IROS2019国际学术会议论文集An Approach of Facilitated Investigation of Active Self-healing System Oriented for Legged Robots

An Approach of Facilitated Investigation of Active Self-healing Tension Transmission System Oriented for Legged Robots Shinsuke Nakashima1, Takuma Shirai1, Kento Kawaharazuka1, Yuki Asano1, Yohei Kakiuchi1, Kei Okada1, and Masayuki Inaba1 AbstractSelf-healing robotics has been of considerable interest. We believe the function will have a major role in legged robots as a typical high-load application of robotics. Some pioneering works have been ongoing on self-healing soft robots. However, the development of large load self- healing component and its system integration with a life-sized legged robot is a challenging task. This study is to try the problem by a constructing self-healing component oriented for facilitated investigation. Proposed part enhances visibility and manufacturing by specializing tension transmission system. The developed module was evaluated by several experiments. First, healing visualization experiment was conducted to evaluate healing progress. In addition, the modules strength was tested using a motor-driven tendon module previously developed in our laboratory. Results of these experiments suggested that the stirring process have a major role in performing self-healing behaviour. Finally, we conducted a preliminary experiment on a tendon-driven legged robot. The experiment demonstrated that the module functioned in a real robot once. I. INTRODUCTION The current progress of legged robotics is intensive. Per- forming dynamic task requires persistent motion which can be done with the help of following approaches.: 1) mechanical structure and control for keeping robots from fracture 2) functional compensation for partial failure 3) healing of broken region Recent legged robots focused on 1 and 2, which made suc- cessful motion performance. Especially, 1 increased its valid- ity based on emerging high-performance actuation strategies. For instance, conventional stiff gear-driven robots 1 and hydraulic robots 2 acquire control-based active compliance. In addition, hardware compliant actuation scheme including SEA (Series Elastic Actuators) and VSA (Variable Stiffness Actuators) provide robots with the anti-impact performance 3. Furthermore, it is a quasi-direct drive that makes agile legged robots more simple, energy-effi cient, and prevalent 4. On the other hand, several trials intended for 2 have been demonstrated. For example, A robot that learns optimal motion pattern after fracture was developed 5. As for 3, self-healing technology provides legged robots with a safety net which gives facilitated trials in a real environment. This approach can push the current border of other approaches in the fi eld of legged robotics. Therefore, 1 S. Nakashima, T. Shirai, K. Kawaharazuka, Y. Asano, Y. Kakiuchi, K. Okada and M. Inaba are with Department of Mechano-Infomatics, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan snakashima at jsk.t.u-tokyo.ac.jp we set the target of this research is to defi ne another performance index, self-healing capability, and implement it to a legged robot prototype shown in Fig. 1. (Ideal) MeltingSolidification Robot Module Fracture Normal Normal (Damage)(Healing)(Recovery) Fig. 1.The research vision of self-healing legged robot. Self-healing behaviour comprises of several steps. Applying the self-healing principle to robotics is getting major in some emerging areas such as soft robotics 6. There are a few variable stiffness structures made from a complex of elastomer and low melting point alloy (LMPA) 7 8. Thermoplastic polymer is an alternative as phase change material to fabricate a self-healing component 9. Additionally, self-healing polymer has been exploited to fabricate self-healing actuator 10 and a robotic hand 11. Another unique approach features liquid circulation for mak- ing continuous reaction between the broken solid part and its uncured ingredients in the liquid state 12. The main focus of current self-healing robots is continuous healing. That is, they keep micro damages from growing to fatal fracture which separate the structure. On the other hand, self-healing after critical damage is of relatively minor intention because of some technical challenges. We attribute it to the diffi culty in system integration rather than material. We believe this type of healing becomes signifi cant for legged robots which have to deal with large impacts. The authors have been trying to tackle the problem by developing “Active self-melting bolt” which appends self- healing capability to bolt structure 13. The component is superior to conventional self-healing components with its initial strength, reaction speed, and self-sensing system. However, its work-intensive manufacturing process and in- visible structure have made it hard to evaluate and improve the component. Therefore, we developed another self-healing component for easier investigation without sacrifi cing its performance for a life-sized robot as possible. 