See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/260712456 A New Actuator With Adjustable Stiffness Based on a Variable Ratio Lever Mechanism Article in IEEE/ASME Transactions on Mechatronics · February 2014 DOI: 10.1109/TMECH.2012.2218615 CITATIONS READS 107 1,014 4 authors, including: Amir Jafari Nikos G Tsagarakis University of Texas at San Antonio Istituto Italiano di Tecnologia 33 PUBLICATIONS 1,376 CITATIONS 484 PUBLICATIONS 10,422 CITATIONS SEE PROFILE Irene Sardellitti Istituto Italiano di Tecnologia 23 PUBLICATIONS 816 CITATIONS SEE PROFILE Some of the authors of this publication are also working on these related projects: WALK-MAN View project WALK-MAN: Whole-body Adaptive Locomotion and Manipulation View project All content following this page was uploaded by Nikos G Tsagarakis on 02 January 2016. The user has requested enhancement of the downloaded file. SEE PROFILE IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 19, NO. 1, FEBRUARY 2014 55 A New Actuator With Adjustable Stiffness Based on a Variable Ratio Lever Mechanism Amir Jafari, Nikos G. Tsagarakis, Irene Sardellitti, and Darwin G. Caldwell Abstract—This paper presents the actuator with adjustable stiffness (AwAS-II), an enhanced version of the original realization AwAS. This new variable stiffness actuator significantly differs from its predecessor on the mechanism used for the stiffness regulation. While AwAS tunes the stiffness by regulating the position of the compliant elements along the lever arm, AwAS-II changes the position of the lever’s pivot point. As a result of the new principle, AwAS-II can change the stiffness in a much broader range (from zero to infinity) even by using softer springs and shorter lever arm, compared to AwAS. This makes the setup of AwAS-II more compact and lighter and improves the stiffness regulation response. To evaluate the aptitude of the fast stiffness adjustment, experiments on reproducing the stiffness profile of the human ankle during the stance phase of a normal walking gait are conducted. Results indicate that AwAS-II is capable of reproducing an interpolated stiffness profile of the ankle while providing a net positive work and thus a sufficient amount of energy as required for the toe-off. Index Terms—Adaptable pivot point, adjustable stiffness, variable ratio lever. I. INTRODUCTION N HUMAN–ROBOT interaction area, compliance has an important role [1]–[4]. Physical passive compliance decouples the inertia of the actuator, which normally has a large value, from inertia of the link, and it decreases the stiffness of the joint [5]–[12]. Adaptable compliance indeed helps to improve the performance in different circumstances. For instance, energy efficiency can be improved by properly tuning the stiffness [13], [14]. In the literature, different design approaches for actuation systems that allow the stiffness regulation are proposed. These variable compliance actuation systems typically employ two actuator units in combination with passive elastic elements to control, independently, the compliance and the equilibrium position of the actuated joint. These systems can be implemented by both antagonistic and series configurations. Antagonistic setups follow the same concept of mammalian anatomy, i.e., a joint actuated by two muscles [15]–[19]. On the other hand, the serial configuration is characterized by the I Manuscript received January 30, 2012; revised April 11, 2012; accepted July 25, 2012. Date of publication October 9, 2012; date of current version January 17, 2014. Recommended by Technical Editor H. Fujimoto. This work was supported by the VIACTORS European Commission under Project FP7-ICT2007-3. A. Jafari is with the Department of Mechanical and Process Engineering, BioInspired Robotic Laboratory, Swiss Federal Institute of Technology, CH-8092 Zürich, Switzerland (e-mail: ajafari@ethz.ch). N. G. Tsagarakis, I. Sardellitti, and D. G. Caldwell are with the Department of Advanced Robotics, Istituto Italiano di Tecnologia, 16163 Genoa, Italy (e-mail: nikos.tsagarakis@iit.it; i.sardellitti@gmail.com; Darwin. Caldwell@iit.it). Digital Object Identifier 10.1109/TMECH.2012.2218615 mechanical series, motor–gear–compliant element and it can be implemented both in linear and rotational designs [20]–[24]. The energy requirements for the compliance regulation depend on the implementation principle. In these systems, both antagonistic and serial configurations, the stiffness adjustment is obtained by tuning the pretension of the spring. As a consequence, the change of the stiffness requires a considerable amount of energy since the actuator, which regulates the stiffness, needs to overcome the forces due to the spring deflection. Recently a new design approach based on the variable lever arm mechanism has been introduced. In these implementations, the pretension of the elastic element is kept constant while the stiffness adjustment depends on the effective ratio of the lever. In both the actuators with adjustable stiffness (AwAS) [25] and the hybrid dual actuator unit [26], the ratio is altered by changing the position of the springs with respect to the pivot point. Alternatively in the vsaUT, proposed by [27], the ratio is set by moving the load point on the lever. Theoretically, by varying the position of the springs, the stiffness can be adjusted from zero to a maximum value, which depends on the spring’s rate and length of the lever. Alternatively, by