Avian‐Inspired Perching Mechanism for Jumping Robots

The integration of multiple locomotion strategies and behaviors allows robots to extend the working environment and enhance the performance of each motion. This work integrates perching to a jumping robot to improve the jumping performance. The developed avian‐inspired perching device has a shock‐absorbing mechanism, which consist of a 3D printable flexible polymer material that absorbs the perching impact. This work characterizes the shock‐absorbing performance of the viscoelastic material as a function of hardness and thickness of the material, initial angles of a mechanism, mechanism length, perching speed, and perching angle. This work also characterizes the performance of mechanical interlocking and penetration as the engagement strategies for vertical surfaces. The performance of perching mechanism as a function of hardness of the target surface, contact angle of the claw, and performance of the shock absorption is observed. Finally, demonstrations to evaluate the perching mechanism's performance on the complete system are conducted, and the robot's performance enhancement with an integrated perching motion is shown. This work provides a design methodology to develop and integrate a perching mechanism into jumping robots.

interlock. [20,25] This works combined the viscoelastic material for the shock-absorbing mechanism (leg part in Figure 1A) and mechanical interlock through claws as the engagement type (foot part in Figure 1A). Although this combination has already been developed, [20] the developed mechanism in this work has advanced features through 3D printing for enhanced perching performance. The shock-absorbing mechanism is a 3D-printable viscoelastic digital material. The digital material can vary the hardness of the shock-absorbing mechanism through changing the ratios of the solid and flexible base materials during printing. The changes in the material ratio and material structure change the viscoelastic properties allowing for optimization of the shock-absorbing characteristics of the material. The engagement mechanism is also a 3D-printable multimaterial. Through the assembly of claws to the flexible part, the mechanism has a passive fastening mechanism inspired by the bird's leg, as seen in Figure 1B-D. Applied loads during leg folding for the shock absorption ( Figure 1C) and weight after perching ( Figure 1D) passively fasten the claw's interlocking. Control of the loads on the wire allows engagement and detachment phases while perching. As a result, the developed bioinspired perching mechanism has simplicity in design, a diversity of designs for optimization, and passive actuation.
This work consists three main sections: first section is about characterization of the shock-absorbing mechanism, which is a 3D-printable viscoelastic material. We first characterized the shock-absorbing performances according to various parameters, such as hardness of the material, foldable angle, thickness, mechanism length, perching speed, and perching angle. Second, we characterized the engagement performance of the gripper within the context of the energy-absorbing mechanism. This work studies mechanical interlocking and penetration as methods for engagement. The characterization of the engagement was conducted at different conditions, such as hardness of surfaces and contact angle of the claw, and includes performance changes of interlocking according to penetration. Experimental test were conducted to determine the performance characteristics of the mechanism both independently and within the complete system's locomotion behavior demonstrating the enhancement to jumping performance. Third, this work provides a design methodology for developing perching mechanisms for a target system.

Shock-Absorbing Mechanism
As perching typically occurs during flying or jumping, the robot has a surface approach speed before perching. The approaching speed generates a perching shock and can cause engagement failures if the mechanism cannot absorb the perching shock properly. One option explored in previous works has been dynamic motion control during perching to minimize the perching shock, [26,27] where orientation and approach speed control while perching can also be observed in birds. [28] However, these strategies do not tend to remove all the impact energy during perching. Therefore, a shock-absorbing mechanism became necessary. Hydraulic compression [17] had been one of the methods studied to absorb energy. Energy from the perching shock compressed the fluid in the mechanism which consumes the energy. Movement of the center of mass while perching had also been explored as a means of absorbing energy during perching. [22] Finally, viscoelastic materials had been employed to absorb energy as they deform [18][19][20] and provided a simple yet robust mechanism for perching energy absorption.
