仿生六足機(jī)器人機(jī)構(gòu)的設(shè)計(jì),仿生六足機(jī)器人機(jī)構(gòu)的設(shè)計(jì),仿生,機(jī)器人,機(jī)構(gòu),設(shè)計(jì) Available online at SCIENCE DIRECTB d Journal of Bionic Engineering 3 (2006) 115-125 A Biomimetic Climbing Robot Based on the Gecko Carlo Menon, Metin Sitti Camegie Mellon University, Pittsburgh, Pennsylvania 15213-389, USA Abstract The excellent climbing performance of the gecko is inspiring engineers and researchers for the design of artificial systems aimed at moving on vertical surfaces. Climbing robots could perform many useful tasks such as surveillance, inspection, repair, cleaning, and exploration. This paper presents and discusses the design, fabrication, and evaluation of two climbing robots which mimic the gait of the gecko. The first robot is designed considering macro-scale operations on Eaah and in space. The second robot, whose motion is controlled using shape memory alloy actuators, is designed to be easily scaled down for micro-scale applications. Proposed bionic systems can climb up 65 degree slopes at a speed of 20 mms-. Keywords: gecko, robotics, biomimetics, climbing, space memory alloy Copyright 0 2006, Jilin University. Published by Science Press and Elsevier Limited. All rights reserved. 1 Introduction The locomotion, sensing, navigation, and adapta- tion capabilities in animals have long inspired humans to emulate them in robots. The purpose of this study was to determine the potential of climbing robots for both ter- restrial and extra-terrestrial explorations. Robots similar to their biological counterparts require extensive sys- tems for power, locomotion, and actuation, with com- putation, sensing, and autonomy. From animal research and current technologies, the possibility of developing biomimetic robots was analyzed. Locomotory abilities and biomimetic properties of lizards provide an advan- tage for climbing vertical surfaces. The development of climbing robots is mainly driven by the desire to automate tasks which are risky. Wall-climbing robots are used for cleaning high-rise buildings and inspection in dangerous environments such as storage tanks for petroleum industries and nu- clear power plants. Recently, there has also been in- terest in using robots which operate in a micro-gravity environment to inspect and repair space vehicles aside from helping astronauts in their risky operations. Sur- Corresponding author: Carlo Menon E-mail: menoncarlo stargatenetit face climbing and walking robots have become crucial for inspection and maintenance of space shuttles, satel- lites, nuclearplants21, pipeand buildings, search- and- rescue41 for homeland security, exploration on planets or hazardous regions, labeling oil tank volume scale, carrying high payloads, cleaning, sand blasting, painting, and microhano-scale manufacturing application-. These autonomous robots encounter mostly unstructured environments, and by legged walking and climbing lo- comotion, can overcome these obstacles easily. Climbing animals may inspire man to develop ro- bots able to access and operate in hazardous environ- ments. Many animals, e.g., cockroaches, beetles, ants21, and cricket, can climb and use mainly cap- illary forces to stay attached to surfaces. Beetles can lift a load up to 20 times heavier than their body when they are attached to a surface firmly. The geckos ability to climb surfaces, however, has attracted attention for decades. By means of compliant microhano-scale high aspect ratio beta-keratin structures on their feet, geckos manage to adhere to almost any surface with a controlled contact area41. This paper presents and discusses new gecko- inspired robots. Strategic solutions which are used by geckos for climbing are investigated and analyzed. The 116 Journal of Bionic Engineering (2006) vo1.3 No.3 paper presents the design, fabrication and test phases that the authors followed to make two robot prototypes. The first robot, called the Rigid Gecko Robot (RGR), was conceived considering operations in space. Reli- ability and robustness are the most important require- ments for the RGR. The second robot, called the Com- pliant Gecko Robot (CGR), has been conceived and designed for miniaturization. As the miniaturization of standard electric motors and pin joints, which connect rigid links of conventional robots, is intrinsically diffi- cult, a new gecko inspired flexible backbone structure actuated by Shape Memory Alloy (SMA) micro-wires was developed. 