IROS2019国际学术会议论文集2540

Morphing Structure for Changing Hydrodynamic Characteristics of a Soft Underwater Walking Robot Michael Ishida1, Dylan Drotman1, Benjamin Shih1, Mark Hermes2, Mitul Luhar2, and Michael T. Tolley1 AbstractExisting platforms for underwater exploration and inspection are often limited to traversing open water and must expend large amounts of energy to maintain a position in fl ow for long periods of time. Many benthic animals overcome these limitations using legged locomotion and have different hydrodynamic profi les dictated by different body morphologies. This work presents an underwater legged robot with soft legs and a soft infl atable morphing body that can change shape to infl uence its hydrodynamic characteristics. Flow over the morphing body separates behind the trailing edge of the infl ated shape, so whether the protrusion is at the front, center, or back of the robot infl uences the amount of drag and lift. When the legged robot (2.87 N underwater weight) needs to remain stationary in fl ow, an asymmetrically infl ated body resists sliding by reducing lift on the body by 40% (from 0.52 N to 0.31 N) at the highest fl ow rate tested while only increasing drag by 5.5% (from 1.75 N to 1.85 N). When the legged robot needs to walk with fl ow, a large infl ated body is pushed along by the fl ow, causing the robot to walk 16% faster than it would with an uninfl ated body. The body shape signifi cantly affects the ability of the robot to walk against fl ow as it is able to walk against 0.09 m/s fl ow with the uninfl ated body, but is pushed backwards with a large infl ated body. We demonstrate that the robot can detect changes in fl ow velocity with a commercial force sensor and respond by morphing into a hydrodynamically preferable shape. I. INTRODUCTION Techniques for locomoting underwater differ signifi cantly between man-made and biological systems. Manned sub- mersibles and remotely operated vehicles (ROVs) often use jets 1 or rotating propellers 2 for propulsion. These strategies work well in open water, but are bulky and noisy compared to biological counterparts and have diffi culty nav- igating through confi ned spaces. Some organisms use shape change to create jet propulsion 3, but many more use an undulating body or oscillating fi ns to swim 4. Regardless of the swimming method, active station-keeping in fl ow (eg. for examining a surface or taking long-duration measurements) is energy intensive. Many marine organisms that live on underwater surfaces like reefs or tidepools employ a form of legged locomotion using rigid 5 or soft appendages (eg. hydraulic tube feet 6) to reduce the energetic cost of station- keeping. This work was supported by the Offi ce of Naval Research awards numbered N00014-17-1-2062 and N00014-18-1-2277. 1These authors are with the Department of Mechanical and Aerospace En- gineering at the University of California, San Diego (UCSD), 9500 Gilman Dr, La Jolla, CA 92093 USA mishida, ddrotman, beshih, tolley 2These authors are with the USC Department of Aerospace and Mechan- ical Engineering at the University of Southern California, 854 Downey Way, Los Angeles, CA 90089 USA markherm, luhar Fig. 1.Soft quadruped robot for walking underwater with morphing body that changes shape to be more hydrodynamically advantageous based on force sensor readings. a) Robot with morphed body in fl ow illuminated by laser for fl ow visualization. b) Robot with symmetrically infl ated body. c) Robot with asymmetrically infl ated body. d) Robot with fl at body. All scale bars are 15 mm. The morphology and locomotion behaviors of animals suggest the importance of hydrodynamic characteristics in underwater walking. Unlike on land, stability in water is affected by fl uid forces like buoyancy, lift, and drag, which infl uence underwater walking 7. For example, amphibious newts alter their gaits when transitioning from land to water to reduce the drag on their legs 8, 9. In addition, crabs searching for food orient themselves toward an odor source when walking in low fl ow, but orient themselves in the hydrodynamically preferable direction when walking in high fl ow, even if that inhibits their olfactory senses 10. Morphing structures can actively change shape during system operation and many marine organisms use them to dynamically change their interactions with fl uid 11. Small fi sh have fi ns that deform as they swim 12 while octopi and squids create large body deformations to both propel and maneuver their bodies 13. Similar morphing structures have been used in robotics such as the octopus-inspired Po- seiDRONE 14 and in robots designed by artifi cial evolution 15. Octopus-inspired soft skins have changed the 2D shape and texture of a surface into a preprogrammed 3D Gaussian structure 16. Morphing elements were used in the fl ippers of a turtle-inspired swimming robot 17 and in the wings of a bat-inspired fl ying robot to increase lift and decrease drag during various sequences of the actuators motion 18. Soft robotics is a growing subfi eld of intelligent systems IEEE Robotics and Automation Letters (RAL) paper presented at the 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) Macau, China, November 4-8, 2019 Copyright 2019 IEEE that leverages inherent material compliance to create highly adaptable robots 19. These systems often use fl uidic elas- tomer actuators (FEAs), which are enclosed chambers made of hyperelastic materials that change shape when actuated with fl uid 20. Flexible but inextensible materials constrain and determine actuator movement when the actuator is sub- jected to internal pressure, while maintaining system com- pliance 21, 22, 23, 24. The material fl exibility of soft actuators gives them infi nite degrees-of-freedom and allows continuous deformation 25 to form smooth surfaces that are advantageous for interactions with fl uids. Robots using these components can adapt to their