IROS2019国际学术会议论文集2151

Robot Finger with Remote Center of Motion Mechanism for Covering Joints with Thick Skin Chincheng HSU1, Alexander SCHMITZ1, Kosuke KUSAYANAGI1and Shigeki SUGANO1 AbstractAn end-effector such as a gripper or multi- fi ngered hand is essential to enable robots to grasp and manipulate objects of various size and shape. Soft skin increases the grasp stability and can provide space for tactile sensors. However, covering the joints with skin is challenging, typically causing a considerable surface area of multi-segment robot fi ngers not to be covered by skin. This also creates the risk that objects get pinched in the joints when fl exing the fi ngers. The current paper suggests using a remote center motion (RCM) mechanism to move the center of joint rotation to the surface of a thick skin layer. In particular, a 6-bar mechanism is used. Thereby, a thick soft skin layer with a continuous surface can be realized. Furthermore, adaptive joint coupling with linkages is implemented. In the current paper a 2-fi ngered gripper is realized, and objects of various size and shape are grasped (from thin paper to objects of 135 mm diameter). The current gripper was manufactured with 3D-printed material to enable rapid prototyping, therefore the payload was limited to only 1 kg for this version. Overall, this paper shows the feasibility of an RCM for a robot fi nger and discusses the benefi ts and limitations of such a mechanism. I. INTRODUCTION For many industrial tasks simple parallel or rotational grippers are suffi cient, but robot grippers and hands are increasingly expected to act in unstructured environments and with a wide variety of objects, and fi ngers with multiple links can adapt better to various object shapes. In addition to fi ngers with multiple links, soft and sensitive skin with tactile sensors can aid the reliable grasping of various objects. In particular, soft skin makes the interaction safer and more robust, and tactile skin sensors provide the most direct information about the contact with the grasped object. However, the integration of soft skin is especially chal- lenging at the fi nger joints 1. In particular, sensitive robot skin is typically not stretchable or even bendable like human skin, which limits the coverage of the joints with skin. Therefore, joints in robots typically have a void in the cover between the links to enable the closing of the fi ngers. For example, in the past our laboratory covered all phalanges of the Allegro Hand with the uSkin sensors, which provide distributed 3-axis measurements integrated in 4 mm thick skin 2. To allow the fi nger fl exion despite of the coverage with uSkin, the already existing gaps between the fi nger phalanges had to be extended each by a distance equal to This research was supported by the JSPS Grant-in-Aid for Scien- tifi c Research No. 19H02116 and No. 19H01130. Corresponding author: Chincheng Hsu 1 TheauthorsarewiththeSuganoLab,Departmentof ModernMechanicalEngineering,SchoolofCreativeScience andEngineering,WasedaUniversity,Tokyo,Japan.contact: chincheng hsufuji.waseda.jp Fig. 1. Prototype gripper with fi ngers that use an RCM mechanism and coupling between the joints. the skin thickness (4 mm), adding in total 12 mm to the 144 mm length of the 3 degree of freedom (DOF) fi ngers. Similarly, the TWENDY-ONE hand was covered with 241 tactile sensors covering most of the palmar surface of the hand, but the joints were not covered with tactile sensors or soft skin 3. Sensorized gloves have been suggested 4, but they need to be thin, which limits the achievable softness and the type of sensors that can be integrated, especially at the joints. In particular, the glove in 4 does not cover the joints with sensors. In the current paper we suggest using a joint mechanism that enables the complete coverage of the palmar fi nger surface with soft material, see Fig 1. In particular, instead of a simple rotary joint we use a 6-bar mechanism, which creates the pivot point at the skin surface. Therefore, the side of the fi ngers that is in contact with the object can be covered with a thick, soft material and a continuous surface layer, which has the following benefi ts: Complete coverage with soft skin increases the grasping stability and the safety both for the robot and the grasped object, because the contact is guaranteed to happen with soft material. There is no principal limit for the thickness of the soft material that can be realized. There is no risk of pinching or squeezing the grasped object in the joints between the fi nger segments when fl exing the fi ngers. 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) Macau, China, November 4-8, 2019 978-1-7281-4003-2/19/$31.00 2019 IEEE3172 Tactile sensors could be integrated in the thick skin, and the continuous sensor coverage would enable to get more information about the grasped object. A higher number of tactile measurements could be exploited for example with deep learning. Furthermore, there is no risk that a contact goes unnoticed because it happens at an area that is not covered with sensors. The tactile sensors do not need to be curved to cover the joints, further easing the possible integration of sensors. The surface area, and therefore the potential contact and sensing area, does not get smaller when the fi nger fl exes. We also integrate a mechanism that allows the adaptive joint coupling between the two fi nger joints, which enables the hand to passively adjust to various object shapes. We thereby demonstrate that an RCM mechanism can be combined with adaptive joint coupling. In the current paper we implement a 2-fi ngered robot end effector, however, we suggest that it would be possible to use the mechanism in the future also for multifi ngered hands. While the fi nger length (123 mm with 2 DOF) is longer than a typical human fi nger, the joint mechanism allows for smaller designs as well. Furthermore, in this paper