水下机器人的机械手

ScienceDirectIFAC-PapersOnLine 48-2 (2015) 294–299Available online at2405-8963 © 2015, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.Peer review under responsibility of International Federation of Automatic Control.10.1016/j.ifacol.2015.06.048© 2015, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.Keywords: Underwater application, AUV, Manipulation, Robot hand, Hand design.1. INTRODUCTIONn the world of scientific research, there is a particular attention to innovative robotic manipulators for next-generation robots. These devices should be able to compete with the human hand in terms of functionality. This field of robotics has recently known a notable development, thanks tothe growing interest towards its multiple applications,including medical implants or automated devices capable ofreplacing human intervention in activities such asmanipulation and exploration in hostile environments.There are two types of robotic hands: humanoid and industrial. The former are based on the anatomical principles of the human hand, and are usually equipped with 2, 3 or 4 fingers, plus an opposable thumb. The latter are instead comparable to grippers for grasping objects, cutting cables, etc. n the last twenty years, many results - sometimes surprising - have been achieved, but the new generation of robotic hands requires a qualitative leap, both in terms of functional ability (dexterity) and application compatibility, i.e. reliability, constructive simplicity, low cost, reduced size and weight, and, in particular, the ability to operate in extreme environments.The Robotiq Robot Gripper, Schunk 2-Jaw Gripper (2013) and flexible robotic Gripper (Sam et al., 2008) models are a good example of industrial robotic hands suitable for performing repetitive tasks with extreme precision, as required in industries such as pharmaceutical research or food processing. The Barret Hand (William T. Townsend et al.,2000) and NU Hand (Kappassov et al., 2013) models are instead an example of anthropomorphic hands, more flexible and suitable for carrying out various tasks. I n fact, these models are able to grasp rigid and non-rigid objects, as they're equipped with multiple fingers. I n the case of the DLR-Hand (Liu et al., 2008), each finger is actuated by an independent mechanism. The motor transfers the motion viatendons, in a way that replicates the function of muscles and ligaments. The Gifu Hand III (Mouri et al., 2002) is instead equipped with 5 fingers - 4 DoF for the thumb and 3 DoF for the fingers. All 16 joints feature an independent motorization. The growing interest shown by oil & gas, military andoceanographic research industries for submarine technologieshas led to the development of new tools for improvingunderwater operations; among them, underwater endeffectors have emerged as indispensable devices for manyvehicles such as ROVs, AUVs, and so on. I n contrast to terrestrial applications, there are few references in the literature for underwater robotic hands (Lane et al., 1997; Gad at al., 2004; Meng et al., 2006; DFKI , 2010). Most models described in literature are simple end effectors equipped with a gripper; therefore they are classified as industrial hands. But if one takes into account the typical tasks to be carried out underwater, the scope of functions offered by industrial hands is quite narrow, since many applications require operations that currently can be performed only by anthropomorphic hands, such as turningvalves, grasping cables, or retrieving objects or archaeological finds. The "Three-fingered gripper" (Bemfica,2014) is an underwater anthropomorphic hand capable of performing the typical functions of this category. A commercially available device is the HDT's Adroit (HDT Global), used by the US Navy's Office of Naval Research (ONR). The main goal of this project was to minimize weight and size of the robot hand, in order to make it usable on small-sized ROVs (when coupled with robotic arms) and with neutral buoyancy. Starting from the study of the literature, the proposed solution presents some original concepts for theAbstract: This paper describes the design process and the construction stages of a multi-fingered robot hand prototype, named GUH14, to be used in underwater environment. The prototype was developed in collaboration by the University of Calabria and the University of Girona, with the goal of designing and building an underwater robotic hand that can be implemented on robot arms. This project is aimed to develop prototype solutions for grasping and manipulation operations in a submarine environment. The hand is composed by a palm and three independent fingers, each with three degrees of freedom (hereinafter, DoF), actuated by servomotors through hybrid transmission, tendons and gearing. The design and manufacturing procedures that allowed for the operation of the prototype will be discussed. This paper summarizes the obtained results.