shifting the load point the stiffness range is defined from a minimum value, which depends on the spring’s rate, up to the infinity. This paper proposes the second version of the AwAS-II. The novelty with respect to other variable lever mechanisms consists in the principle implemented for adjusting the stiffness. In the AwAS-II, the position of the pivot point changes while both the position of the springs and the load point are kept constant. The stiffness range does not depend on the spring’s rate and length of the lever, and it can be adjusted from zero to completely rigid. This allows for the introduction of shorter lever compared to AwAS which permits AwAS-II to be a lighter and more compact setup. The reduced lever length also allows for a faster regulation of the stiffness level since the total distance that the pivot needs to travel is reduced. Even though using softer springs can also help to reduce the size and weight of the system, energy storing capacity in the AwAS-II will decrease if softer springs are used. Currently, the regulation of the compliance in assistive systems, such as prosthetic devices and exoskeleton is gaining interest due to the energy efficiency [28]–[32]. In addition, since these devices require a significantly fast stiffness adjustment according to the human performance, the speed of stiffness adjustment is vital. For instance, during normal walking, the ankle stiffness rapidly changes over the stance phase in order to provide the net positive work which is required for the toeoff [33]. However, most of the commercially available prosthetic feet have not the capability to change the stiffness [29]. Therefore, they are not widely accepted by the users since an 1083-4435 © 2012 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications standards/publications/rights/index.html for more information. 56 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 19, NO. 1, FEBRUARY 2014 Fig. 1. Schematic of the several adjustable stiffness concepts based on the lever mechanism. (a) Stiffness is altered by changing the effective arm length through variation of the spring location (AwAS [24], hybrid dual actuator unit [25]). (b) Stiffness is adjusted by changing the point where the force is applied on the lever (vsaUT [26]); (c) Stiffness is changed through the variation of the lever ratio by moving the location of the pivot position. inappropriate stiffness profile of the ankle prosthetic may lead to increase the muscle activities that contribute to body forward propulsion. Some other prosthetic systems have a prescribed stiffness variation profile. The effect of different preset stiffness profiles in energy storage and return prosthetic ankle on muscle activity has been studied in [32]. However, since the stiffness profiles are predefined, these kinds of prosthesis are only suitable for particular conditions, for instance certain walking speed. To realize an ankle prosthetic device applicable to different situations, variable stiffness actuators are expected to be a promising approach. However, when a variable stiffness actuator is introduced to drive the ankle prosthetic, it needs to guarantee a reasonably fast stiffness adjustment as well as the energy storage capacity required for the toe-off. To this purpose, experiments are conducted to verify the effectiveness of AwAS-II in reproducing the ankle stiffness profile behavior during the stance phase. This paper is organized as follows. In Section II, the concept of AwAS-II is explained. The mechanical realization of AwAS-II prototype is presented in Section III. Section IV discusses the modeling of AwA-II. The performance of AwAS-II prototype in tuning the stiffness is analyzed in Section V. Experimental results are discussed in Section VI. Finally, the conclusion and future works are presented in Section VII. II. CONCEPT OF THE MECHANISM IMPLEMENTED IN AWAS-II To explain the advantages of the AwAS-II concept with respect to the other variable stiffness actuators which employ a lever mechanism, a comparison is presented in this section. The mechanism for adjusting the stiffness applied to AwAS [24] and to the hybrid dual actuator unit [25] is based on a variable lever arm mechanism. Assuming a lever which is rotating around its pivot [see Fig. 1(a)], with two springs antagonistically attached and able to move with respect to the pivot, the lever arm is defined as the distance between the pivot and the springs. Therefore, the apparent stiffness at the joint can be tuned by changing the length of the lever arm. In detail, the joint is stiffer when the length of the lever arm increases. The range of the stiffness is defined between zero and a maximum value which depends on the total length of the lever and also on the spring’s rate. Alternatively, in the stiffness adjustment mechanism applied to the vsaUT [26], as it is shown in Fig. 1(b), the springs are fixed and the load point varies on the lever. The length of the effective lever arm, in this case is the distance between the pivot and the load point. As a consequence, also in this case, the stiffness of the lever can be tuned by changing the length of the effective lever arm. However, the shorter is the length of the lever arm, the higher is the apparent stiffness. The range of the stiffness is ultimately defined between a minimum value, which depends on the spring’s rate, up to infinity. Important is to note that this mechanism is only applicable