we use thick skin, but do not integrate tactile sensors in it. However, the integration of tactile sensors like uSkin would be possible 5. The current paper therefore focuses on the presentation of the joint mechanism, and demonstrates the capability to integrate thick skin at the joints. We present the joint coupling mechanism and show the grasp adaptiveness with 1 or 2 fi ngers for various object sizes and shapes. We presented a fi rst prototype fi nger with RCM joints in 5, integrated tactile sensors and showed the continuous sensing ability, but used sliders instead of a 6-bar mechanism, did not implement adaptive joint coupling, and did not inte- grate the fi nger in a gripper or perform grasping experiments. The rest of this paper is organized as follows. In Section II we review related robot grippers and hands. Section III introduces the conceptual design of our robot gripper and its features. Section IV presents the experimental procedure that was used to evaluate the gripper and shows the results. Section V draws conclusions and presents future work. II. RELATEDWORKS Many grippers and multifi ngered robots have been de- signed 67, and they employ various actuation techniques. The actuators can be placed directly in the fi ngers or remotely in the palm or forearm. Various transmission techniques have been used, including tendons, linkages and cams, various gears, belts, twisted strings and fl exible shafts. If the actu- ators are placed directly in the fi ngers, the actuator can be either in the joint 3 or in one of the links 2, and rotary or linear actuators can be used. Placing the actuator close to the joint increases the size and mass of the critical distal phalanges, but decreases the transmission chain complexity. Underactuation is commonly employed, especially for the distal and proximal interphalangeal (DIP and PIP) joints, similar to humans. Underactuation in general creates chal- lenges for dexterity, but has the benefi t of passive shape adaptation. It can be achieved in various ways, for example with tendons or with linkage driven mechanisms, with or without springs. Underactuation is discussed for example in 8910. To realize ordinary revolute joints in robot hands, rotary bearings at the joints are used, and they function in a reliable and space effi cient way. Furthermore, elastic elements like springs have been used as joints 1112. Moreover, another type of non-revolute joint with contact-aided surfaces has been used 13141516. However, all these joints create challenges for the coverage with skin, as the surface area changes with the joint angle. It is not feasible to completely cover these joints with a thick, fl at, continuous skin. Regarding tactile sensors 1, thin tactile sensors exist and have been integrated for example in the Gifu hand 17, which could help to cover the hands with sensors without increasing the fi nger length. Furthermore, stretchable sensors exist 1819, and have been used to cover the joints of arms. However, even with thin, stretchable sensors it seems to be challenging to cover small fi nger joints 4. Furthermore, thicker sensors like uSkin could provide more information such as distributed force vectors, and thicker soft material is benefi cial for increased compliance even without tactile sensors. Recently, there has been an increasing number of soft robotics grippers and hands 202122, which unlike tradi- tional robots do not have rigid links and defi ned joints, and therefore are highly compliant and have an excellent ability to passively adapt to the grasped object, making them an attractive choice for many applications. However, precise and strong manipulation is challenging with soft grippers, as well as the coverage with dense tactile sensors like uSkin. RCM mechanisms 23 can achieve pivot points outside the mechanical joint structure. They have been used for example for surgical needle-insertion devices 24 and for exoskeletons 2526, for which it is crucial that the ex- oskeletons center of rotation matches the humans one. A straightforward method is to use a circular slider to im- plement a circular-prismatic joint 23, and the exoskeleton in 25 used a circuitous joint. Both these mechanisms use sliders, which are challenging to produce in a sturdy and space effi cient way. Another way to implement an RCM mechanism is with a 6-bar 2326, which relies only on common rotary bearings. Using an RCM mechanism for robot fi ngers could achieve the benefi ts discussed in the introduction, but to the best of the authors knowledge has previously not been used for robot hands or grippers. III. HARDWAREDESCRIPTION This section will describe the mechanical design of the prototype fi nger and gripper. In particular, the RCM mecha- nism and passive coupling mechanism will be explained. The overall size of the fi nger was designed to roughly correspond to the one of the Allegro hand from Wonik Robotics, a robot hand used in our lab, so that in future work the performance of the Allegro hand and our gripper can be compared. Please also note that our Allegro hand is 3173 Fig. 2. Design of robot fi nger. The coupling mechanism is shown in orange and the RCM mechanism in yellow (PP = proximal phalange, MP = middle phalange, DP = distal phalange). The red dots show the location of the virtual proximal and distal interphalangeal joints (PIP and DIP). instrumented with tactile sensors, making our Allegro hands fi ngers longer and wider than the standard hand 2. A. Finger Design with RCM The conceptual design of the grippers fi nger can be seen in Fig. 2. The distal phalange is 43 mm and the medial phalange is 48 mm long. The proximal phalange houses the motor and will not move in our gripper design. As a result, each fi nger has two DOF. The part of the proximal phalange covered with soft skin for the implementation in the gripper is 32 mm long. The width of the fi nger is 22 mm. Fig. 2 shows that the center of rotation for