2 Problem definition The unique features of a novel climbing robot are: 0 Climbing on any surface roughness and material in any environment (on buildings, rocks and trees, in desert and space, under sea, etc.); 0 Longer operating time and range: currently, the range of autonomous micro-robots is limited by their small on board power sources. By increasing the effi- ciency with which the robots locomote by using the low power attachment and detachment of dry synthetic ad- hesives, the operating range will be increased. Fur- thermore, the self-cleaning nature of the dry adhesive material combined with the high tensile strength of the fibers allows non-degrading performance with a very long lifetime; 0 High maneuverability, speed and agility due to fast attachment and detachment in any orientation; 0 Possibility of carrying higher payloads (a gecko can carry a payload the same weight as its body while climbing a wall); 0 Accessing small areas due to their miniaturiza- tion; 0 Generating very high controlled attachment forces for realizing mechanical work during the robotic mission especially for maintenance applications; 0 Autonomous and on-line monitoring, inspection and maintenance of surfaces by integrated sensing and manipulation tools. In order to develop a vertical climbing robot with high performance, the kinematics of the most agile climbing animal, the gecko, was studied and analyzed. Kinematic data51 were analysed and modeled to simu- late the two dimensional motion of the gecko. Those simulations suggested a climbing robot having the fol- lowing characteristics: (1) Centre of mass close to the vertical surface, (2) Light structure, (3) Reliable system, (4) Robust locomotion. While reversing engineering ideas from nature, the level of biomimetic abstraction must be defined in order to design a system that is valuable from an engineering prospective. For this reason, the authors, who aimed at developing real prototypes able to climb inclined sur- faces, decided to consider only the functional charac- teristics of the gecko adhesive. Commercially available and economically convenient adhesives with good at- taching properties and repeatable behaviour were tested. These adhesives, which have the same attaching func- tionality of gecko hair and are suitable for extended testing, enabled the development of novel climbing ro- bots. We wanted to develop a robot with only the nec- essary degrees of freedom and using only essential components necessary for climbing. Therefore, the robot is not made heavier by auxiliary motors and sensors which may compromise its climbing performance. Here, the proposed designs assume relatively flat surfaces for climbing with no obstacles. 3 Robot design The gecko differs from other climbing animals in three main aspects: dry adhesive pads, foot geometry and gait. These three aspects were studied in order to design a gecko-inspired robot. In this section, the strat- egy for developing a climbing robot prototype is pre- sented and discussed. 3.1 Adhesive pads Much work has been devoted to the development of adhesion mechanisms for climbing robots. Suction ad- hesionV9101 requires the robot to carry an onboard pump to create a vacuum inside cups which are pressed against the wall or ceiling. Many grops developed Car10 Menon, Metin Sitti: A Biomimetic Climbing Robot Based on the Gecko 117 wall-climbing robots using mainly vacuum suction. However, a suction mechanism consumes high power and is relatively slow at detachment. In addition, any gap in the seal can cause the robot to fall. Last, the suction adhesion mechanism relies on ambient pressure to stick to a wall, and therefore is not useful in space applications because of the zero pressure space environment. Another common type of adhesion mechanism is magnetic Magnetic adhesion has been implemented in wall climbing robots for specific applications such as in- spection in nuclear facilities. Despite that, magnetic attachment is useful only in specific environments where the surface is ferromagnetic, so for most applications it is an unsuitable solution. Another strategy is to study passive attachment mechanisms, like those used by climbing animals. The Tokay gecko, for example, can weigh up to 300 g and reach length of 35 cm, yet is still able to run inverted and cling to smooth walls. Unique adhesive pads give the gecko incredible movement and climbing performance. Recently, nano-technology has enabled novel fabrica- tion techniques for gecko-inspired dry adhesives. Geckos have compliant micro- and nano-scale beta-keratin structures of high aspect ratio on their feet which adhere to any surface with a pressure-controlled contact area51. This adhesion is mainly due to molecu- lar forces such as van der Waals forces. Foot-hairs have a branch structure starting from the micrometer scale (stalks) and arriving in nano-scale (spatula stalks). The hairs can bend and conform to a wide variety of surface roughness. Since dry adhesion is based on van der Waals forces, surface chemistry is not of great im- portance. This means that dry adhesion will work on almost any surface. Synthetic adhesive mimicking gecko structure has been developed and exciting results are expected. Using micro-molding techniques, 4 pm di- ameter polymer micro-fibers are already available201, and high performance is possible. However, the devel- opment of a climbing robot prototype needs reliable and commercially available adhesives that could be used for a large number of tests. For this reason two commercial adhesives were tested: Silly