This work studied the viscoelastic material, digitalized 3D printing rubber-like material, as a passive shock-absorbing mechanism. The mechanism was integrated at the front side of the robot like Figure 2B. Because polyjet 3D printing provided modifiability of the digitalized materials ( Figure 2C), we could easily vary material properties, such as hardness and angle, for the characterization of the shock-absorbing performance. The characterization was also performed for other design parameters, such as the number of the shock-absorbing material and the length of the leg, and initial perching conditions, such as initial angle and speed. As a result, this characterization Figure 1. Avian-inspired perching mechanism to integrate perching and jumping locomotion modes. A) Perching on a vertical wall (a. perching mechanism, b. robot body, c. four-bar jumping mechanism, d. viscoelastic shock absorbing material (bird's muscle), e. tendon for the passive fastening mechanism (bird's tendon), f. outer shell to fix the wire for the passive fastening mechanism, g. carbon load to assemble leg and foot parts, h. flexible claw holder for the fastening mechanism, i. claws for the mechanical interlocking). B) Bird's leg anatomy (j. muscle, k. tendon). C) Fastening mechanism through the leg folding. D) Fastening mechanism through the applied weight of the robot while perching.
www.advancedsciencenews.com www.advintellsyst.com provided the design parameter's effects on shock-absorbing performances; in addition, 3D-printable material provided advantages of ease and simplicity for producing the shockabsorbing material. Figure 2 depicts the experimental setup to characterize the shock-absorbing performance of the mechanism. The pendulum-like equipment was able to precisely set the perching speed and pitch angle as seen in Figure 2A. The experimental procedure was as follows. We initially lifted the rotatable rod to a specific height manually to set the potential energy, then let it rotate freely to transform the potential energy into kinetic energy. A quick-release mechanism was attached at the end of the rod ( Figure 2B), which released the robot at a target speed. We could change the target speed by changing the initial potential energy. The trigger point was determined by a wire length connected between the quick-release mechanism and equipment frame. In this case, the robot was released at the horizontal position, yielding only horizontal velocity. Figure 2C shows the foldable leg part, which is a viscoelastic digital material. As the robot contacted the surface, the inertia of the robot's body and reaction force folded the leg and deformed the viscoelastic material, absorbing or consuming the impact energy. Figure 2D depicts the overall impact process. The leg folding not only absorbed the overall kinetic energy (E O ) but transformed some of the energy into pitch (E P ) and vertical (E V ) velocities. This was characterized as the first absorption. At the point where the robot's body hitted the surface the second absorption phase began. The new reaction force compensated for the energies generated by the first absorption causing the robot to pitch toward the surface, which created additional leg folding and additional energy absorption. The body impact also absorbd energy by dispersing it throughout the body structure. The second absorption ended as the foot left the surface, and any remaining energy must eventually be absorbed by the attachment mechanism itself. Although Figure 2D has gravitational effects after the foot's leaving because the experiments for the characterization of the shock-absorbing performances have not considered the foot engagement to the surface, the kinetic energy became zero after the successful perching with the foot's interlock.
A high-speed camera recorded the whole impact process and took the position data to calculate energy changes of the body and foot contact stability. To calculate the pure shock-absorbing performances of the leg, we only considered the first absorption before the body hitting had occurred. Then, the absorbed energy of the body is calculated as follows Figure 2. Experimental setup and method to characterize the shock-absorbing performance of the viscoelastic digital material. A) Pendulum-like setup to generate a perching motion (a. rotational joint, b. rod to connect the rotational joint and quick release mechanism, c. quick-release mechanism, d. wire to trigger the quick release mechanism, e. target surface (flat timber plate), f. light for high-speed recording, g. high-speed camera). B) Detailed view of the quick-release mechanism (h. quick-release mechanism, if the wire triggers the mechanism, upper and lower holders open together to minimize an interaction to robot's behaviors, i. robot body, j. perching mechanism; the foot part is replaced as a circular tip to minimize effects from the foot contact with surfaces). C) Viscoelastic 3D-printed material part of the leg (k. UV curable flexible material (TnagoBlackPlus, Stratasys), l. UV curable solid material (VeroBlackPlus, Stratasys), m. digital material structure). D) Energy changes during perching (n. approaching, o. first absorption, p. body hitting between the first and second absorption, q. second absorption, r. behavior after shock absorption, E O is the overall kinetic energy of the robot's planar motion, E H , E V , and E P are robot's kinetic energy calculated by the horizontal, velocity, and pitch velocity, respectively.). E) Parameters for the perching motion shown on the robot side-view photo.