*University of Calabria – Italy (e-mail:[ francesco.spadafora, maurizio.muzzupappa, fabio.bruno]@unical.it)**University of Girona – Spain ( e-mail:[dribas, pere]@) Design and Construction of aRobot Hand Prototype for Underwater ApplicationsRobot Hand Prototype for Underwater Applications Spadafora, F.*; Muzzupappa, M.*; Bruno, F.*; Ribas, D.**; Ridao, P.**actuation mechanisms and the manufacturing process, featuring an intensive use of rapid prototyping techniques. This paper is organized as follows: the first part will introduce the design specifications and the construction technologies. I n the second part, we will describe the proposed design solutions, focusing on the gearing elements. The last section presents the construction and testing of the prototype.2. REQUIREMENTSThe project goal was the creation of an underwater anthropomorphic hand capable of being coupled with robot arms on small to medium size ROVs and AUVs.In order to achieve this goal, the hand was equipped with two fingers, plus an opposable thumb. This solution was intended to satisfy the functional requirements related to operations such as opening and closing valves, connect or disconnect connectors, activating and deactivating switches, recovering or laying cables and/or hoses on the seabed, and collecting objects - also of irregular shape - and samples of the seabed. At the same time, it was also possible to minimize as much as possible the problems related to waterproofing and space/weight requirements of the hand.n this phase of development of the first prototype, we decided to design a hand capable of operating up to 60 meters; so, we have defined materials and geometric dimensioning of the hand for this operative condition. Another very important issue was the neutral buoyancy of the joint in the water, in order to prevent any actual reduction of the force generated by the arm and to facilitate the manipulation of large and neutrally buoyant items. The selected materials have subsequently driven the choice for the manufacturing technologies to be employed for the various components of the hand. These techniques were: a) machining with conventional machine tools, and b) additive manufacturing (Kruth, 1998). The choice of the most suitable technology has been driven by the analysis the design constraints and the different functional characteristics for each component. t should be noted that the techniques described above are very different from each other, in terms of design approach, times and systems of production.In order to satisfy the requirements for the gripping of objects of any size and shape on the seabed, the hand has been designed to allow the fingers to flex on the palm. Therefore, the solution required 3DoF and an independent actuation for each finger. Finally, the implementation of the finger kinematics required two different transmission architectures typically used in terrestrial applications, i.e. "tendons" and gears, so as to ensure the maximum flexibility, a good clamping force, a greater reliability and simplicity of construction. It must be noted that while the use of a motor for each of the 3 joints ensures a greater control, the complexity of the joint is also increased, not to mention the risk of possible failures of the prototype if any water infiltration occurs. For this reason, the tendon transmission was adopted for the last two phalanges, which require no special precautions to avoid contact with the water, while the gearing has been implemented only on the first phalanx. The mixed solution proposed in this paper, based on servomotors, gearings and tendons, appears to be an original and cost-effective solution to solve the problems related to the articulation of the underwater robotic hand.3. HAND DESIGNFor the articulation of the first phalanx, the architecture of the specific finger was built inside the palm of the hand, featuring a driving gear (A) connected to the servomotor (B) that transfers the motion to a driven gear (C) by means of a belt (D); the drive torque, suitably amplified by a gearing (E), is transmitted to the articulation of the finger (see Figure 1).Fig. 1. Layout of the handFor the actuation system, the choice fell on programmable servomotors hitec hs5646, not specific to underwater applications, but capable of expressing a maximum torque of 12.5 kg*cm at 7V; this value can ensure a good tightening for each finger, and also a good speed of opening and closing (equal to 0.2 sec/60°). The chosen servomotors work in a range that varies approximately of ± 60°. Belts are used to couple servo motors and drive shafts for the transmission of power. They are made with elastomer reinforced by glass fibers and they are particularly suitable to ensure a correct kinematics and the flexibility of the fingers.For the articulation of the last two phalanges, a comparative analysis was conducted on tendons made of nylon and carbon fibre: the former has proven to be the best choice for ensuring a greater elasticity of the filament and minimizing the danger of breakage of the finger in case of accidental impacts. The filaments have been designed to pass through two ducts, located in the front part and in the back of the finger itself (Figure 2). I n this way, each tendon, while appropriately balancing the pulling forces, is able to open or close the phalanges.CDFig. 