to prismatic joint (where the stiffness is defined as the force required for one unit of linear displacement). Whereas, in the revolute joints (where the stiffness is defined as the torque required for the unit of angular displacement) changing the location of the load point, also implies a variation of the torque, thus the apparent stiffness at the output link remains unchanged. Differently from the principles described earlier, the concept of AwAS-II is based on a lever mechanism with variable ratio. In this case, as it is shown in Fig. 1(c), to tune the stiffness the position of the pivot is changed while the location of both the force application and the springs is kept constant. When the pivot point moves, the ratio of the lever changes. This ratio is defined as L1 /L2 , where L1 is the distance between the pivot and the point where the springs are connected and L2 is the distance between the pivot and the load point. When the pivot is aligned with the springs (L1 = 0), the ratio becomes zero thus the lever stiffness is zero. On the other hand, when the pivot is aligned with the load point (L2 = 0) the stiffness increases up to infinity and the link becomes completely rigid. III. MECHANICAL REALIZATION OF AWAS-II The concept of the lever with variable pivot point is implemented in AwAS-II as it is shown in Fig. 2(a). The lever is connected to the springs from one end and from the other end it is connected to the output link through a rotary joint which allows transmit of force between the lever and the output link while the lever can rotate around this rotary joint when the output link is loaded. Therefore, this rotary joint corresponds to the force point in Fig. 1(c). The CAD view of AwAS-II, as it is shown in Fig. 2(b), consists of a motor M1 devoted to set the link position and a motor M2 inserted to tune the stiffness. M1 is a brushless frameless dc motor [Emoteq:HT02300] with a peak torque of 2.35 N·m connected to a harmonic drive gearbox with a ratio of 50:1. The output of the harmonic drive is then connected to the intermediate link. The motor M2 mounted on the intermediate link is a miniature brushed JAFARI et al.: NEW ACTUATOR WITH ADJUSTABLE STIFFNESS BASED ON A VARIABLE RATIO LEVER MECHANISM 57 Fig. 2. (a) Implementation of the lever with adjustable pivot point in AwAS-II, (b) cross section of the CAD design while the output link is unloaded, (b) stiffness regulation system and the prototype of AwAS-II (d). frameless torque motor [Kollmorgen:QT-0707] with a peak torque of 0.05 N·m. It is connected to a miniature harmonic drive gearbox with ratio of 100:1. The output of the miniature harmonic drive is connected to a ballscrew drive. The pivot is a cam follower [McGill:MCFR] which has a diameter of 13 mm mounted on the ballscrew’s nut and can be linearly moved by motor M2 [see Fig. 2(c)]. A linear guide, parallel to the ballscrew, while preventing the nut to rotate, constraints the pivot to move along the lever. The linear guide also facilitates for the lateral loads applied to the ballscrew. In the AwAS-II prototype [see Fig. 2(d)], two torsion springs with rate of 10 N·m/rad are antagonistically placed with a preangular deflection equal to half of their maximum allowable deflection. This guarantees the continuous engagement of the springs between the lever and the output link for different angular deflections of the joint. The angular difference between the intermediate link and the output link is the angular deflection. A mechanical lock constrains this angular deflection to be in the range of [–0.3, 0.3] rad. When the link deviates from its equilibrium position, the end of the lever, which is connected to the springs, slides along the spring’s legs. To have a frictionless sliding motion, two cam followers are placed between the lever and each spring [see Fig. 2(b)]. It should be mentioned here that when the pivot is aligned with the rotary joint between the lever and the output link [see L2 = 0 in Fig. 2(a)], the force which is transmitted between the lever and the output link cannot produce any torque around that rotary joint to rotate the lever and is directly transferred from the output link to the intermediate link (without involving the springs). In this case, the AwAS-II joint is completely rigid. On the other side, when the pivot is aligned with the center of rotation of the AwAS-II joint (L1 = 0 in Fig. 2(a), which is also where springs are attached to the lever), by loading the Fig. 3. Connections between essential elements of AwAS-II. output link, the lever (which always rotates around the pivot), will rotate around the center of rotation of the AwAS-II joint together with the output link and so there would be no rotation between the lever and the output link. Therefore, springs will not be deflected and so there would be no elastic torque and thus the stiffness is zero. Fig. 3 shows the block diagram of the connections between essential elements of AwAS-II with a partial section on the motor M1 components. As it is clear from this graph, the stator of M1 is connected to the base frame and the rotor M1 is connected to the intermediate link (through the harmonic gear). The pivot is sliding along the intermediate link. The lever is connected to the output link through a rotary joint and from the other end it is attached to the springs. Springs are located between the lever and the output link. The output link is also connected to the base frame through another rotary joint (see Fig. 2) which allows the rotation of the output link around the same axis of the M1. 