Putty and polydimethyl siloxane (PDMS). We chose these two materials as they work on the same functional principle of the gecko ad- 13,14,17 hesive: by preloading the material against a surface, the contact area is maximized and intermolecular bonds are established. Fig. 1 shows results obtained using a customized measurement test-bed. Adhesives had a size of 95 mm2, they were loaded against a glass surface using a preload of 75 mN, an approach velocity of 0.08 mms-, and a retracting velocity of 0.4 111l11.s-l. The contact time was one second. Fig. 1 also shows that, during the one second contact phase, the preload slightly decreases caused by the plastic behaviour of the adhesive materials. During this phase, adhesives comply to surface roughness and fill nano-scale hollows. In addition, Fig.l shows that Silly Putty exerts the highest normal adhesive force and therefore this material was chosen for our robotic ap- plication. Plastic behavior 7- XiEiiq 0 Time (s) 20 I -320 Fig. 1 Silly Putty and flat PDMS adhesive force using 75 mN preload. 3.2 Foot design The adhesive pad of the climbing gecko is opti- mized for power efficiency and fast attachment and detachment cycles. In the attachment phase, the foot approaches the surface and the pad is preloaded and dragged on the surface. Thus, the pad fibers adapt to the surface roughness and maximize the contact area for high adhesion. In the detachment phase, the foot is twisted to peel the adhesive pad from its tip part. Then, the pad pops off and separates from the surface after a critical angle (about 30 degrees). Fig. 2 shows an ideal robotic foot movement. Using a compliant foot, the robot can take advantage of the properties of the adhesive pads. Fig. 3 shows the realistic solution of the attaching-detaching mechanism which was designed for our robot prototypes. Some simplifications were carried with respect to the ideal case of Fig. 2: 118 Journal of Bionic Engineering (2006) Vo1.3 No.3 Fig. 3 Foot mechanism. The adhesive is Silly Putty since tests shows that it Drag motion is not used since Silly Putty does not The approaching, preloading and peeling phases are carried out using the configuration suggested by Fig. 3. The foot mechanism is composed of an electrical solenoid motor, a rigid leg and an elastic foot material. has the highest normal force. have microhano hairs which need to be oriented. 3.3 RGR design The two-dimensional kinematic model of the RGR prototype has ten degrees of freedom (DOF), as shown in the left of Fig. 4. The first four-DOFs are used to lift the legs by means of four motors; one-DOF, in the middle of the robots back, is necessary for locomotion and it is controlled by another motor. The other five-DOFs are passive revolute joints. The right of Fig. 4 shows that the planar kinematics of the robot can be represented by a four-bar-linkage. The dynamics of the RGR, in vertical climbing mode, were studied using multi-body software (Visual Nastran Desktop 4D) and a three-dimensional model with realistic specifications. The model was 0.1 m long, 0.1 m wide and weighed 80 g. The graph on the left of Fig. 5 shows the rotation of the motor which controls the robots back displacement (number 5 in Fig. 4). This rotation is the input for the dynamic simulation. The graph on the right of Fig. 5 shows the torque output of the same motor. This torque is necessary for counterbalancing both the robot weight and dynamic forces caused by the robot motion. Fig. 4 Picture of the rigid gecko inspired robot. z30pL on the right, a schematic representa- tion of the gecko robot showing the model to be stud- ied for understanding its unstable configuration. (FL,J=Fore Left Joint; HRJ=Hind Right Joint; FRJ=Fore Right Joint; HLJ=Hind Right Joint; BRJ=Back Right Joint; MRJSMiddle Revolute Joint.) (3) Changing the position of the motor, (4) Decreasing the angle range of the BRJ rotation. For the RGR prototype, the fourth solution was chosen since a symmetrical configuration of the robot was preferred. 3.4 CGR design The CGR was designed aiming at miniaturization of climbing robots. For this purpose, an innovative compliant system has been developed. This robot has a composite frame and SMA wires which provide motion that mimics muscles. The back, Fig. 8, is flexible, and SMA wires are attached to both sides. The back is able to recover the initial length of the SMA wires during their cooling phase. Unlike revolute electronic motors and rigid links connected by pin joints used in the RGR, the flexible structure and the simple linear SMA actuators can be easily and efficiently scaled down for miniatur- ized climbing robots. The geometry of the robot was optimized both to have long robot step and amplify SMA wires force. With regard to step optimization, analytical kinematics equations were derived taking into account flexible back characteristics. Analysis was necessary to obtain AL, the step length, as a function of all the other parameters, a, b, c, and m of Fig. 8. SMA wire Fig. 8 