IROS2019国际学术会议论文集0327

Abstract Animals in nature, such as geckos, inchworms, and felines can climb on various surfaces using different mechanisms and serve as references for the study of bio-inspired robots. This paper presents an inchworm-inspired climbing robot that consists of soft body and feet. The soft robot is actuated by shape memory alloy wires and utilizes microspine arrays to attach its feet to rough or soft surfaces. A series of experiments to test the functionality of the feet and torso of the designed robot have verified the theoretical feasibility of the robot. Results have shown that the designed bio-inspired robot can climb on inclined or vertical curved surfaces and flat surfaces. The robot can also adapt to the underwater environment. Thus, this robot has great potential for various applications such as pipeline inspection. I. INTRODUCTION Many climbing animals, such as geckos, inchworms, and felines, exist in nature. Through their physical flexibility, these animals can use different mechanisms to climb a variety of surfaces. Numerous rigid climbing robots have been designed that are composed of biomimetic adsorption mechanisms (e.g., Van der Waals forces 1-4 or interlocking 5-8), rigid detachment (e.g., motors 9 or tendons 10), and inflexible bodies. All these developed robots are easily damaged due to accidental fall, and encounter difficulty in adapting to curved surfaces or underwater environments. Despite much rapid development of soft crawling robots in recent years, which use ionic polymer metal composites 11, shape memory alloys 12, 13, or dielectric elastomer 14 as actuators, few studies have been conducted on soft climbing robot. Verma et al. 15 created a soft tube-climbing robot by using pneumatic actuators, which could only climb inside tubes. Tang et al. 16 designed an amphibious climbing soft robot by using pneumatic actuators; this robot could move on the ground and underwater but not on curved or rough walls. This research is supported in part by the National Natural Science Foundation of China (No. 61773358). Qiqiang Hu is with the CAS Key Laboratory of Mechanical Behavior and Design of Materials, Department of Precision Machinery and Precision Instrumentation, University of Science and Technology of China, Hefei, China, and also with the Department of Biomedical Engineering, City University of Hong Kong, Hong Kong SAR (email: .hk) Erbao Dong*, Hu Jin, and Jie Yang are with the CAS Key Laboratory of Mechanical Behavior and Design of Materials, Department of Precision Machinery and Precision Instrumentation, University of Science and Technology of China, Hefei, Anhui, 230026, PRC (*corresponding author e-mail: ebdong, phone: +86-551-63601482) Gang Cheng is with the Beijing Key Laboratory of Intelligent Space Robotic Systems Technology and Applications, Beijing Institute of Spacecraft System Engineering, CAST, Beijing 100094, PRC Dong Sun is with the Department of Biomedical Engineering, City University of Hong Kong, Hong Kong SAR (email: .hk) Gu et al. 17 developed a soft wall-climbing robot by using dielectric-elastomer actuators and electroadhesive feet; the robot could achieve a climbing speed of 63.43 mm/s at sinusoidal voltage (6 kV and 16 Hz). However, this robot requires high voltage and could only climb on flat walls. Three challenging problems exist in the research on the soft climbing robot. First is properly designing soft actuators and adsorption mechanisms. Second is assembling the actuators and the sticking feet appropriately for effective propulsion. Third is controlling the body movement to ensure a stable gait. This paper presents an inchworm-inspired soft climbing robot based on modules of shape memory alloy (SMA) wire. The robot can climb on rough surfaces by using microspine arrays. With its unique waterproof and compliant characteristics of SMA modules, the robot is also capable of adapting to the underwater environment and climbing on curved walls. The paper first presents the design and fabrication of each part of the robot, including the soft actuators and microspine arrays. Then, experiments are conducted to validate the theoretical feasibility of the climbing robot, followed by a discussion of various factors that affect the movement of the robot. Finally, a soft climbing robot prototype is developed and tested. II. DESIGN AND FABRICATION A. SMA actuators The geckos feet are capable of pronation and flexion, which improves the adaptability to different surfaces. Traditional actuators hardly implemented the compliant attachmentdetachment movements, as shown in Figure 1. In the attachment phase, the compliant foot is preloaded and is capable of adapting to raised or curve surfaces. By increasing contact areas, the foot is able to provide higher adhesive forces. In the detachment phase, the foot is peeled from its tip part and easily recovers to its original morphology. Figure 1. Ideal process of attachment and detachment of bio-inspired compliant foot. Inchworm-inspired soft climbing robot using microspine arrays Qiqiang Hu, Erbao Dong*, Gang Cheng, Hu Jin, Jie Yang, Dong Sun 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 IEEE5800 Figure 2. Design and fabrication of climbing robot. (a) Skeleton of actuators, (b) casting process, (c) unidirectional and bi-directional actuators, (d) retractable torso, (e) compliant foot with symmetrical microspine arrays or dry adhesives, and (f) climbing robot. In our previous works 18-20, we investigated the soft and smart modular structures actuated by SMA wires, and proposed a soft actuator capable of bi-directional reciprocal motion. The flapping of the soft actuator was close to the ideal movements of a compliant foot and capable of adapting to the curved surface. Figure 2 shows the process of design and fabrication of the compliant foot using SMAs (Dynalloy, with diameter of 0.15 mm and