58 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 19, NO. 1, FEBRUARY 2014 TABLE I GENERAL SPECIFICATIONS OF AWAS-II TABLE II PARAMETERS FOR AWAS-II Fig. 5. spring. Fig. 4. Schematic of AwAS-II. The sensing system of AwAS-II includes four position sensors and one torque sensor. In detail, an optical encoder measures the angular position of the motor M1, two absolute magnetic encoders measure the position of the output link and the intermediate link and an incremental encoder measures the rotation of the ball screw. In addition, a torque sensor has been installed to measure the torque applied to the intermediate link and it is based on a custom load cell element equipped with semiconductor strain gauges. A comparison between the specifications of the original AwAS [24] and AwAS-II is shown in Table I. AwAS-II design achieves a larger range of motion and stiffness regulation. In addition, it can rapidly adjust the stiffness level due to the reduced length of both the ballscrew and the lever. AwAS-II is capable of storing more energy even though it has a significantly reduced size. IV. MODELING OF AWAS-II The schematic of AwAS-II model is illustrated in Fig. 4. The dynamics of this system, neglecting the gravity contribution, can be described as are the external, elastic, and resistant torques, respectively. The other parameters in (1) are described in Table II. The joint stiffness K = ∂τE /∂φ, is defined as the derivative of the elastic torque τE with respect to the angular deflection ϕ = q − θ1 where the elastic torque τE = ∂E/∂φ is the derivative of the elastic energy E with respect to the angular deflection ϕ. Therefore, to formulate the joint stiffness, the elastic torque has to be obtained which can be determined from the elastic energy. To derive the elastic energy of the AwAS-II, the torsion springs are initially replaced by equivalent virtual compression springs. In this case, the rate of the compression springs has to be expressed in terms of torsion spring’s rate. As it is shown in the Fig. 5, the force Ft is applied to the torsion spring at the distance rt from the center of rotation of the torsion spring. To simplify the model, it is assumed that the force Ft is always applied in alignment with the axis of the compression spring. The force Ft deflects the torsion spring by δγ x (2) δγ ≈ rt where x is the linear displacement of the torsion spring’s arm as it is shown in the Fig. 5. The resultant torque applied to the torsion spring is then given by Tt = Ft rt . B1 θ̈1 + ϕ1 θ̇1 − τE = τM 1 (1) where θi and τM i are the position (after the gear reduction) and the torque associated with the motor Mi with iε [1], [2], respectively; q is the position of the output link, τext , τE , and τr (3) Using (2) and (3), the rate of the torsion spring can be obtained such as Kt = Ks rt2 I q̈ + N q̇ + τE = τext B2 θ̈2 + ϕ2 θ̇2 + τr = τM 2 Replacing the torsion spring by an equivalent virtual compression (4) where the term Ks represents the rate of virtual compression spring. Since the torsion springs are assembled with a predeflection, the virtual compression springs are accordingly assumed to have a precompression at the no load equilibrium position. Therefore, the elastic energy stored at the equivalent compression springs due to the angular deflection of the output link is JAFARI et al.: NEW ACTUATOR WITH ADJUSTABLE STIFFNESS BASED ON A VARIABLE RATIO LEVER MECHANISM Fig. 6. 59 Schematic of the stiffness adjustment principle. given as E = Ks x2 (5) where x = P Ā sin α (6) is the displacement of the virtual spring, P Ā is the distance of the pivot with respect to the springs, and α is the rotation angle of the lever with respect to the intermediate link, as shown in Fig. 6. This can be formulated such as α = sin−1 [(1 + r) sin φ] (7) where r= P Ā L − P Ā (8) is the ratio of the lever and L is the length of the lever. The location of the pivot is adjusted by rotating the ballscrew by the motor M2, therefore, P Ā = nθ2 (9) where n (0.0025/2π m/rad) is the transmission ratio between the motor M2 and the ballscrew. By considering (2)–(9), the elastic energy can be derived such as 2 nθ2 Kt E = 2 L2 sin2 φ. (10) rt L − nθ2 Therefore, the elastic torque can be formulated such as 2 nθ2 Kt τE = 2 L2 sin(2φ) (11) rt L − nθ2 and thus the stiffness can be described as 2 nθ2 Kt 2 K=2 2L cos(2φ). rt L − nθ2 (12) From (12), it can be seen that the joint stiffness K depends on the lever ratio r, the rate of the springs Kt , the length of the torsion spring’s arm rt , the total length of the lever L, and the angular deflection ϕ of the output link with respect to the intermediate link. However, the range of the stiffness, which is between zero to infinity, is due to the range of the lever ratio r. The other parameters in (12) only affect the nonlinearity of the stiffness curve. Fig. 7. Stiffness regulation for the AwAS-II prototype based on the traveling distance of the pivot inside the lever. To adjust the stiffness, the motor M2 needs to overcome a resistant torque τr in addition to the friction and inertia components related to the motor M2 and the ballscrew. The energy spent by the motor M2 to overcome the resistant torque is the same as the one stored into the springs. Therefore, the resistant torque τr = ∂E/∂θ2 defined as the derivative of the elastic energy E with