environments, which allows them to perform actions like conforming to irregular surfaces and objects 26. This behavior is especially benefi cial for a legged system as a soft actuator can bend to various different shapes that would require many joints to replicate in a rigid system 27. Previous work on underwater walking robots has pri- marily focused on rigid jointed walkers. The six-legged AQUAROBOT was developed for measuring seabottom roughness via pressure sensors in the feet of the robot 28. An inexpensive autonomous legged underwater vehicle was created for physically locating mines as a largely independent agent of a swarm 29. Little Crabster, another robot with six rigid legs was developed specifi cally for exploring shallow water in high tidal environments 30. To test an underwater walking gait based on the punting gait of a crab, a one-legged robot was designed to show a self-stabilizing locomotion mode combining swimming and pushing gaits 31, while another bioinspired crab robot was created to analyze the underwater dynamics and fl uid forces on the gaits of a hybrid swimming-walking robot 32. However, the aforementioned underwater walking robots have diffi culty walking over un- even terrain and adapting to enclosed environments, which could be addressed with soft robotics components. Here we present a soft robot capable of sensing local fl ow conditions and morphing a portion of its body to maintain a hydrodynamically advantageous profi le (Fig. 1). The shape of the body is altered using soft fl uidic actuators that switch the body profi le between different states to change hydrodynamic characteristics. A commercial off-the- shelf force sensor is used to determine large changes in fl ow speed, allowing the robot to opportunistically change its body shape. When walking in the same direction as the fl ow, the robot can increase the drag on its body and when walking in still water or against fl ow it can reduce the drag. In addition, under higher velocity fl ow the robot can decrease lift to give it additional traction. In Section II, we describe the design of the soft hydraulic quadruped robot and the morphing body. In Section III, we detail the experimental methods used and in Section IV, we discuss the effect of different robot bodies attached to the same legged base on hydrodynamic forces, robot traction, and robot walking speed in fl ow. Finally, we demonstrate a soft robot that can sense changes in fl ow speed and morph its body shape in response. II. ROBOT DESIGN The underwater walking robot consists of four soft hy- draulic legs and a morphing body attached to a rigid frame. Each leg is made of three chambers with bellows. Pressur- izing a chamber causes a net bending motion and actuating the chambers in a prescribed sequence creates a walking gait 33. Similarly, the robot changes the shape of its morphing body by pressurizing separate pouches with water. This produces a reversible shape change to improve the robots hydrodynamic profi le. Morphing the body changes the shape of the surface interacting with fl ow which affects both the lift and drag and morphing into an asymmetric shape creates an orientation dependence. Because different fl ow conditions necessitate different body shapes, the robot must know when to morph its body so we include a sensor that provides information about the fl ow around the robot. A. Hydrodynamic Model The fl uid forces of drag and lift on the body are functions of the body shape, the density of the fl uid, the cross-sectional area of the body in contact with the fl ow, and the fl uid velocity around it. By infl ating and defl ating the morphing body on top of the robot, the robot alters its shape and its cross-sectional area, changing both drag and lift. Increasing drag will increase the walking speed of the robot when walking in the same direction as fl ow, but will decrease walking speed when walking against fl ow. Decreasing lift will increase the normal force applied by the ground, which increases the friction between the ground and the robot; increasing lift will lower the friction between the ground and robot, causing it to more easily slip or move upward. To predict the critical velocity UCat which the underwater walking robot begins to slip, we consider a simple force balance. At the leg-substrate interface, we assume that the drag force generated by the fl uid Fd= 1 2CDAU 2 C and the static friction Ff= N are in balance at the point of incipient motion. From this we obtain 1 2CDAU 2 C= N = ? (MV)g 1 2CLAU 2 C ? (1) where CDand CL are the drag and lift coeffi cients for the morphing body (respectively), A, V, and M are the planform area, volume, and mass of the robot (respectively), is the static friction coeffi cient for the leg-substrate contact, N is the normal force on the robot from the substrate, is the density of the surrounding fl uid, and g is the acceleration due to gravity. We used the planform area for both the lift and drag calculations because the planform and frontal area were similar and would simplify notation. Rearranging the above expression yields the following relationship for the critical velocity: UC= s 2(MV)g A(CD+CL) (2) To increase the critical velocity, a soft robot capable of morphing could decrease planform area A or alter its shape to decrease CD+ CL. Along with minimizing the drag Fig. 2.Images of robot design. a) Image of the soft quadruped with morphing body. b) Schematic of the hydraulic system in which four three-way valves infl ate and vent the legs and a volumetrically-controlled syringe pump infl ates and defl ates the morphing structure. coeffi cient CDby morphing into a smooth hydrodynamically- effi cient shape, another viable strategy is to alter its shape to reduce lift. B. Robot Components We have previously presented a soft quadrupedal robot ca- pable of navigating unstructured terrain on land 34. In this work, we made a new version of this robot to