www.advancedsciencenews.com www.advintellsyst.com where KE is a measured kinetic energy of the body's center of mass, m b is the mass of the body, v b is the linear velocity of the body, I b is the body's moment of inertia in the pitch direction, ω p is the angular velocity of the body in pitch direction, E ab is the absorbed energy, KE i is the kinetic energy at the foot's first contact with surfaces, and KE t is the kinetic energy after the first or second absorptions. Foot contact stability is also an essential factor in successful perching. If the foot can stay in contact with the surface during the entire impact process, the claws will have more opportunity for engagement than if the foot is bouncing off the surface. We observed the foot's contact behavior from the recorded videos and determined the foot's contact stability.
For the characterization, we measured the shock-absorbing performance under different conditions. Figure 2E depicts the parameters used to determine the perching performance. The design parameters were the leg angle, θ l , hardness of the leg material, leg length, l p , and the number of legs. The initial conditions for perching were approaching angle, θ p , and speed, v p . We could easily change the leg angle and hardness during the polyjet 3D printing and vary the approaching angle and speed in the experimental setup.

Engagement Mechanism
Perching is achieved by attachment to a target high surface. Although there are various engagement types, such as grasping, [21,22] dry adhesive, [23] and electrostatic adhesion, [24] this work studied the performance of mechanical interlocking and penetration, as one of the common and challenging types. The surface parameters, such as surface roughness and hardness, are important in determining the performance of the engagement. The surface roughness is one of the essential factors for mechanical interlocking. For example, birds and insects have claws and hairs on their feet to increase the chance of mechanical interlocks with rough surfaces, which have been studied by roboticists. [25,[29][30][31][32][33] The hardness of a target surface is another important factor in interlocking through penetration. Regardless of the surface's roughness, a sharp end-tip can penetrate the soft surface to make an engagement. [34,35] This work covered the characterization of mechanical interlocking performance of the developed mechanism but also included additional studies on the performance changes with penetration. Interlocking through penetration is more robust than without the penetration, as asperities are not necessary. As a result, we can strategically use the penetration according to the surface's hardness and enhance the engagement's success rate. Figure 3 depicts parameters for the engagement. A design parameter is the claw angle calculated as follows where θ c is the claw angle of the claw, θ l is a leg angle, and θ f is a foot angle which is fixed at 40°; leg angles are varied to vary the claw angle. Experiments were then conducted to characterize the performance changes of the engagement according to the claw angle. Additionally, performance changes according to shockabsorbing performance were also experimentally characterized by varying the leg angle and hardness together. Surface roughness is another parameter in determining the engagement performance. While engagement methods, such as adhesion [36] or suction [37] prefer a flat surface, mechanical interlocking requires rough surfaces to have abundant asperities for interlocking. Tree bark was selected as the rough surface. The engagement characterization was conducted not only on real tree bark, but also on the same duplicated pattern with harder (Shore D 80, safety helmet) and softer (Shore A 30, rubber) materials as shown in Figure 3. Varying the hardness of the surface highlights the penetration affects on interlocking and perching performance.

Passive Fastening Mechanism
In addition to the claws themselves, the foot structure was also able to passively enhance its grip on the surface. This was achieved through the incorporation of a cable which curls the foot under an applied load, as shown in Figure 1D,E. Therefore, the weight of the robot and leg folding resulted in further curling of the claw, while subsequent jumping reduced this force, allowing it to more easily release.