respect to the position of the second motor θ2 can be expressed as 3 L Kt τr = 2 2 n2 θ2 sin2 φ. (13) rt L − nθ2 Therefore, on the basis of (13) the resistant torque τr is zero when there is no angular deflection (ϕ = 0). As a consequence, in this case, to set the stiffness the motor M2 only needs to overcome the friction and accelerate the inertia of the motor M2 and the ballscrew. This is an important advantage of variable stiffness actuators which regulates the stiffness using the lever mechanism. Other types, both in antagonistic and serial configurations, which tune the stiffness through variation of the spring’s pretension, require more energy since they have to work against the force due to the spring’s deflection. V. PERFORMANCE ANALYSIS To show the variation of the joint stiffness with respect to the movement of the pivot, a parameter λ = PA/L is introduced. This parameter is the normalized traveling distance of the pivot and using (8) can be reformulated such as λ= r . 1+r (14) Fig. 7 shows the stiffness variation of the AwAS-II prototype (12) as a function of the pivot normalized traveling distance λ. As this graph shows, for λ < 0.6, the stiffness increases slightly and the range of the pivot traveling distance which effectively tunes the stiffness is about 40% of the total length of the lever. Fig. 8 shows the required energy to tune the stiffness which has to be provided by the motor M2 as a function of angular deflection ϕ and ratio r calculated on the basis of (9). 60 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 19, NO. 1, FEBRUARY 2014 Fig. 10. Fig. 8. Energy required to change the stiffness. AwAS-II torques/angular deflection as a function of the ratio r. uation of link oscillation. The feed forward term g(q) is added for gravity compensation. The controller of the motor M2 is given by τM 2 = kM 2p (θ2d − θ2 ) + kM 2d (θ̇2d − θ̇2 ) Fig. 9. where the reference position θ2d for the motor M2, which adjusts the pivot position for a given desired stiffness Kd is obtained from (12) such as ⎛ ⎞ Kd 2 K t L cos(2φ) L ⎠. (18) θ2d = ⎝ √ Kd 2 n + Maximum angular deflection as a function of joint stiffness. rt As it is clear from the graph when the angular deflection is small, a little amount of energy is required to change the stiffness. In AwAS-II, the maximum angular deflection ϕm ax is constrained to be 0.3 rad using a mechanical lock. However, as it is shown in Fig. 9, after a certain level of the stiffness and while the output link is loaded, the springs become fully deflected. In this case, the potential energy of the output link reaches the maximum value of 5.8 J, as it was mentioned in Table I. Therefore, if the stiffness is further increased, this results in a reduction of the maximum angular deflection while the maximum energy stored at the output link remains constant. VI. EXPERIMENTAL RESULTS In order to verify the performance of the AwAS-II, both motors M1 and M2 were controlled through position control strategy running at 1 kHz. The reference position θ1d for the motor M1 was formulated such as θ1d = qd + φ (17) (15) where qd is the reference link position and ϕ = τE /K is the passive deflection computed through the measured joint elastic torque τE and the model-based joint stiffness K from (12). The controller of the motor M1 is τM 1 = kM 1p (θ1d − θ1 ) + kM 1d (θ̇1d − θ̇1 ) + kld q̇ + g(q). (16) The controller for the motor M1 positioning in (16) contains an additional active damping term which is used for the atten- K t L 2 cos(2φ) A. Stiffness Tuning To evaluate the ability of AwAS-II in changing the stiffness, the intermediate link position is fixed by setting the M1 reference to a given value while the output link is manually moved from its no load equilibrium position. The angular deflection between the output link and the intermediate link was computed from the encoder measurements and the elastic joint torque was monitored by the torque sensor. The experiment was repeated for different sets of the ratio r. The experimental results (solid lines) are plotted in Fig. 10 together with the theoretical results (dotted lines). As it is clear from the graph, the experimental data match the predicted values. The slope of each curve represents the joint stiffness K for several values of the lever ratio r. The arrow in Fig. 10 is toward increasing the ratio. It is clear that by changing the ratio, the stiffness can be tuned in a wide range. When the ratio is 0.05 the joint stiffness is almost zero. However, when the ratio becomes bigger, the slop of the curve increases. B. Tracking Simple Periodic Trajectories Initially to examine the ability of AwAS-II to control position and stiffness independently, both motors M1 for position and M2 for stiffness were simultaneously controlled to follow sinusoidal position and stiffness trajectories of different frequencies. To follow a sinusoidal position, the stiffness motor has to vary according to (18). Fig. 11 presents the motor positions and stiffness trajectories against the reference ones revealing the capability of the actuator to control both variables independently with good fidelity. JAFARI et al.: NEW ACTUATOR WITH ADJUSTABLE STIFFNESS BASED ON A VARIABLE RATIO LEVER MECHANISM 61 Fig. 13. Stiffness variation during the stance phase based on the typical ankle torque/angle profile. Fig. 11. Tracking a sine wave trajectory for the position of motor M1, trajectory of the motor M2 in order to achieve