2. Cross-section of the fingerA kinematic simulation allowed for an appropriate design of the ducts where the tendons pass through (which were subsequently verified by physical tests, see sect. 4). 3D printing allowed us to construct the fingers with this particular configuration, so that each component is cheap, easily replaceable, lightweight, water-resistant and able to adequately protect the tendons. The ducts of the tendons are designed to allow a complete opening and closing of the last two phalanges when the gear, connected to the first phalanx, turns full. Although this solution doesn’t guarantee the stiffness to the system, it makes the system adaptable to grasp objects of different sizes and shapes. In figure 3, we show the theoretical maximum workspace of the hand: a cylinder with maximum diameter of 120mm and a box with a dimension of 220mm along the direction of gripping. An extensive experimental campaign will be carried out to define a specifictaxonomy about this.Fig. 3. Workspace of the handOnce the key elements of the kinematic mechanism havebeen defined, we moved on to design the geometry of thepalm, which led to the consideration of several factors. Thefirst issue was the optimization of the arrangement of thethree servomotors, the related gears and some calibratedshims able to function as belt tensioners. Also the appropriatedimensioning as a function of operating pressure was takeninto account, in order to assess both the structural strength ofthe components and the watertight integrity of the internalcomponents.To ensure the dry conditions inside the palm, both during thestatic phase of rigid displacement and the motion of thefingers, two different problems have been addressed: thestatic seal between the palm and the cover, and the dynamicseal in the link between the palm and the fingers. So wedetermined the exact dimensions of the grooves for thesealing components and the dimensions and materials of theO-rings to be used. In particular, we have used NBR (Nitrile)o-ring for static seal and bronze bulkheads equipped with adouble O-ring for dynamic seal.For the structural sizing, we have conducted a series of FEMsimulations that allowed for the identification of the areas tobe stiffened, both in the palm and in the cover. The prototypewas tested in first instance at low depths (6 bar), in order toavoid excessively burdensome operating conditions. For thisreason, a number of FEM analyses using Delrin wereconducted. This acetal resin is robust enough to ensure agood structural response at not very high pressures. Theresults show that the geometry does not present high stressvalues and the maximum deformation amounts to 0.15 mm.During the design of underwater components, it is importantto remember that the deformation of the component shall belimited, as strong deformations may result in a failure of theO-ring, leading to a complete or partial filling of water in thedevice.CNC (computer numerical control) machines have proven tobe suitable for the construction of the palm. The 3D printingmaterials, such as ABS o PLA, were not able to ensure agood strength and stiffness against the same pressure load: asa consequence, dry conditions inside the palm were notachievable.3.1. Gearing designA belt transmission with toothed pulleys, aimed to amplifythe motion (see Figure 4), was designed for transmitting themotion from the servomotor (1) to the shaft (5).In order to make the finger - the first phalanx in particular -complete a rotation of 120°, the gearing was sized by fixingthe distance between (1) and (4) to 40.12 mm, the diameter ofthe driving gear located downstream of the servomotor to24.5 mm, and the diameter of the driven gear to 16mm. Thetransmission ratio τ1 amounts to 1.53, so it is capable ofamplifying the motion of the servomotor, whose effectiverotation is ± 60°.Fig. 4. Transmission system of the handThe size of the belt chosen for actuating the mechanism has a pitch length of 144mm. Table 1 summarizes the data of the first kinematic mechanism.Tab. 1. Data of the first transmissionBelt transmission servomotor rotation 1 120° shaft rotation 4 183.6° τ1 1.53 With respect to the tendon transmission for handling and gripping the objects via the three fingers, the main problem has been the identification of the actual values of the force to be applied. Though the use of Nylon wires for the transmission of motion inside the finger can improve theeasiness of assembly of the fingers on the palm, it can't ensure an optimum tightening of the three phalanges (once the closure mechanism is actuated) because