both the virtual PIP and DIP joint are at the surface of the soft skin layer. For each virtual joint, two 4-bar parallelograms, connected together, form a 6-bar mechanism. The shorter links of the 6-bar mechanism are physically incorporated as part of the fi nger housings. The center of rotation is at the point where those links would intersect if they were extended. Thereby, by adjusting the placement of the corresponding bearings, the fi nger can easily be designed for different thicknesses of the soft skin. In the current implementation, the soft skin layer is 5 mm thick. The longer links of each parallelogram are designed to touch each other when the joint is bent 90 deg, and therefore act as a hard limit. The more the parallel links align, the more the mechanism loses structural stiffness, practically limiting the achievable range of motion for such a joint. In the current implementation we limit the rotation to 90 deg, but also about 120 deg would be achievable with good structural stiffness. More complex RCM mechanisms can achieve larger joint ranges, but we consider 90 deg to be suffi cient for our current version. Fig. 3. (a) Links of the mechanism (not including the fi nger housings, which also act as 2 links). (b) Diagram of the 6-bar RCM mechanism. To prove that each 6-bar mechanism has 1 DOF, we use Grublers equation of mobility for a planar linkage system: M = 3(L 1) 2J(1) In this equation M is the mobility (the number of DOFs), L is the number of links, J is the number of joints in the mechanism. Fig. 3 is the diagram of the 6-bar RCM mechanism. Fig. 3(a) shows the links attached to the two phalanges separately. Each of the letters A-G represents a rotary joint of the linkage. In both groups you can see the letters B, C, and E, as they represent the same joint, as shown in Fig. 3(b). The left side of Fig. 3(b) shows the extended state, and the right side shows the fl exed one. The linkage includes 6 links, which are A-C, D-E, B-F, C-G, A-D and F-G. A-D and F-G are formed by the housings of the corresponding phalanges. Furthermore, there are 7 joints in the mechanism. Thus, the mobility equals to 1, proving that each of the 6-bar mechanisms forms 1 DOF. The red dot shows the location of the virtual joint corresponding to this 1 DOF. We place two 6-bar mechanisms in each fi nger, one for each fi nger joint (DIP and PIP), enabling 2 DOF for each fi nger. Two more 6-bar mechanisms are placed in parallel to the fi rst two, on the other side of the fi nger (see Fig. 6), which does not infl uence the number of DOF of the fi nger, but increases the structural stiffness. B. Actuation and Coupling Mechanism The fi nger is underactuated, and the two virtual joints are coupled with linkages, as shown in Fig. 4. Thereby, 3174 Fig. 4.The actuation mechanism is presented. The linkage converts the linear motion to a rotary one, and also implements the adaptive coupling of the two joints. As shown in the lower image, the fi nger adapts passively to the object shape. the actuation force is distributed on the phalanges, and the robot can passively adapt to a wide variety of object shapes, as will be described below and demonstrated in Section V. In particular, the current paper demonstrates that a linkage- based coupling mechanism can be also used in combination with RCM joints. Fig. 5 shows the mechanism to transform the motion of the linear actuator to drive the two virtual RCM joints PIP and DIP. The Roman numbers I-VIII are the joints of the actuation and coupling mechanism. PP acts as the ground of the linkage. The position of joint I relative to PP is defi ned by the stroke of the motor, corresponding to the active DOF of the system. If the position of joint I relative to PP is defi ned, the system has 8 links (PP, MP, DP, I-II, II-III-IV, III-V, V-VI-VII, and VI-VIII) and 10 joints (I to VIII, PIP, and DIP). Using Equation (1), the mobility of the system is 1, representing the passive DOF of the system. In particular, joints PIP, III, IV, V and VII form a 5-link system with 2 DOF, one of which is active (the location of III), and one passive (the coupled location of V and VII). Commonly, 4- bar linkages are used for the passive coupling of joints, but the RCM mechanism necessitates that the location of the virtual joint PIP is different to the real joint IV, which does not infl uence the kinematics however, as both IV and PIP Fig. 5. The actuation mechanism: (a) links of the mechanism (not including PP, which acts as the ground frame, MP and DP), (b) linkage diagram. Fig. 6. CAD design of the 2-fi ngered gripper. In the current implementation each fi nger is actuated by one individual motor. are fi xed respective to PP. In our current implementation, we do not implement a spring for neither the PIP nor the DIP joint, which could increase the necessary actuation force of one joint respective to the other, and thereby establish which of the two joints fl exes fi rst when no contact with an object is established. Therefore, only the friction determines with joint bends fi rst, and with our prototype the PIP joint fl exes fi rst during free- hand grasping, likely because it involves the motion of less and more proximal links. Flexing the PIP joint fi rst can be benefi cial for grasping various objects. To actuate each fi nger, we use one linear motor from iR Robot (MightyZAP, L12-20F-3). Each fi nger is equipped with one motor so that it can be actuated independently. This also allows us to quickly add fi ngers in the future, i.e. for a multi-fi ngered robot hand. Alternatively, one motor could also actuate both fi ngers. The motors are indicated in blue color in Figs. 2 and 4, and the specs are listed in Table I. The motors are housed in the proximal phalange, which is statically mounted to the 3175 base of the gripper, see Fig. 6. The proximal phalanges are 66 mm apart at the gripper base. The motor output shaft is connected to the linkage mechanism for t