desired sine wave trajectory for the stiffness. Fig. 12. Typical and interpolated ankle torque/angle profile of a 75 kg person during normal walking [33]. C. Reproducing Position and Stiffness of Human Ankle The stiffness adjustment has an important role in human ankle over the stance phase of a gait cycle [31]. In this experiment, we verified the performance of the AwAS-II as a variable stiffness actuator used for driving an orthosis for the human ankle. During walking, the ankle has to provide a net positive work over the stance phase in order to supply sufficient energy for the toe-off. The stance phase can be divided into two steps; loading and unloading [32]. During the loading operation, the ankle stores energy and over the unloading operation it releases the energy. However, in order to generate a net positive work over the stance phase, the amount of released energy has to be more than the amount of energy stored. In a human ankle this is performed through the stiffness adjustment. Fig. 12 shows the ankle/position profile during a gait cycle for a healthy person of 75 Kg [33]. It is important to note that Fig. 12 is only a typical graph since each person has a different gait pattern. In this graph, the network is positive since the profile is a CCW closed loop chain (abcdefgha). The area inside the loop represents the amount of the work. The stance phase is from a to h and the swing phase of the gait cycle is from h to a. From a (Heel Strike) to b (Foot Flat), is controlled plantarflexion period in which ankle behaves like a linear spring and so does not contribute on the mechanical work [35]. However, the ankle produces the net positive work by changing the stiffness during the stance phase from b to h. The stiffness profile associated with the torque/position characteristic of the stance phase after the foot flat (b to h) is shown in Fig. 12. In a normal speed of walking, this period is about 90% of the stance phase interval [33]. As it is clear, stiffness variation in the human ankle is very fast. The goal of this experiment is to generate the same net positive work as the human ankle does by regulating the stiffness and not tracking the stiffness profile accurately. The tracked stiffness profile should be as close as possible to that human ankle with minimum stiffness variation since changing the stiffness requires energy consumption. Therefore, in Fig. 12 the data are initially divided into several groups and then interpolated through linear least square method. This pricewise interpolation of torque/angle generates almost the same net positive work but it requires less stiffness variation compare to human ankle. It is important to note that the ankle position in Fig. 12 is given with respect to the equilibrium position in which the shank is perpendicular to the foot. In addition, it should be clarified that the position calculated for the ankle assumes a single stationary rotation axis. However, this assumption is often introduced in prosthetic ankle studies [32], [34]. As a consequence, to experimentally reproduce the ankle stiffness and position profile, the output link of AwAS-II is locked in a given position while the intermediate link (motor M1) is free to rotate around its axis (see Fig. 14). Therefore, the reference position sent to the motor M1 corresponds to the ankle position during the stance phase. Fig. 15 shows the result of the stiffness and position tracking over the stance phase. In this graph the interpolated stiffness is tracked, however the goal of this experiment is to generate the same amount of positive work over the stance phase by tuning the stiffness and not to accurately track the stiffness profile in Fig. 13. This graph demonstrates that AwAS-II is capable of reproducing position and interpolated stiffness of the human ankle over the stance phase in order to provide the same amount of net positive work required for the push-off in human ankle. 62 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 19, NO. 1, FEBRUARY 2014 of the ankle stiffness profile. It is important to mention here that the capability of AwAS-II (and in general any other actuator) in generating the net positive work depends on the torque and speed characteristics of its main motor and existence of the spring in compliant actuators helps to reduce the required power by the main motor. The capability of AwAS-II in this optimal scenario is constrained due to its limited energy storage capability. Therefore, AwAS-II may need to exert additional torque through its main motor for generating larger net positive works. Fig. 14. Experimental setup for reproducing the ankle position and stiffness during the stance phase; the output link if fixed horizontally while the intermediate link (motor M1) can rotate around its axis. Fig. 15. Reproducing the human ankle during the stance phase: (a) interpolated stiffness and (b) position. VII. CONCLUSION AND FUTURE WORK In this paper, a new actuator with adjustable stiffness AwAS-II was presented, which is an improved version of the original actuator AwAS [24]. The proposed actuator has a wider range of stiffness (from zero to infinity) which is achieved through a variable ratio lever mechanism. The stiffness is mostly affected by the lever ratio with a minor contribution coming from the load. Compared to the original prototype of AwAS, the new actuator can adjust the stiffness faster due to the reduced pivot travel distance. When