Shock-Absorbing Performance
This section characterizes the shock-absorbing performance by varying the leg design parameters and perching conditions; these www.advancedsciencenews.com www.advintellsyst.com include the leg material hardness, leg angle, leg length, number of legs, perching angle, and perching speed. Figure 4 and 5 show the change in the energy according to different conditions during the impact process divided into the first and second absorptions.

First Absorption
The leg folding not only transforms initial horizontal energy into pitch and vertical energies but also absorbs energy through the deformation of the viscoelastic material ( Figure 4A, left of the dotted line). At the end of the first absorption the body makes contact with the surface, this is the starting point of the second absorption.
At the initial point of contact, the robot has a set amount of kinetic energy that must eventually be absorbed for a successful perch. The first absorption time varies with the leg parameters. The ideal shock absorbing behavior is where these parameters are varied such that the leg folding is terminated at the onset of the second absorption (body contact) to avoid over or under folding designs, for example, shock-absorbing behavior of hardness = 70 and θ l = 30°in Figure 4A. Figure 4 and 5 provide changes in shock-absorbing behaviors according to the design parameters and perching conditions. Energy absorption of the leg is determined by the characteristics of the leg: the hardness, Figure 4A (rows), leg angle, Figure 4A (columns), number of legs, Figure 4B, and mechanism length, Figure 4C. Multiple legs and stiffer material Figure 4. Measured shock-absorbing performance according to the design parameters. A-C) Energy changes of the body according to the leg properties (A), the number of legs (B), and mechanism length (C) during the entire impact process, the foot first contacts to the surface at zero second, The results are with the perching speed and angle of 1.5 m s À1 and 0°, respectively, All results are averaged from 10 trials. In A), robots have double legs and mechanism length of 40 mm. In B), the mechanism length is 40 mm. In C), the robots have double legs and the leg material properties are Shore A hardness of 30 and leg angle of 30°. Overall KE is kinetic energy of the robot's planar motion. Horizontal, vertical, and pitch KE are kinetic energies calculated by horizontal, vertical, and pitch velocities of the robot.
www.advancedsciencenews.com www.advintellsyst.com increase the stiffness of the shock-absorbing mechanism and thus shortens the time for energy reduction. However, stiffening the digital material results in changes to the material's microstructure and thus properties while doubling the leg increases stiffness through the addition of more material with the same micro-structure and properties (left to right in Figure 4A,B). Large leg angles provide folding space, which increases the shock-absorbing time and changes energy absorption (top to bottom in Figure 4A). The length of the mechanism also plays a significant role in determining the absorption time. As longer leg lengths increase the absorption time due to the body's increased distance from the surface at the point of contact, this also results in increased time for gravity to accelerate the robot vertically ( Figure 4C). Although the presented results show hard material (Shore A hardness = 95) and large leg angle (θ l = 30°) absorb the most energy, early termination of the leg folding causes loss of adaptation ability for the body's behavior and unstable foot contact with the surface (Section 3.2.1). Because the leg folding behaviors can be changed by the perching conditions and interaction between the claw and surface, we provide the best parameters for our system in Section 3.2.2. The first absorption time and shock-absorbing behaviors are also affected by the initial conditions: perching angle, Figure 5A and perching speed, Figure 5B. The perching angle defines the folding torques, τ f from the reaction force of the surface, F r where F f is the tangential force and F fr is the frictional force. Angled perching, having both vertical and horizontal inertial components, generates more folding torque as the friction between the foot and surface cause by sliding down the surface, add to the total folding torque, and therefore increases leg folding. As shown in Figure 5A, folding angles after the same time (25 ms) increase with angled perching. However, excessive perching angle (Perching angle = 30°in Figure 5A) transforms most of the reaction force from the surface to axial forces, F a , causing less leg folding. Although energy absorption through the axial force is possible due to the deformation of viscoelastic material in the axial direction and energy dispersion in the whole body, this work only considers the leg folding as a shockabsorbing strategy during the first absorption. Angled perching has a longer impact time due to the body's increased distance from the surface, which causes early termination of the leg folding and unstable foot contact stability (Section 3.2.1). If we consider the foot contact stability necessary for