Compliant gecko Compliant 1. Hind legs inspired robot model. Fig. 9 shows results when wires of both sides of the back are alternately contracted to perform one full step. In order to compare the effects of a and m and obtain the corresponding physical solution, the condition: a+m=constant (2) was used. In addition, the maximum contraction of the wires was limited to 4% of their length because of the inherent SMA wire characteristics. For simplicity, fore 120 Journal of Bionic Engineering (2006) voi.3 No.3 Fig. 9 Relation among L, c, (I and Ap while the variables a and m were constrained by equation a+m=constant. The SMA wire contraction was constrained to the 4% of the wire length. and hind legs were considered of the same lengths (m=b). The following considerations are deducible from the graph on the left of Fig. 9: (1) L increases with u; (2) The variation of L (a) increases with a; (3) The variation of L increases with the variation The graph on the right of Fig. 9 shows that if the length (parameter a) increases, then the step size Ap decreases (see Fig. 8). In addition, the condition a+m= constant means that the step increases with the length of the legs. The ideal robot must therefore have long legs and a short back. The second analysis focused on CGR back de- flection during the contraction of the SMA wires. Since the CGR back is fixed differently to the fore and hind legs (Fig. 4), the compliant back was modeled as a can- tilever with an external normal force, R, and a moment, M, applied to its end (Fig.10). Both R and A4 are func- tions of the cantilever deflection and their values were therefore computed in an iterative procedure during CGR back deflection. The effects of the distance spacer, s, on the distance, d, and force, F, (see Fig.10) were studied using large deflection theory221. The ffow-chart of Fig.11 shows the iterative pro- cedure which was used. Parameters ro and Fa the ap- proximated cantilever curvature and the estimated SMA constant forces respectively, represent the initial soft- ware inputs. For simplicity, the flow-chart of Fig.11 does not show all software subsystems, e.g. subsystems for computing elliptic integrals, which are involved in the cantilever large deflection computation. of c. The graph on the right of Fig. 11 shows results obtained using realistic data of the CGR prototype back, Youngs Modulus = 226 GPa; back length=lO cm, back width = 24 111111. This graph is critical for control strate- gies. In fact, the developed cantilever deflection model can be used in a feed-forward control loop. For the CGR locomotion design, weight and dynamic forces were neglected as the prototype was designed to be very light and to climb slowly. Fig. 10 Model for the SMA force analysis. The CGR can be reduced to the study of a cantilever contracted by a SMA wire. The distance spacer(s) introduces a variable moment M. 51 41 I I/ Displacement (mm) Fig. 11 On the left side: flow-chart of the software developed for the iterative computation of CGR back deflection. Large deflection theory was used. On the right side: force that the SMA wires exerted for bending the CGR back. Different curves correspond to different values of the distance spacers. 4 The prototypes and experiments In this section, actual RGR and CGR prototypes are presented. Robot specifications and characteristics are also discussed. 4.1 RGR prototype The chassis of the RGR, which was designed to operate in macro-scale and for space applications, was Car10 Menon, Metin Sitti: A Biomimetic Climbing Robot Based on the Gecko 121 built using aluminum alloy. The frame was made by folding aluminum sheets. RGR was equipped with five solenoid motors, four for lifting the legs, and one for locomotion. The maximum torque of each motor, which was amplified by a 8 1 : 1 gearboxes, was 25 Nmm ob- tained using 5 V. The RGR was controlled by a PIC 16F877 micro-controller integrated in a customized electronic board. Fig. 12 shows the control strategy used for one-full step. All five motors were controlled in sequence in order to detach one foot at a time minimiz- ing the risk of robot falling. Fig. 13 Picture of the compliant gecko inspired robot. The use of glass fiber had two different purposes: reinforce the compliant body structure; electrically iso- late the CGR frame when in contact with SMA wires. A thin layer of epoxy, obtained by the use of a spinner machine, was also spread over the composite back in order to increase the electrical insulation. Composite theory was used to compute the mechanical properties of the CGR back laminate (Table 1). Table 1 Mechanical properties of the CGR back laminate Fig. 12 Control strategy for one-full robot step: time evo- lution of the rotations of each motorized joint. 4.2 CGR prototype The fabrication of the CGR, shown in Fig. 13, was very challenging due to the use of SMA wires and com- posite material chassis. The CGR back was equipped with 50 pm diameter SMA wires with a transition tem- perature of about 90 C (Flexinol