the link is not loaded, the force due to the spring’s predeflection is perpendicular to the pivot displacement, thus the requested energy to adjust the stiffness is minimum. In addition, the new version results in a lighter and more compact structure with an increased energy storage capacity. Experiments results verified the performance of the AwAS-II in adjusting the stiffness. Initially the tracking of primitive profiles such as sine waves for both the position and the stiffness of the link were tested, showing good performance. Then, experiments were carried out to verify the ability of AwAS-II in reproducing the stiffness variation of the human ankle over the stance phase. For this purpose, a profile of human ankle stiffness was considered and then sent as input reference to the stiffness motor. Results indicated that the AwAS-II changed the stiffness rapidly and provided a sufficient amount of net positive work required for the push-off. Future work consists in the implementation of AwAS-II for driving the ankle joint of a leg exoskeleton. REFERENCES Fig. 16. phase. Power generated by the human ankle and AwAS-II over the stance To verify the positive work, the power generated at the human ankle is plotted in Fig. 16 together with the power at the AwAS-II joint by using the torque (through the stiffness and position) and the velocity (time derivative of the position) profiles. This graph shows that AwAS-II was able to generate almost the same amount of positive work (5.1 J) as the human ankle (5.8 J) over the stance phase by tracking the linear interpolation [1] A. Albu-Schaffer, S. Haddadin, C. Ott, A. Stemmer, T. Wimbock, and G. Hirzinger, “The dlr lightweight robot: Design and control concepts for robots in human environment,” Int. J. Ind. Robot, vol. 34, no. 5, pp. 376– 385, 2007. [2] A. Bicchi and G. Tonietti, “Fast and soft arm tactics: Dealing with safetyperformance trade off in robot arm design and control,” IEEE Robot. Autom. Mag., vol. 11, no. 2, pp. 22–23, 2004. [3] H. O. Lim and K. Taine, “Collision tolerant control of human friendly robot with viscoelastic trunk,” IEEE/ASME Trans. Mechatronics, vol. 4, no. 4, pp. 417–427, 1999. [4] J. Heinzman and A. Zelinsky, “Quantitative safety guarantees for physical human–robot interaction,” Int. J. Robot. Res., vol. 22, no. 7, pp. 479–504, 2003. [5] G. Pratt and M. Williamson, “Series elastic actuators,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots and Systems.Human Robot Interaction Cooperative Robots, 1995, vol. 1, pp. 399–406. [6] S. Hyon, “A motor control strategy with virtual musculoskeletal systems for compliant anthropomorphic robots,” IEEE/ASME Trans. Mechatronics, vol. 14, no. 6, pp. 677–688, Dec. 2009. [7] D. Sun and J. K. Mills, “Manipulating rigid payloads with multiple robots using compliant grippers,” IEEE/ASME Trans. Mechatronics, vol. 7, no. 1, pp. 23–34, Mar. 2002. JAFARI et al.: NEW ACTUATOR WITH ADJUSTABLE STIFFNESS BASED ON A VARIABLE RATIO LEVER MECHANISM [8] A. M. Dollar and R. D. Howe, “A robust compliant grasper via shape deposition manufacturing,” IEEE/ASME Trans. Mechatronics, vol. 11, no. 2, pp. 154–161, Apr. 2006. [9] J. Y. Lew and S. M. Moon, “A simple active damping control for compliant base manipulators,” IEEE/ASME Trans. Mechatronics, vol. 6, no. 3, pp. 305–310, Sep. 2001. [10] K. Kong, J. Bae, and M. Tomizuka, “Control of rotary series elastic actuator for ideal force-mode actuation in human–robot interaction applications,” IEEE/ASME Trans. Mechatronics, vol. 14, no. 1, pp. 105–118, Feb. 2009. [11] K. Kong, J. Bae, and M. Tomizuka, “A compact rotary series elastic actuator for human assistive systems,” IEEE/ASME Trans. Mechatronics, vol. pp, no. 99, pp. 41–50, 2011. [12] F. Carpi, C. Menon, and D. De Rossi, “Electroactive elastomeric actuator for all-polymer linear peristaltic pumps,” IEEE/ASME Trans. Mechatronics, vol. 15, no. 3, pp. 460–470, Jun. 2010. [13] G. Cabodevilla, N. Chaillet, and G. Abba, “Energy-minimized gait for a biped robot,” in Proc. Fachgespräch Autonome Mobile Systeme, 1995, pp. 90–99. [14] A. Jafari, N. Tsagarakis, and D. Caldwell, “A novel intrinsically energy efficient actuator with adjustable stiffness (AWAS),” IEEE/ASME Trans. Mechatronics, vol. 18, no. 1, pp. 355–365, Feb. 2013. [15] A. Bicchi, G. Tonietti, M. Bavaro, and M. Piccigallo, “Variable stiffness actuators for fast and safe motion control,” in Proc. Int. Symp. Robot. Res., 2003, pp. 100–110. [16] S. A. Migliore, E. A. Brown, and S. P. DeWeerth, “Biologically inspired joint stiffness control,” in Proc. IEEE Int. Conf. Rob. Autom., 2005, pp. 4519–4524. [17] G. Tonietti, R. Schiavi, and A. Bicchi, “Design and control of a variable stiffness actuator for safe and fast physical human/robot interaction,” in Proc. Int. Conf. Robot. Autom., Apr. 2005, pp. 526–531. [18] R. Schiavi, G. Grioli, S. Sen, and A. Bicchi, “Vsa-ii: A novel prototype of variable stiffness actuator for safe and performing robots interacting with humans,” in Proc. IEEE Int. Conf. Robot. Autom., May 2008, pp. 2171– 2176. [19] R. Hyouk, K. Choi, J. Ryew, S. Jae-Do, N. Jaewook, J. J. Choon, and K. Tanie, “Biomimetic soft actuator: Design, modeling, control and application,” IEEE/ASME Trans. Mechatronics, vol. 10, no. 5, pp. 581–588, Oct. 2005. [20] S. Wolf and G. Hirzinger, “A new variable stiffness design: Matching requirements of the next robot generation,” in Proc. IEEE Int. Conf. Robot. Autom., May 2008, pp. 1741–1746. [21] J. W. Hurst, J. Chestnutt, and A. Rizzi, “An actuator with mechanically adjustable series compliance,” Robotics Inst., Carnegie Mellon Univ., Pittsburgh, PA, Tech. Rep. CMU-RI-TR-04-24, 2004. [22] T. Sugar, “A novel selective compliant actuator,” Mechatronics, vol. 12, no. 9, pp. 1157–1171, 2002. [23] B. Vanderborght, N. Tsagarakis, C. Semini, R. Van Ham, and D. G. Caldwell, “Maccepa 2.0: Adjustable compliant actuator with stiffening characteristic for energy efficient hopping,” in Proc. IEEE Int. Conf. Robot. Autom., 2009, pp. 12–19. [24] A. Jafari, N. Tsagarakis, B. Vanderborght, and Darwin Caldwell, “AwAS: A novel actuator with adjustable stiffness,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Systems, 2010, pp. 4201–4206. [25] B.