use hydraulics instead of pneumatics for walking underwater. We fabricated the soft fl uidic elastomer actuators that make up the legs of the robot using a commercially available multimaterial 3D printer (Connex3, Stratasys) out of a photocured elastomer (TangoPlus and VeroClear mixture). Each fl exible actuator consisted of three chambers with bellows that extended when internal pressure was applied, causing a net bending motion 34 (see Supplementary Video). Since operation of the legs required a consistent con- nection to a hydraulic pressure source, a tether of tubing connected the robot to a land-based pump and solenoid valve system. We connected the tether only during the experiments that involved walking or active body morphing. Four three- way solenoid valves switched between pressurizing and venting chambers of each leg and a volumetrically controlled syringe pump infl ated and defl ated the morphing body (Fig. 2). The valves were cyclically opened to infl ate the chambers of the leg to create the walking gait (see Supplementary Video). A rigid, 3D-printed frame held the four legs at a 45 angle relative to the body. An infl atable pouch formed a soft morphing body on top of the 3D printed frame that changed the hydrodynamic characteristics of the robot when pressurized. To fabricate the pouch, we attached two sheets of taffeta fabric together with a heat-activated adhesive using an impulse sealer to form an enclosed volume. We then punctured one sheet to insert a tube and fi tting for the hydraulic line and we used a thermoplastic epoxy to seal the opening around the fi tting. We then adhered the bottom of the pouch to a rigid cover that could be reversibly attached to the top of the frame. We created a morphing body with two pouches capable of taking three shapes to explore different hydrodynamic pro- fi les. When infl ated, the large pouch spanned the entire top surface of the robot (Fig. 1b) and the small pouch spanned half of the top surface of the robot (Fig. 1c). When both pouches were defl ated, they formed a fl at top (Fig. 1d); in no confi guration were both pouches infl ated simultaneously. The smaller pouch was 62 mm in length and 43 mm in height while the larger pouch was 125 mm in length and 55 mm in height. The mass of the rigid robot frame and legs combined was 356 grams and the underwater experiments were performed with 267 grams of ballast for an underwater weight of 293 grams when accounting for buoyancy. The pouches were infl ated and defl ated with a volumetrically- controlled syringe pump system 35. To detect fl ow speed changes, we attached a commercial off-the-shelf fl ow meter (Adafruit) to the body of the robot. Sensing the changes in fl ow allowed the robot to respond to unfavorable fl ow conditions by morphing its body. III. METHODS A. Water Channel and Flow Generator We conducted experiments with the robot remaining sta- tionary for visualizing fl ow, measuring the lift and drag on the robot, and measuring the critical velocity causing sliding in a freshwater channel with a length of 762 cm, a width of 91 cm, and a water depth of 48 cm (Fig. 3). The maximum fl ow velocity of the channel was 60 cm/s with a background turbulence level of less than 1%. The ambient fl uid temperature in the channel was 23C, corresponding to a kinematic viscosity of 0.93102cm2 /s. The fl ow parameters differed by a maximum of 5% across the width of the channel. We conducted the tests of the robot walking and of the active response with the fl ow sensor in another similar freshwater fl ow channel. B. Measurement We measured fl ow velocity using Laser Doppler Velocime- try (LDV) near the bottom of the channel at the height of the robot to capture boundary layer effects. We subjected the Fig. 3.Diagrams of the water channel and load cell. a) Layout of the water channel with instrumentation for imaging and measuring fl ow over the robot. b) Confi guration of six-axis load cell for measuring hydrodynamic forces on the robot in fl ow. three morphed confi gurations to increasing fl ow speeds to determine the critical velocity at which the stationary robot would begin to lose traction and slide. Stationary cameras recorded the robot from the side to track the position of the robot in the direction of desired motion and from the bottom to track the out-of-plane motion. We then used open-source optical tracking software (Tracker, Open Source Physics) to determine the walking speed of the robot with different body confi gurations when in still water and when in fl ow. We also estimated the friction coeffi cient between the robot and the glass channel by tilting a glass pane in the water and measuring the angle at which the robot slid due to gravity. To measure the effect of shape change on the hydrody- namic forces experienced by the body, we attached the static robot to a six-axis load cell in the water channel (Fig. 3b). We measured the lift and drag on the four confi gurations of the morphing body (fl at, infl ated half on the leading edge, infl ated half on the trailing edge, and symmetrically infl ated) with and without the legs to isolate the specifi c contribution of the body shape. Because the robot was not walking, we did not attach the hydraulic tether to the robot for these experiments, which would have interfered with measurement of the forces on the body. For each of these cases, we recorded baseline forces with the robot in still water so we could subtract out the effects of gravity. After increasing the fl ow speed to the desired level, we waited two minutes for the fl ow speed transients to subside and collected and averaged lift and drag data at 100 Hz for sixty seconds at three different fl ow speeds. We also imaged the fl ow around the robot using a Particle Image Velocimetry (PIV) system with a laser fi eld projected into the water from above the channel. Light scattered by particles in the water was tracked by software from