanism is shown in Fig.2. The three tendons of the soft fingers are connected tothe same wheel and their motions are coupled. In addition,the motor horn and the wheel rotate together when thequick-release mechanism is not triggered (the mechanismis engaged) and the wheel rotates freely (disengaged) whenthe mechanism is triggered. This allows the fingers to closerapidly. When the IR sensor detects an object at a distancelower than a threshold, the quick-release mechanism istriggered. Therefore, the object can be quickly grasped bythe proposed hybrid robotic gripper. As the quick releasemechanism requires little force to activate (to be triggered),a small Dynamixel motor with precise position control wasused for triggering, while a larger, more powerful motor wasused to drive the tendons.

Fig. 4. Motion recording of the telescopic, pneumatic soft mechanism with respect to heights: Left diagram shows the motion of the first layer; middlediagram shows the motion of the second layer and right diagram shows the motion of the third layer.

Fig. 5. The molding process of the telescopic, pneumatic soft palm: PartA and part B are made out of the same silicone material.

C. Telescopic, Pneumatic Soft Palm

The telescopic, pneumatic, soft palm proposed in thispaper is used to avoid damage to the IR distance sensorthat could be caused by a potential collision. The squarepalm has a side length of 25 mm and an initial height of 10mm, which is extendable up to 50 mm (total extension is 40mm). In order to manufacture the telescopic palm, the hybriddeposition manufacturing (HDM) technique is applied duringthe molding process [20]. To achieve greater flexibility anddecrease the manufacturing complexity, we selected a high-performance silicone rubber (Smooth-On, Dragon Skin 30).The multi-stage molding process of the telescopic, pneumaticsoft palm contains three main steps, as shown in Fig. 5.

Firstly, the bottom and middle chamber molds are con-nected so that silicone material can be poured into them.Once the air bubbles have been removed from the siliconematerial, the top half of the mold is inserted into the restof the mold with excess silicone overflowing out of themold where required. Meanwhile, silicone material is alsoused to attach part A and B together to obtain the completetelescopic palm. In order to empirically evaluate the expan-sion performance of the telescopic, pneumatic soft palm, anexperiment was conducted by recording the motion of the

Fig. 6. Three different states of the soft palm: a) shows the top view oftelescopic mechanism’s initial state, b) shows the side view when the secondlayer has been extended, and c) shows the side view when the third layerhas been extended.

different layers of the telescopic mechanism, as shown in Fig.6. Markers were installed on each layer of the mechanism tohelp record the corresponding motion during the extensionprocess. Data was gathered for five trials involving a widerange of pressures from 0 to 35 kPa, as shown in Fig. 4.Based on our observations, we notice that the height of thefirst layer of the telescopic mechanism is relatively stableduring the entire inflation process. The heights of the secondlayer and the third layer of the soft, telescopic mechanismexperience significant displacements. At 15 kPa (between 40s and 60 s in Fig. 4), the second layer extends ∼20 mm, andat 35 kPa (between 100 s and 120 s in Fig. 4), the third layerextends ∼15 mm. The total extension of the third layer is40 mm (between 0 s and 130 s in Fig. 4).

D. Plastic-reinforced Finger

The plastic-reinforced finger that we used in this designis based on the design described in [15]. It contains twolayers (inner skin and outer skin), a low stretch braid that iswrapped around the inner skin from the top the to bottom,and a 3D-printed inextensible plastic layer that is embeddedbetween the inner skin and outer skin. The length of thefinger is 150 mm and the radius of the semicircular surfaceat both ends of the finger is 15 mm.

The HDM technique is also used during the molding pro-cess of the plastic-reinforced finger [20]. High-performancesilicone rubbers (Smooth-On, Dragon Skin 30 and DragonSkin 10) were used as the outer and inner skin mate-rial, respectively. Different materials have been used for

Fig. 7. The molding process of the plastic-reinforced finger: In the end,before obtaining the entire plastic-reinforced finger, the two ends of theouter skin of the finger are sealed by using the same material.

manufacturing the finger as the inner skin needs to havehigher stiffness to resist high pressures and the outer skinneeds to have higher extensibility under a similar bendingangle. The multi-stage molding process of the soft fingercontains four main steps, as shown in Fig. 7. Firstly, theinner skin is molded onto a 12.5 mm diameter rod. Oncethe inner skin has cured, the inextensible 3D-printed plasticlayer is placed on the skin with silicone glue. Then, lowstretch braid is wrapped around the skin with the purposeof reducing the lateral expansion of the finger normal to thedesired curvature. Finally, the outer skin is molded aroundthe existing structure.