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 IEEE2567 This paper is organized as follows. Sec. II describes the concept of active self-healing tension transmission system and the contribution of our work. Sec. III explains the proposed module improvement and difference between previ- ously developed ones. Sec. IV validates the proposed system with the observation of the self-healing process and tensile tests. Sec. V shows the result of the preliminary experiment on a life-sized tendon-driven legged robot. Sec. VI draws the conclusion and future works of the paper. II. ACTIVESELF-HEALINGTENSIONTRANSMISSION SYSTEM This section mentions the merits of self-healing tension transmission system for a legged robot. Next, it also explains our research contribution. A. Tendon-driven legged robot of life-sized scale Tension measuring part Wire fixed at Motor pulley Wire at end point Wire fixer bolt & nut Foot plate(bent) Another configuration Fig. 2.Landing impact breaks components of the tendon-driven system. Wire ropes break most easily and they protect other components from fractures. Especially, parts of ankle extensor muscle will out of order frequently. Fig. 2 shows a tendon-driven legged robot and broken parts after landing impact. A motor-driven tendon which rotates ankle joint in plantar fl exion motion, like an Achilles tendon, got out of order most often. Broken components are followings: 1) wire fi xed at motor-driven pulley 2) wire fi xed at robot link 3) aluminium axis inside the tension measurement unit 4) steel bolt for fastening wire at robot link Wire failures occur prior to that of robot link parts and other metal components. Because wire failure is easier to repair than other parts, tendon-driven actuation system itself enhances the availability of a robot system. Applying self- healing capability to tension transmission system leads the actuation system to function continuously without an external operator. In contrast to a motor-driven tendon, other components of a robot are hard to repair. One major example is removing a piece of the broken bolt from the threaded hole. Another example is a robot link because a lot of life-sized robots adopt custom-made machined ones from hard metal or CFRP aiming at large stiffness and strength per mass. These facts indicate that self-healing tendon-driven actua- tion can be an effective approach for a more available life- sized legged robot without fatal fracture. B. Related works and contribution of the research Similar motivation led researchers to investigations of various self-healing tension transmission systems 7 9 10. However, there have been some design challenges to develop self-healing system for high-load applications such as legged locomotion. We consider this as the tradeoffs between component evaluation and system integration. First, current self-healing robots with good healing perfor- mance extend its fi eld by gaining automatic decision making, software platforms and decreasing human intervention. On the other hand, we have developed “Active Self- melting Bolt” aiming at integration to a real robot. It was advantageous in initial strength, compact scale and the same control environment with the real robot. However, a highly integrated structure itself made the manufacturing pro- cess complex. Especially, electronics components including wiring and a microheater got rupture during the process. This problem made it diffi cult to improve and evaluate the component. This work is to solve these tradeoffs partially by intro- ducing an approach to facilitate the investigation of self- healing behaviour. Another signifi cant condition is to keep the modules performance as a tension transmission system for a life-sized legged robot. III. DESIGN APPROACH OF THE SELF-HEALING MODULE FOR FACILITATED INVESTIGATION In our previous works 13 14, the main objective was to develop the self-healing bolt. Hence, we developed the bolt structure with highly integrated electronics for self-healing and self-sensing function. Additionally, the bolt had micro- heaters in its center for the concentration of thermal input. Although the bolt has the potential of a self-healing joint for generic mechatronic systems, its current implementation is problematic. Especially, setting microheaters in the center causes the following diffi culties during the manufacturing process. 1) Molding integration of microheaters with accurate positioning required multi-step casting followed by shaping, post-processing, and assembly. 2) Heater wirings broke frequently during the shaping procedure of the threads. 3) The modules healed strength ratio was only 10%, and invisible bolted joint made it diffi cult to observe and evaluate the reaction process. Therefore, we developed another self-healing module for dealing with the problems. First, aluminium joints equipped with heaters made the manufacturing process more simple. 2568 That is, shaping and post-processing steps became unneces- sary. Next, cartridge heaters made the heater wirings sepa- rated from the LMPA part. Third, 2D bolt shape with side cover plates facilitated the inspection of the healing process. (B) Cartridge heater (C) Helical spring Neodymium magnet Heater wiring (Top)(Bottom) (Prototype model) After filled with low melting point alloy (A) Transparent cover (cross-section view) 4 mm 6 mm 10 mm (CAD)(Prototype) Tension transmission region mm2 Fig. 3.Overview of 2D bolt module. Capital letters state major differences from the previously developed modules. A. Components Fig. 3 shows the components of the proposed module. Although the basic components of the proposed module are similar to that of our previous works, its main differences are oriented for easier manufacturing and inspection of self- healing behaviour. 1) 2D bolt structure made from low melting point alloy (LMPA) part changes its phase below 50C. Phase transition reconnects broken solid part by melting. 2) Aluminium joint part connects LMPA 2D bolt part and motor-driven tendon. The part also functions as a connection for other components of the module. This part is assumed to be unbreakable. 3) Internal cartridge heaters generate thermal input and melt LMPA 2D bolt. 4) Clear cover plates function as a sealant of the module. The plates are necessary to seal liquid metal which has low viscosity and large density. The plates also enable an operator to observe inner structure during the self- healing process. 