successful perching, the optimal scenario therefore being perching angles of close to zero degrees, and we can vary other parameters to get the desired shock-absorbing performance. The perching speed directly affects the energy that must be absorbed; however, the perching speed is determined by the robot's locomotion characteristics. Ideally, the locomotion could be adjusted such that at the perching position, the velocity would be minimized; in this case, at the top of the jump the robot is moving at 1.5 m s À1 horizontally. Since the absorption time determines the amount of kinetic energy added due to Figure 5. Measured shock-absorbing performance according to the initial conditions. A,B) Energy changes according to the perching angle (A), and perching speed (B) during the entire impact process, the foot first contacts to the surface at zero second. The results are with mechanism properties as double legs, Shore A hardness of 30, leg angle of 30°, and mechanism length of 40 mm. All results are averaged from 10 trials. In A), F r is reaction force from the surface, F f is folding force, F a is axial force, and the perching speed is 1.5 m s À1 . In B), the perching angle is 0°. Overall KE is kinetic energy of the robot's planar motion. Horizontal, vertical, and pitch KE are kinetic energies calculated by horizontal, vertical, and pitch velocities of the robot.
www.advancedsciencenews.com www.advintellsyst.com gravitational acceleration, lower than expected speeds can greatly increase absorption time. This therefore results in overall lower performance energy absorption. While higher speeds do not increase absorption time, the mechanism's absorption saturates quickly, absorbing a small percentage of the total energy of the system.

Second Absorption
After the first absorption, there is a second chance to fold the leg for further shock absorption ( Figure 2E q). As the first absorption create a pitching moment away from the surface, the second absorption must compensate for this pitch energy as well as the remaining translational energy. The body's impact with the surface is used to change the pitch direction (item p in Figure 2D); however, alternative options, such as a tail, [38] have been explored by others. During the second absorption, the remaining energy is absorbed by not only the body making contact with the surface, dispersing the energy throughout the body structure and additional appendages, but also through a secondary leg folding, due to the change in the pitch direction. As a result, we can strategically use the second absorption to maximize the shockabsorbing performances. Figure 4 and 5 show dramatic reduction of body's kinetic energy during the second absorption.

Engagement Performance
Leg folding behaviors during shock absorption determine foot contact status to surfaces, which is foot contact stability in this work. Adding claws for interlocking also affects the shockabsorbing behaviors and foot contact stability. For the design of the foot engagement performance, this section describes additional design considerations by observing the foot contact stability and interlocking performances. Figure 6 presents the results of the foot contact stability tests. Overfolding, Figure 6A, and underfolding, Figure 6B, can both be observed to cause loss of foot contact due to bouncing; however, the bouncing characteristics are unique to each scenario. Bouncing in the former is due to bottoming out of the mechanism and is more localized in the arm itself, while the latter transfers significant energy into pitch which must be absorbed during body contact (second absorption). The leg angle is the most significant parameter for foot stability when perching, as shown in Figure 6C. For a robot expected to perch around 1.5 m s À1 , a leg angle of 10°shows normal folding behavior across nearly all tested material hardness. However, as the perching angle changes the foot contact stability rapidly decreases, as can be seen in Figure 6D, where the softer material shows the best performance as it is able to better handle the misalignment.

Foot Contact Stability
In the case of the presented prototype, the best design for foot contact stability is a leg angle of 10°and a shore A hardness of 30, allowing for perching speeds between of 1 to 2 m s À1 and perching angles of 0 to þ10°. However, we also need to consider the www.advancedsciencenews.com www.advintellsyst.com shock-absorbing and interlocking performances to find the best design for perching performance. Figure 7 shows engagement performances through interlocking between the claws and surfaces, where Figure 7A shows an example of a successful perching procedure. If the foot contact is unstable, the first established interlock may be lost ( Figure 7B). However, the robot has another chance to make a second interlock during the second absorption due to the change in the pitch direction. As long as one of the interlocks is maintained, the robot will successfully perch on the surface. As a result, the perching performances are the results of the coupled shock-absorbing performances (Figure 4 and 5), foot contact stability (Figure 6), and the claw's interactions with the surfaces.