In order to evaluate the bending performance of the plastic-reinforced, pneumatically actuated, soft finger, an experimentwas performed recording its motion as the pressure wasincreased. Similar to the performance evaluation of the palm,data was gathered for five trials of responses to variouspressures, from 7 to 84 kPa with a ten second interval. Thebending profile of the finger is shown in Fig. 8. Furthermore,this finger is able to bend approximately 270 degrees basedon our observations of Fig. 9.

IV. GRASPING EVALUATION

In order to evaluate how well the proposed gripper de-sign performs grasping tasks, a robot arm (UR5, UniversalRobots) was used and the developed gripper was installedon its end-effector in order to perform various experiments.Table I shows objects of various dimensions, weights, shapes,and materials that were used during these experiments.

A. Daily Grasping Tasks

The vertical grasping capability of the proposed gripperwas first tested by using it to grasp various everyday objectsand to hold them stably in the air for approximately fiveseconds, as shown in Fig. 10. Furthermore, each object from

Fig. 8. Bending profile of the plastic-reinforced, pneumatically actuated,soft, robotic finger. The lines are used to denote the links of the finger andthe circles denote the joints.

Fig. 9. Two different states of the soft finger. The left subfiguredemonstrates its initial state and the right subfigure demonstrates its fullybent state.

Fig. 10. Grasping experiments: From subfigure a) to f), the gripper isgrasping a red cardboard box, a Starbucks cup, a yellow plastic bottle, ayellow baseball, a red football, and a potato chips can.

the object set presented in Table I was grasped ten timesso that the average grasping time could be calculated. Inthese experiments, all the target objects were grasped by theproposed gripper with a high success rate and short averagegrasping time (see Table II).

TABLE IOBJECT SET

Object Dimension (mm) Weight (g)Paper box 215x60x160 56.46

Starbucks cup 70x143 16.28Plastic bottle 200x100x55 46.31

Baseball 95 173.26Football 150 150

Paper can 80x235 51.13

Note: The objects that have only one value in the dimensionsection are spheres, and the objects that have two values in thedimension section are cylinders (the first value represents thediameter of the top and bottom surfaces and the second valuerepresents the height).

TABLE IIAVERAGE TIME REQUIRED FOR GRASPING

Object Time (s)Paper box 0.46

Starbucks cup 0.61Plastic bottle 0.70

Baseball 0.65Football 0.34

Paper can 0.41

Fig. 11. Impact absorption capability tests with objects of different shapes.Subfigures a) and c) show that the soft telescopic mechanisms with twolayers extended successfully prevents the target objects from damaging theIR distance sensor at the beginning and the end of the grasping process.Furthermore, subfigures b) and d) show that the soft telescopic mechanismwith three layers extended has the same capability as well as the abilityto act as an assistant finger at the beginning and the end of the graspingprocess.

B. Impact Absorption Capability

The impact absorption capability of the soft telescopicmechanism was evaluated during grasping. A red cube anda yellow polyhedron were used during the experiments, asshown in Fig. 11. Subfigures a) and c) demonstrate that whenthe soft telescopic mechanisms have two layers extended theysuccessfully prevent the target objects from damaging the IRdistance sensor at the beginning and the end of the graspingprocess. Subfigures b) and d) demonstrate that when the softtelescopic mechanisms have three layers extended they havethe same capability as well as the ability to act as extrafingers during grasping.

C. Video Demonstration

A video presenting the soft robotic gripper and the exper-iments conducted can be found at the following URL:

www.newdexterity.org/softaerialgrasper

V. CONCLUSIONS AND FUTURE WORK

In this paper, we presented a highly adaptive, 3-fingered,soft robotic gripper that is capable of grasping variouseveryday objects. The use of a compliant soft finger allowedgrasping of objects of different shapes, sizes, and materialswithout damaging them. The extendable, soft, telescopicmechanisms are able to absorb the impact forces of oncomingobjects, protecting the on-board IR distance sensor andproviding better grasping conformability. The experimentsalso demonstrated that the tendon-driven system combinedwith the quick release mechanism was able to help thegripper to grasp the objects with reasonably high speedswithout damaging them.

Regarding future directions, we plan to improve the man-ufacturing process of the soft structures, decreasing thenumber of steps required to fabricate them, and reducing thestress concentration regions to avoid leakages in the system.We also plan to optimize the position of the IR distancesensor and the positions of the soft telescopic mechanismsto better grasp objects.

REFERENCES

[1] F. Ilievski, A. D. Mazzeo, R. F. Shepherd, X. Chen, and G. M. White-sides, “Soft robotics for chemists,” Angewandte Chemie InternationalEdition, vol. 50, no. 8, pp. 1890–1895, 2011.

[2] Y. Hao, Z. Gong, Z. Xie, S. Guan, X. Yang, Z. Ren, T. Wang,and L. Wen, “Universal soft pneumatic robotic gripper with variableeffective length,” in 2016 35th Chinese Control Conference (CCC),pp. 6109–6114.