of its elasticity. For this reason, as mentioned earlier, we used a "hybrid"solution in order to make the mechanism suitable for flexingthe fingers on the palm. So the choice fell on a tendon actuation for the motion of the two last phalanges, and on a gear actuation for the first phalanx. This hybrid solution allowed for ensuring the desired flexibility and, at the same time, for gripping objects of different size and shapes. A series of experimental measurements was conducted to definethe actual length of the Nylon wires during opening and closing. The obtained results are shown in Table 2.Tab. 2. Experimental measurements of the transmissionGear transmission Finger α max rotation120°distance (mm) 20.86Servo motor β max rotation 120°Outer ΔL (mm) 13.8Inner ΔL (mm) 22A very accurate design of the transmission ratios between the three gears (no.6 of Figure 5) keyed on the axis (no.5 of Figure 4) and the three gears mounted on the first phalanx (no.7 of Figure 5) was carried out to respect the above mentioned operating parameters.Fig. 5. Transmission system of the first phalanxThe gearing connected to the palm (no.6 of Figure 4) is characterized by the presence of three gears: two have the same number of teeth (21) and module 1, and one presents a greater number of teeth, (26) always with module 1. Other three gears (no.7 of Figure 4) are present on the side of the finger; the central gear of the finger (with a smaller diameter,15 teeth and module 1) is free to slide within the two larger gears (21 teeth and module unit), which are integral with the first phalanx. The transmission ratio is equal to 1.73 for the outer gears and to 1 for the inner gear. In this way, in equalrotation of the entire finger, the nylon wire rotates 1.73 times faster around the provided tracks, which are integral with the rotation of the central gear (7). Thanks to the increased speed, it is possible to obtain a complete rewind and release of the two tendons, both for opening and closing the fingers. This configuration makes the articulation motion synchronized. Table 3 shows the values of the second transmission present in the hand. Tab. 3. Data of the second transmission Distance (mm) Driving gear diameter (mm) Driven gear diameter (mm) Transmission ratio 20.8626 15 1.73 21 21 1Moreover, this hybrid solution has allowed for a total independence of the three fingers and the entire palm, ensuring a good simplicity in the stages of assembling and disassembling, thus facilitating the maintenance works.3.2. Electronic design A control board, mounted on a watertight compartmentplaced at the wrist, actuates the fingers. The board consists ofa DC-DC converter, which powers the servos and theelectronics, a PI C16F1825 microcontroller, which generatesthe control signals, and a RS485 communication driver that receives the angular set-points from the vehicle control system. Each servo is controlled by a square signal with a 20ms period, where the duration of the high level state (between 0.9 and 2.1ms) determines its angular position(from -90 to 90 degrees). Three different rates of change of the high level state values have been defined to control the rotation velocities of the servos. 4. CONSTRUCTIONA number of production modalities based on CNC machines and 3D printers have been employed for the realization of the prototype, as mentioned in the previous section.I n order to define dimensions, assembly of all the components, functions and/or interferences of moving parts, all the PLA components were manufactured in first instance by means of the 3D FDM printer at the University of Girona, a RepRap BCN (Figure 6).A very important issue was the experimental assessment of the finger kinematics, since the presence of the nylon tendons did not allow for performing a virtual simulation to predict the actual response of the finger (CAD systems, in fact, are not capable of simulating effectively the behaviour offlexible elements).Fig. 6. Assembly testSpecifically, it was necessary to identify any possible interpenetration of the phalanges and evaluate the correctshape of the internal channels in the finger, so as to ensure its operation in any configuration. The entire system was testedby means of a special support that has allowed for checking the travel required for the tendons in opening and closing,and also the appropriate sizing of the gearing to which each finger is connected.Once the proper assembly of all the designed parts was checked, the first prototype of the palm was manufactured with a 4-axis CNC milling machine, starting from a block of Delrin. The connection parts and the seals have been made by turning bronze elements to ensure the sliding among rotatingparts (Figure 7).Fig. 7. Construction of some hand componentsFigure 8 shows the final prototype of the hand. Fingers aremade of PLA, while Derlin is used for the palm.Fig. 8. Final prototype5. TESTINGOnce all the processing stages have been completed, the subsequent