transition temperature of 70 C). The torso was designed to achieve the axial reciprocating motion. We assembled the feet and torso into a robot by 3D printer adapting pieces. The bending angle and contact force of the feet were controllable by changing the number and length of the SMAs and the distance from the spring steel 19. The compliant foot could achieve the pronation and flexion in the same manner that geckos do, and could help the robot move on the curved or flat surface. B. Microspine arrays Animals that use spines, such as cockchafer 7, beetles 21, and cockroaches 22, are always able to climb on rough and dirty surfaces. As mentioned in the Introduction, many climbing robots have been designed to mimic these animals. The feet of these robots have two common characteristics: one is using rigid materials (such as acupuncture needles 23, steel hooks 24, and fishing hooks 25) as microspines and the other is compliant suspension to connect the robot body. The former enables the robot to latch onto asperities on rough surfaces effectively, while the latter is generally flexible and passive to enable the spines to adapt to various asperities and avoid the risk of rigid collisions. Different from the fabrication technique of shape deposition manufacturing 26 or selective laser sintered 7, module casting is simple and accessible. Figure 3 shows we used the polydimethylsiloxane (PDMS) to mould the microspine arrays (3 5), which had a length of 17 mm, width of 17 mm, height of 2 mm, and glues (Valigoo, V-80) to strengthen the integrations. The exposed spines were made of disposable acupuncture needles (Huanqiu) with 250 m shaft diameter, 60100 m tip diameter, 1 mm height, and 2.5 mm spacing (Figure 3(a-c), the dense distributions of which could improve the probability of valid attachment. As the verification testing, the base angle between spines and the surface of PDMS was set to 60 . Figure 3(d) shows that the compliant foot was manufactured by assembling the microspine arrays and bi-directional SMAs actuator using adhesive bonding technology (HJ-T326). The contact area between the actuator and microspine array surface was guaranteed to be maximized due to the negative effect of decreasing contact area 27. III. EXPERIMENTS To understand the flexural properties of the actuators, a fluorescence test was conducted. The actuators were coated with fluorescer and placed in the dark. Using an interchangeable lens digital camera, we captured the motion of actuators heated under the voltage 14V, and then analyzed the pictures with image processing algorithms. As shown in Figure 4, the neutral surface of the actuators was recognized by image processing (red line) and approximated a circular arc. The two endpoints of the arc were connected to determine the course angle. The course angle could be calculated in real time and decreased as the actuator bended (video in Attachment). The unidirectional actuator had a minimum course angle of 52 and a maximum central angle of 76 . The central angle was equal to (180-2) and directly defined the arc radius as R = 90L/(90-) , where L is the length of the neutral surface of the actuator. When the actuator was in the initial state, the course angle was 88 and the arc radius was Figure 3. Design and fabrication of microspine arrays. (a) Front view, (b) side view, (c) back view, and (d) adhesive bonding between microspine arrays and bi-directional SMAs actuator. 5801 Figure 4. Flexural properties of unidirectional and bi-directional actuators. Figure 5. The course angle and arc radius R in flexural experiment. The length L was 70 mm and the heating voltage was 14V in our experiments. approximately 14 times L. When the actuator was in the end state, the arc radius became approximately 0.75 times L. As shown in Figure 5, the course angle and arc radius were recorded during the heating and cooling process. The start and end point of the heating process were also marked in the upper figure (red dot). There was a mapping relationship between the course angle and arc radius. We could observe that the course angle decreased quickly followed by approximately constant in the heating process, and increased slowly in the cooling process. While the arc radius R could reach a minimum of 5.3 cm, and decreased greatly in the heating process. It was worth mentioning that the radius R and course angle did not return to the initial state due to the accumulated heat and pre-stretch of the SMAs. Furthermore, the length of the actuator could be changed according to actual needs, and the differences also affected the course angle of the actuator. By comparison, the bidirectional actuator had a minimum course angle of 61 and a maximum deflection angle of 58 . The deflection angle could be calculated by the course angle and showed the adaptation range of the actuator for the curved surfaces. According to Figure 5, the bi-directional actuator could adapt to the minimum radius of curvature of 6.9 cm. To ensure that the robot can travel, the forefoot or hind foot needs to adhere closely to the ground in turn, and the torso has to provide displacement, restore force, and drawing force. The torso in the initial state was attached to an ATI force sensor (Gamma SI-130-10), which measured the drawing force of the torso. As shown in Figure 6, the force was directly related to the input voltage. At a certain heating time, the force increased up to 12 N as the voltage improved. As shown in Figure 6(c), the slope of the force curve is large under high voltage, thereby proving that the drawing force rises faster when the input voltage improves. Then, the contraction and release ability of the torso under no load and load conditions were measured. As shown in Figure 7, the heated torso produces the deformation of contraction, which results in axial displacements under the constraints of the guide rail. The displacements