Interlocking Performance
Including claws for interlocking with rough surfaces changes the shock-absorbing behaviors and foot contact stability as compared with the circular tip on flat surfaces. As the claws slide over the bumps of the rough surface, dynamic reaction forces are generated, which are not seen in flat surfaces. Due to the dynamic reaction forces, instantaneous leg folding generates more pitch torque because of the material's viscoelastic characteristics, which results in underfolding and accompanying loss of foot contact. For this reason, underfolding on rough surfaces occurs in softer materials as compared with flat surfaces (comparing the perching speed of 1.5 m s À1 in Figure 7C and 6C). Although establishing an interlock during the underfolding behavior is possible due to the second interlock, typically, the www.advancedsciencenews.com www.advintellsyst.com robot has low success rates for perching because the foot's bouncing during the underfolding behaviors requires more interlocking forces. The claw's interlocking force can supplement deficiencies in the shock-absorbing performance for successful perching. Several parameters affect the interlocking performance. First, the leg angle changes the claw angle and interlocking forces. Presented angles in the bracket of Figure 7C are the final claw angles after a successful perch. If the claw angles are close to parallel (θ c = 0°) or perpendicular (θ c = 90°) with the surface, interlocking becomes difficult. Although Figure 7C has a narrow range of the final claw angles, Figure 7D has clear differences in the final claw angles as compared to Figure 7C. As a result, the final claw angle of approximately 45°has high success rates due to the robust interlock and reduced underfolding issues. Second, the interlocking force is a function of the surface characteristics, specifically hardness and robustness. The claws can penetrate a soft surface, so they have a more robust interlock and successful perching ( Figure 7H) than a hard surface ( Figure 7C). The robustness of the surface determines whether the interlock can be maintained or not. Real tree bark can be torn by the interlocking force, and the claws lose their hold as seen in Figure 7F. For this reason, success rates on the real tree bark ( Figure 7G) have the same trend as on the hard surface ( Figure 7C) due to the similar hardness, but have lower success rates due to the low surface robustness.
The interlock provides a robust perching behavior, which is able to withstand changes in initial conditions. Initial perching speed determines the initial energy that must be absorbed by the shock-absorbing mechanism ( Figure 7B). Therefore, each combination of leg properties has different success rates, and the robust combination has high success rates at various perching speeds ( Figure 7C,I,J). The leg angle of 40°and shore A hardness of 50 have 100% success rate under perching speed from 1 to 2 m s À1 . The initial perching angle, which is the initial pitch angle, also has different shock-absorbing behaviors and foot contact stability;, where angled perching has an increase chance of underfolding issues. The claw's interlocking forces are able to hold the grasp during underfolding behaviors allowing the mechanism to have a range of successful perching angles ( Figure 7K). A leg angle of 40°and shore A hardness of 50 show 100% success rate under perching angles from 90 to 95°with perching speeds from 1 to 2 m s À1 .