[3] Y. Kuriyama, Y. Okino, Z. Wang, and S. Hirai, “A wrapping gripperfor packaging chopped and granular food materials,” in 2019 IEEEInternational Conference on Soft Robotics (RoboSoft), pp. 114–119.

[4] S. Li, J. J. Stampfli, H. J. Xu, E. Malkin, E. V. Diaz, D. Rus, andR. J. Wood, “A vacuum-driven origami “magic-ball” soft gripper,” in2019 International Conference on Robotics and Automation (ICRA),pp. 7401–7408.

[5] R. R. Ma, L. U. Odhner, and A. M. Dollar, “A modular, open-source 3dprinted underactuated hand,” in 2013 IEEE International Conferenceon Robotics and Automation, pp. 2737–2743.

[6] G. Gao, A. Dwivedi, N. Elangovan, Y. Cao, L. Young, andM. Liarokapis, “The new dexterity adaptive, humanlike robot hand,”in IEEE International Conference on Robotics and Automation, 2019.

[7] J. Shintake, V. Cacucciolo, D. Floreano, and H. Shea, “Soft roboticgrippers,” Advanced Materials, vol. 30, no. 29, p. 1707035, 2018.

[8] J. Hughes, U. Culha, F. Giardina, F. Guenther, A. Rosendo, and F. Iida,“Soft manipulators and grippers: a review,” Frontiers in Robotics andAI, vol. 3, p. 69, 2016.

[9] A. M. Dollar and R. D. Howe, “The highly adaptive sdm hand:Design and performance evaluation,” International Journal of RoboticsResearch, vol. 29, no. 5, pp. 585–597, 2010.

[10] A. G. Zisimatos, M. V. Liarokapis, C. I. Mavrogiannis, and K. J.Kyriakopoulos, “Open-source, affordable, modular, light-weight, un-deractuated robot hands,” in 2014 IEEE/RSJ International Conferenceon Intelligent Robots and Systems, pp. 3207–3212.

[11] R. Ma and A. Dollar, “Yale openhand project: Optimizing open-sourcehand designs for ease of fabrication and adoption,” IEEE Robotics &Automation Magazine, vol. 24, no. 1, pp. 32–40, 2017.

[12] H. Dong, E. Asadi, C. Qiu, J. Dai, and I.-M. Chen, “Geometric designoptimization of an under-actuated tendon-driven robotic gripper,”Robotics and Computer-Integrated Manufacturing, vol. 50, pp. 80–89, 2018.

[13] S. Shirafuji, S. Ikemoto, and K. Hosoda, “Development of a tendon-driven robotic finger for an anthropomorphic robotic hand,” TheInternational Journal of Robotics Research, vol. 33, no. 5, pp. 677–693, 2014.

[14] Y. Sun, Y. S. Song, and J. Paik, “Characterization of silicone rubberbased soft pneumatic actuators,” in 2013 IEEE/RSJ InternationalConference on Intelligent Robots and Systems, pp. 4446–4453.

[15] D. P. Holland, E. J. Park, P. Polygerinos, G. J. Bennett, and C. J.Walsh, “The soft robotics toolkit: Shared resources for research anddesign,” Soft Robotics, vol. 1, no. 3, pp. 224–230, 2014.

[16] K. C. Galloway, K. P. Becker, B. Phillips, J. Kirby, S. Licht, D. Tcher-nov, R. J. Wood, and D. F. Gruber, “Soft robotic grippers for biologicalsampling on deep reefs,” Soft robotics, vol. 3, no. 1, pp. 23–33, 2016.

[17] P. Polygerinos, S. Lyne, Z. Wang, L. F. Nicolini, B. Mosadegh, G. M.Whitesides, and C. J. Walsh, “Towards a soft pneumatic glove forhand rehabilitation,” in 2013 IEEE/RSJ International Conference onIntelligent Robots and Systems, pp. 1512–1517.

[18] J.-H. Low, I. Delgado-Martinez, and C.-H. Yeow, “Customizable softpneumatic chamber–gripper devices for delicate surgical manipula-tion,” Journal of Medical Devices, vol. 8, no. 4, p. 044504, 2014.

[19] A. McLaren, Z. Fitzgerald, G. Gao, and M. Liarokapis, “A passiveclosing, tendon driven, adaptive robot hand for ultra-fast, aerial grasp-ing and perching,” in 2019 IEEE/RSJ International Conference onIntelligent Robots and Systems (IROS), pp. 5602–5607.

[20] R. R. Ma, J. T. Belter, and A. M. Dollar, “Hybrid DepositionManufacturing: Design Strategies for Multimaterial Mechanisms ViaThree-Dimensional Printing and Material Deposition,” Journal ofMechanisms and Robotics, vol. 7, no. 2, 05 2015.