5) Helical springs generate restoring force after breaking of LMPA 2D bolt structure. It also keeps separated half modules connected, which makes the liquid metal not to leak. This part is assumed to be unbreakable too. Each spring can be replaced by another elastic element such as a rubber band. 6) Neodymium magnets also generate restoring force during the self-healing process. Especially, it functions when the force generated by springs are so little. B. Design Design of self-healing module focuses on that of 2D bolt part made from low melting point alloy (LMPA). Maxi- mum tension value determines the size and material of the structure, and basically, the dimensions are decided based on the authors previous research. We set 1500N as the maximum tension value for enduring large tension exerted during legged locomotion of life-sized robot. Low melting point alloy U-47 is the main material of 2D bolt part. Table. I mention the alloys specifi cation. Its tensile strength is as the same as that of polycarbonate, and its melting point is below that of the clear cover plates made from PMMA and PET. Self-healing process does not affect the covers. Dimensions of the aluminium joint part are shown in Fig. 3. Red hatched area transmits tension. Its cross-sectional area and maximum tension values are calculated as Eq. (1) and Eq. (2), respectively. S mm2 = 47.4 mm2 (1) TMAXN = MAXMPa S mm2 = 3081N(2) where MAXis the tensile strength of the material. The value is appropriate when considering the safety factor of 2.0. TABLE I SPECIFICATION OF“U-47” Melting point46.7 C Density8.8 kg/m3 Specifi c heat capacity0.147 J/(g K) Heat of fusion13.986 J/g Tensile strength65 MPa Elongation at break12.4 % C. Assembly Assembly process has a few variations because of the casting step. Shaping low melting point alloy into 2D bolt structure utilizes casting, and the module itself functions as a mold. Hence, post-processing procedures such as threading is unnecessary. There are following two options illustrated in Fig. 4, and each of them has different advantages and disadvantages. 1) Whole casting process features single-step pouring into the assembled module. Before pouring liquid metal onto it, the upside cover of the module should be eliminated. 2) Half casting process has several pouring steps. Addi- tionally, one more heating process is required to make uniform structure just like self-healing behaviour. Spring attachment Stainless cup Liquid metal Assembled module without top cover Half modules with cover plates i. Whole castingii. Half casting (heating) Overfilled(failure) modified Stainless cup Liquid metal Fig. 4.Casting process has a few variations. (i) Whole casting (ii) Half casting First, whole casting process is advantageous for its sim- plicity. However, it has an experimental issue caused by the pouring process. Overfi lling of liquid metal easily occurred, 2569 which made it impossible to attach the top clear cover plates on the module. Removing the small quantity of liquid metal from the solid structure requires some trials. On the other hand, half casting process features easy adjustment. Pouring liquid metal into upright half modules employ its weight of the metal for casting. Weight of the metal pressurizes the metal itself, which enables near net shape casting. Adjusting after each casting and overfi lling is easier than that of the whole casting process. However, a few additional procedures are necessary for half casting process. One is pouring and the other is heating after spring attachment. In addition, uniform mirror surface on the plastic plates makes it diffi cult to observe the phase transition. IV. VERIFICATION OF THE DEVELOPED MODULE This section mentions the validation experiments of the proposed self-healing module. Experiments focused on fol- lowing functions. 1) visualization of internal self-healing thermal reaction 2) large tension transmission ability A. Visualization of self-healing process Fig. 5 describes an experimental setup for inspection of the self-healing process, that is, melting and solidifi cation. A human operator sent PWM reference values to embedded cartridge heaters via the controller board. Thermal input of each cartridge heater is shown in Eq. (3). PinW = V 2 in V2 R 3.8W(3) where Vin= 40, and R = 420. Fig. 5. Healing observation system includes a microscope to inspect healing reaction. Fig. 6 shows the melting progress of the proposed module. After starting of cartridge heaters, phase transition started from both sides towards the center part directionally. It took less than 6min for the broken part to be molten. Middle crack was healed during melting. A heater with less resistance value accelerates melting reaction with the same voltage input. Fig. 7 illustrates solidifi cation progress of proposed mod- ule without stirring. solidifi cation started and proceeded from various points simultaneously and heterogeneously. The op- erator confi rmed the whole surface solidifi ed within 11min after powering off the heaters. Moreover, the temperature of aluminium joint decreased to room temperature in 30min after powering off the heaters, which was shown using an infrared temperature sensor. Further acceleration of thermal reaction can be performed using retrofi tted cooling such as forced air circulation. Fig. 8 shows the appearances of the samples after the solidifi cation process. The surface of (b) is rougher than the other two specimens. In contrast, (a) and (c) shows uniform structure with little surface perturbations. This result indicates that stirring contributes to improved solidifi cation condition. As for (a), liquid metal droplets confl ict to the module with a certain velocity, which generates vibration and mixing. As for (c), a human operator injected sloshing vibration to the experiment setup manually. In a robot system, making tendon vibration by the actuator module would be effective. Crack 00:00:0000:01:5200:05:52 Crack Fig. 6. Melting process of the proposed self-healing module. The characters “S” and “L” stand for solid and liquid, respectively. The reaction proceeded unidirectionally from right and left sides to the center. The center crack disappeard during the process. 00:00:0000:08:0300:10:44 Fig. 7. Solidifi cation process of the proposed self-healing module