were logged by a laser displacement sensor (Keyence, LK-H1W). At a certain heating time, the displacement also increased as the voltage improved up to 10 V and remained at approximately 10.5 mm thereafter. We also tested the restore force of the torso after it produced various displacements. The restore force was produced by the bending deformation of spring steels and gradually enlarged as the voltage increased. The force was also strong under small deformations. The torso could produce the restore force of 1.9 N at heating voltage 14 V, which was able to propel the robot that weighed 32 g. Figure 6. Drawing force of torso (L=80mm). (a) Model of testing device, (b) maximum drawing force varies with input voltage at a certain heated time, and (c) drawing force varies with time. Figure 7. Contraction ability of torso (L=80mm). (a) Model of testing device, (b) maximum displacement under no-load condition and the restore force vary with the input voltage at a certain heating time, and (c) maximum displacement varies with loads at heating voltage of 14 V. 5802 Figure 8. Adaptive and grasping ability of compliant foot with microspine arrays As shown in Figure 7(c), we measured the contraction ability of the torso under different loads at heated voltage 14 V, which increased by 60 g each time until 480 g. The torso could provide the displacement of approximately 2.9 mm under 1 kg load. The results proved that the torso was capable of lifting loads and providing power to travel. The compliant foot consisted of an SMAs actuator and two symmetrical microspine arrays. The adhesion ability of the foot was the important factor to ensure that the robot could climb inclined surfaces. Owing to the flexural properties of the actuators, the foot could adapt to surfaces of different curvatures, as shown in Figure 8. The adaptive capacity was related to the size of the actuator that affected its deflection angle, and the grasping ability had been greatly improved due to the symmetrical microspine arrays. The foot could grasp the scotch tape wrapped with rough sandpaper at heated voltage 14V, which weighed 222 g. Furthermore, we tested the tangential and normal forces produced by the compliant foot on the polylactic acid plate, as shown in Figure 9. The testing methods were also shown in the figure, including testing of single and symmetrical microspine arrays. For the effect of single configuration, the foot generated normal force while accompanying the generation of tangential force. The tangential force is always less than the normal force due to the 60 base angle of the microspine. For the effect of symmetrical configuration, the maximum normal force is close to twice the single configuration and the tangential force is close to zero. The foot produced two tangential forces in opposite directions and normal forces in the same direction. Therefore, the tangential forces cancelled out, and the superposition of normal forces doubled. According to the experiments, the compliant foot could provide 1 N preload and lift the 222 g scotch tape. IV. RESULTS The forefoot and hind foot of an inchworm are claw-like structures that are loosened or grasped sequentially in motion. The torso of the inchworm completes a contraction and release process to produce displacements, as shown in Figure 10. Figure 9. Tangential and normal forces provided by compliant foot. Figure 10. Physical structure of inchworm As shown in Figure 11, our robot mimics the structure and motion of the inchworm. Furthermore, the robot consists of two compliant feet and a torso. The foot and torso were connected with four 3D printed hinges, which ensured that the end of the torso could rotate freely. The guide rail limited the movement direction of the torso. The robot shown in the picture was on the outer surface of the pipe (left) and inside the pipe (right). The parameters of the robot are presented in Table 1. The robot was controlled by five sets of switching circuits, STM32, and two-way power supplies (11 V and 14 V). We controlled the switching time of each actuator according to the different gaits. As shown in Figure 12, the robot climbed on the inclined curved surface. The corresponding gait analysis is presented in Figure 13. In the gait analysis, the upper part of 5803 TABLE I. PARAMETERS OF ROBOT Parts Size (mm) Weight (g) Compliant foot 70 15 2.5 10 Torso 90 10 15 17 Microspine arrays 17 17 2 0.8 Robot 120 70 20 37 Figure 11. Climbing robot the dotted line indicates the foot bent downward, and the lower part shows lifting up of the corresponding feet. For the torso, the upper part of the dotted line indicates heating contraction. In a cycle, the robot experienced the four states (1, 2, 3, and 4), which were marked in the two figures. State 5 indicated the beginning of the next cycle. State 6 occurred 13 cycles later and went through several ineffective cycles because the foot would not attach to the surface. The cycle is the important factor that influenced the performance of the robot. The short power-on time of the actuator prevented the foot from attaching to the surfaces effectively, while the long power-on time caused overheating problems of the SMAs. Thus, control methods should be improved in future work. In our experiment, cycle T was 4 s and the robot could climb on the inclined surface by 0.2 cm/s. We found that the moving speed of the robot was mainly due to the slow recovery of the spring steel, and the insufficient lateral stiffness of the robot caused it to swing left and right, thereby resulting in detachment and fall. In the vertical experiments, the insufficient lateral stiffness directly caused the robot to lose balance and fall. We solved the problem in two ways. On the one hand, we extended the guide rail and limited its movement except the axial movement at the top of the rail, which ensured that the robot could climb