Employing the selected parameters, close up views of the interlocking between the claws and various surfaces and the behaviors of the fastening mechanism are provided in Figure S1, Supporting Information. Supplementary video S1, Supporting Information, showcases demonstrations on possible surfaces around us, such as a rough surface, tree bark, timber, card board, and towel. Success rates from 10 trials are 80% (rough surface), 90% (tree bark), 90% (timber), 100% (cardboard), and 100% (towel). The rough surface is bonded small rock particles, which have a grain size of 1.27-2.08 mm. Figure 8A,B shows robot configurations and the integrated jumping and perching locomotion modes, respectively. The robot has a four-bar jumping mechanism ( Figure 8A.c), which produces jumping energy through the spring (Figure 8A.d) and mechanism folding. A clutch mechanism can wind wire ( Figure 8A.e), which is connected to the end of the jumping mechanism, to store the jumping energy ( Figure 8A.q). Rotation in the opposite direction disengages the clutch and releases the jumping energy ( Figure 8A.p). Opening the jumping mechanism is necessary for successful perching as the legs would otherwise impede to interaction of the perching leg and surface ( Figure 8B.s-u). The jumping mechanism has a rotational connection to a body, and torsion springs provide a passive opening motion ( Figure 8A.g). The mechanism opening produces proper pitch torques with a jumping direction of 80°, creating a parallel perching angle to the vertical surface at the apex of the jumping trajectory ( Figure 9). Therefore, for experimental testing, we selected the parameters of the shock-absorbing mechanism for this scenario, which results in a 1.78 m s À1 perching speed at the apex. To maintain the initial position of the robot prior to jumping, a experimental rig was developed ( Figure 8A.f ). For stable jumping without the holder, work has been done to study interactions between the foot and ground; [39] however, we need further studies to employ the results of the work. Figure 8C provides success rates of demonstrations on different surfaces. Soft surfaces (duplicated bark pattern with soft material, card board, and towel) have higher success rates than hard surfaces (duplicated bark pattern with hard material and rough surface). This is expected as penetration can create interlocks on soft surfaces, while interlocks must be found by the claws on hard surfaces. The procedure for jumping after a perch, including how the robot stores the jumping energy and releases the interlock with the surface, is described in Figure S2, Supporting Information and supplementary video S4, Supporting Information. In addition, the Figure S2, Supporting Information shows an example of a task after perching on the vertical surface, such as taking pictures for exploring purposes at an elevated position.

Design Methodology
The best design parameters create robust perching performance under various conditions, such as perching speed, angle, and surface hardness. The perching motion of this work occurs after the jumping motion. We selected the jumping direction as 80°to have the proper perching angle at the apex of the jumping trajectory, and the perching speed and angle were 1.78 m s À1 and parallel with the surface, respectively. The jumping behavior does produce small variations in the perching speed and angle as seen in Table S1, Supporting Information. In the section for the interlocking performance (Section 3.2.2), we varied the design parameters of the perching mechanism to observe the success rates. A leg angle of 40°, shore A hardness of 50, double leg, and claw angle of 40°had robust perching performances on various surface conditions. To have robust performance under various perching speeds, given a leg designed for a specific perching velocity, low velocities will result in less absorption; however, the lower energy overall requires lower interlocking forces. Perching velocities above the designed velocity will require higher interlocking forces as the additional energy will not be absorbed by the leg mechanism and therefore will remain in the system. Therefore, ideally the leg should be designed for the maximum perching speed observed as lower speeds will create more successful perching as compare to higher speeds. In this case, the selected parameters show stable perching performances between the perching speeds from 1 to 2 m s À1 , which is a sufficient range for this work's perching scenario. In the case of the perching angles, θ b , perching angles under zero degrees reduce the first absorption, causing less shock absorption ( Figure 8B.w). At perching angles greater than zero degrees, shock absorption was improved but caused underfolding and unstable foot contact stability ( Figure 5A and 6D). Therefore, we need to avoid negative perching angles. The selected Figure 8. Demonstration of the integrated jumping and perching locomotion modes. A) Configurations of a jumping base and robot: a. robot body, b. four-bar jumping mechanism under energy stored mode, c. four-bar jumping mechanism under energy released mode, d. spring, e. wire connected between the end of jumping mechanism and winding mechanism inside of a robot, f. jumping mechanism holder to avoid slip during jumping, g. torsional spring for opening motion of a jumping mechanism, h. perching mechanism (Shore A hardness: 50, leg angle: 40°, double legs, and leg length : 40 mm), i. electronics [controller: Teensy3.2 (PJRC), IMU sensor: MTI-1 (xsense), DC motor driver: DRV8833 (Pololu)], j. battery (3 Â 200 mAh), k. marker for a motion capture system: vintage (VICON, smapling: 800 Hz), l. connector for an active tail, m. DC motor for a clutch mechanism, n. clutch, o. winder, p. jumping energy release mode, q. jumping energy-storing mode. Supplementary Video S2 shows how the clutch mechanism works. B) Integrated motion from jumping to perching on a vertical surface: r. robot stores jumping energy in leg mechanism, s. released jumping energy takes off a robot from the ground, t. leg mechanisms are opened for a perching motion, u. leg mechanisms are fully opened before contact with surfaces, v. parallel perching angle to vertical surface (jumping direction : 80°), w. not parallel perching angle to vertical surface (jumping direction : 90°). C) Success rates of jumping-perching on various surfaces, the success rates come from 10 trials. Supplementary Video S3 showcases jumping-perching performances of the robot at various surfaces.