reassembly of the hand allowed for conducting a series of tests under pressure, in order to validate the water-tightness of the entire device.The first tests were conducted in water tank at the University of Girona. The maximum depth of 5 meters has allowed for a first testing of the static tightness of the device (figure 9). As expected, the hand turned out to have a positive buoyancy in water, so it was ballasted with a weight of about 300 gr. In a subsequent test, the hand was mounted on a robotic arm of the Girona 500 AUV, in order to verify the completecompatibility of the connection.Fig. 9. Testing stage in water tank The second series of tests was conducted in static conditions,at a pressure of 6 bar, inside a hydrostatic test chamber at the University of Calabria. The last series of tests was carried outat a depth of 15 meters to validate the dynamic tightness of the hand (figure 10). At the moment, we have not measured the actual maximum force of the hitec hs5646 servos. Starting from a theoretical study, the maximum force applied on the finger found to be about 1.25 kg/cm. Experimentalmeasurements will be carried out to assess the real force acting on the prototype.I n order to avoid the loss of strength of device during the grasping phase, the servo motors must always power thefingers because the hand is backdrivable.Fig. 10. Testing in sea environment6. FUTURE WORKSThe design of the hand is a work in progress that requires more tests in order to finalize the prototype. At present, some initial goals of the project have been achieved: the hand has neutral buoyancy and a low manufacturing cost. We still have to conduct further testing of the link with the robot arm and to assess the functional requirements related to operations such as opening and closing valves, connect or disconnect connectors, etc.I n order to assess the forces actually provided by the servomotors and the related tightening torque, the hand will be equipped with load cells and pressure sensors placed on the fingers. Future plans include the incorporation of a set of current sensors, designed to detect the intensity peaks resulting from the finger blockage during a grasping, as well as the incorporation of tactile sensors.A future development is the design of the interface between the hand and the arm, so as to equip the entire system with a new wrist, substituting the current rigid one. A joint with one or more degrees of freedom would allow for a better management of the approach and manipulation of an object. Finally, a further development is the use of a 3D printing technique based on sintering of metallic powders for the production of components so that the hand can be assembled and tested on the arm of the "Girona 500" at an operating depth of 300 meters.. This would combine excellent mechanical properties with an easier neutral buoyancy in water.REFERENCESBemfica, J. R., Melchiorri, C. et al. (2014). A three-fingered cable-driven gripper for underwater applications. InRobotics and Automation(ICRA), EEE I nternational Conference, pp. 2469-2474.DFKI, http://robotik.dfki-bremen.de/en/research/robot-systems/seegrip-manipulator-1.htmlGad, R., Naik, G., and Aralgedad, N. (2004). A design of 2-dof gripper circuit for deep-sea objects. In OCEANS ’04.MTTS/IEEE TECHNO-OCEAN ’04, volume 3, pp.1402 –1409.Hdt global-aviable:/services/robotics/underwater/. Kappassov, Z., Khassanov, at al. (2013). Semi-anthropomorphic 3D printed multigrasp hand for industrial and service robots. I n Mechatronics and Automation (ICMA), IEEE International Conference, pp.1697-1702.Kruth, J-P., M. C. Leu, and T. Nakagawa. (1998). Progress in additive manufacturing and rapid prototyping. CIRPAnnals-Manufacturing Technology. 47.2, pp. 525-540. Lane, D. M., Davies, et al. (1997). MADEUS: advanced manipulation for deep underwater sampling. Robotics & Automation Magazine, IEEE, 4(4), pp. 34-45.Liu, H., Wu, K., Meusel, at al. (2008). Multisensory five-finger dexterous hand: The DLR/H I T Hand I I.In Intelligent Robots and Systems. IROS 2008. IEEE/RSJInternational Conference, pp. 3692-3697. Mouri, T., Kawasaki, et al. (2002). Anthropomorphic robot hand: Gifu hand III. In Proc. Int. Conf. ICCAS, pp. 1288-1293.Meng, Q., Wang, H., Li, P., Wang, L., and He, Z. (2006).Dexterous underwater robot hand: Heu hand ii. I n Mechatronics and Automation, Proceedings of the 2006 IEEE International Conference, pp. 1477 –1482. Robotiq - Adaptive Gripper 2-F I NGER model – 85 Aviable:/media/Robotiq-2-Finger-Adaptive-Gripper-Specifications.htmlSchunk. - 2-Finger Parallel Gripper WSG.Available:/en/produkte/produkte/2-fingerparallel-gripper-wsg.html (April 13). Sam, R., & Nefti, S. (2008). Design and development of flexible robotic gripper for handling food products.In Control, Automation, Robotics and V ision.ICARCV08. 10th I nternational Conference, pp. 1684-1689).William T. Townsend, (2000). The Barrett Hand grasper.Industrial Robot, Vol. 27(3), pp.181-188.。