Performance Improvement
Interactions between integrated motions can improve the overall performance of the robot. Examples include: The first improvement is focused on the jumping trajectory characteristics. Jumping with angle and integrated perching and jumping allow for different strategies in using the initial jumping energy to create horizontal distance. While an angled jump divides initial jumping energy into horizontal and vertical kinetic energies (θ j = 30°-90°, Figure 9), the robot can generate a very different trajectory by first jumping vertically and perching then jumping horizontally (θ j = 0°, Figure 9). The maximum jumping distance for this robot is observed at a jumping angle, θ j , of 60°, as seen in Figure 9. However, the shallower the jumping angle, the higher the possibility of hitting an obstacle and therefore certain obstacles may be impassable by an angled jump, e.g., being in a hole. Assuming that a robot can only jump at a single angle, a vertical jump with perching has the highest apex point to overcome obstacles while perch jumping can still yield significant horizontal distances. Procedure of how the robot jumps on the vertical surface after perching is described in Figure S2, Supporting Information. Additionally, assuming that the robot is able to alter its jumping angle while perched, a perched angled jump can create the maximum horizontal distance traveled to overcome a wide obstacle; this will be explored in future work.
The second improvement is focused on improving glide performance in integrated jumping and gliding robots, such as the MultiMo-Bat, [8,9] which could also integrate a perching mechanism. As gliding performance is dependent on the initial conditions, a perched horizontal jump can provide the initial gliding velocity and height potential energy, which would greatly enhance the overall gliding distance as compared a vertical jump which provides zero initial horizontal velocity.
The final improvement is focused on observation and nonlocomotion tasks. Vertical jumping and perching at the apex of jumping trajectory maintain the high position. The high position provides wide range of view to the robot for exploring tasks. We did a demonstration to take a picture after perching as the exploring purpose ( Figure S2d, Supporting Information). In addition, robot can do other tasks such as battery charging through solar cells.

Conclusion
This work integrates perching locomotion mode into a jumping robot. We developed a perching mechanism inspired by a bird's leg, which has shock-absorbing and foot-engagement parts. A 3D-printable viscoelastic material absorbs energy through its deformation (folding). The perching mechanism is characterized according to the design parameters, such as leg angle, hardness, mechanism length, number of legs (thickness), and initial perching conditions, such as perching speed and angle. Claws establish the engagement through mechanical interlocking and penetration. The engagement performance is characterized according to the leg angle and hardness, which determine shock-absorbing performances, surface hardness, and perching conditions. Finally, we proved the performance of the perching mechanism by demonstrating a jumping-perching motion. With the developed perching mechanism, we proposed how the motion integration improves the robot's performances, such as jumping and gliding distances and tasks after perching.
We found that the developed perching mechanism absorbs a perching shock properly and has robustness on perching speeds of 1-2 m s À1 and perching angles of 0°to 5°to various surface conditions. Through the case study, we have learned how we can improve the perching performance, such as pitch control to overcome limited perching angles. In future work, we will design an active tail, specifically an aerodynamic tail, and improve the flexibility of perching performances for various jumping behaviors. In addition, we will maximize the synergy of motion integration through an additional integration of gliding.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.