-S. Kim and J.-B. Song, “Hybrid dual actuator unit: A design of a variable stiffness actuator based on an adjustable moment arm mechanism,” in Proc. IEEE Int. Conf. Robot. Autom., 2010, pp. 34–40. [26] L. Visser, R. Carloni, R. Unal, and S. Stramigioli, “Modeling and design of energy efficient variable stiffness actuators,” in Proc. IEEE Int. Conf. Robot. Autom., 2010, pp. 4321–4327. [27] S. Kajikawa and K. Abe, “Robot finger module with multidirectional adjustable joint stiffness,” in Proc. IEEE/ASME Trans. Mechatronics, 2010, vol. 99, no. 1, pp. 1–9. [28] L. Zollo, S. Roccella, E. Guglielmelli, M.C. Carrozza, and P. Dario, “Biomechatronic design and control of an anthromorphic artificial hand for prosthetic and robotic applications,” in Proc. IEEE/ASME Trans. Mechatronics, 2007, vol. 12, no. 4, pp. 418–426. [29] G. K. Kulte, J. S. Berge, and A. D. Segal, “Heel-region properties of prosthetic feet and shoes,” Rehabil. Res. Develop., vol. 41, no. 4, pp. 535–546, 2004. [30] G. K. Klute, C. F. Kallfelz, and J. M. Czerniecki, “Mechanical properties of prosthetic limbs: Adapting to the patient,” Rehabil. Res. Develop., vol. 38, no. 3, pp. 299–307, 2001. View publication stats 63 [31] D. Russell and M. McTavish, “A bench-top prototype of a variable stiffness prosthetic,” in Proc. MyoElectric Controls/Powered Prosthetics Symp., Fredericton, NB, Canada, Aug. 17–19, 2005. [32] J. D. Ventura, G. K. Klute, and R. R. Neptune, “The effect of prosthetic ankle energy storage and return properties on muscle activity in belowknee amputee walking,” Gait Posture, vol. 33, pp. 220–226, 2011. [33] A. Bogert, “Exotendons for assistance of human locomotion,” BioMedical Eng. Online, vol. 2, no. 1, pp. 2–17, 2003. [34] Y. Ehara, M. Beppu, S. Numara, Y. Kunimi, and S. Takahashi, “Energy storing property of so-called energy-storing prosthetic feet,” Arch. Phys. Med. Rehabil., vol. 74, pp. 68–72, 1993. [35] S. K. Au and H. Herr, “Powered ankle-foot prosthesis,” IEEE Robot. Autom. Mag., vol. 15, no. 3, pp. 52–59, Sep. 2008. Amir Jafari received the B.S. degree in industrial engineering and the M.S. degree in mechanical engineering from Isfahan University of Technology, Khomeynīshahr, Iran, in 2004 and 2006, respectively, and the Ph.D. degree in robotics from the Italian Institute of Technology, Genoa, Italy, in April 2011. He is currently a Postdoctoral Researcher at the Bio-Inspired Robotic Laboratory, Swiss Federal Institute of Technology, Zurich, Switzerland. His research interests include design and control of variable stiffness actuators, rehabilitation robotics, and exoskeletons. Nikos G. Tsagarakis received the D.Eng. degree in electrical and computer science engineering from the Polytechnic School of Aristotle University, Greece, in 1995, and the M.Sc degree in control engineering and the Ph.D. degree in robotics from the Univeristy of Salford, U.K., in 1997 and 2000, respectively. He is currently a Senior Researcher at the Italian Institute of Technology (IIT) (Head of Humanoids Lab, ADVR), Genoa, Italy, with overall responsibility for humanoid design (iCub and cCub). Before joining IIT, he was a Research Fellow and then a Senior Research Fellow in the Centre for Robotics and Automation, University of Salford, where he worked on haptic systems, wearable exoskeletons, humanoid robots, mechanism design, and actuation systems. He is the author of more than 90 papers published in research journals and presented at international conferences. Irene Sardellitti received the M.S. degree in biomedical engineering (cum laude) from Campus BioMedico University of Rome, Rome, Italy, in 2004, and the Ph.D. degree in bioengineering from the Scuola Superiore Sant’ Anna of Pisa, Pisa, Italy, in May 2008. In January 2005, she joined the Advanced Robotic Technology and System Laboratories, Scuola Superiore Sant’Anna. In June 2006, she joined, as a visiting student, the Stanford Robotics Laboratory where she pursued her research for two years. Since July 2008, she holds a Postdoctoral position in the Advanced Robotics Department, Italian Institute of Technology, Genoa, Italy. Her research interests include humanfriendly robotics, safety analysis, and control strategies for compliant actuation. Darwin G. Caldwell received the B.Sc. and Ph.D degrees in robotics from the University of Hull, U.K., in 1986 and 1990, respectively. In 1994, he received the M.Sc. degree in management from the University of Salford, U.K. He is currently the Director of Robotics at the Italian Institute of Technology, Genoa, Italy. He is a Visiting/Honorary/Emeritus Professor at the Universities of Sheffield, Manchester, and Wales, Bangor. His research interests include innovative actuators and sensors, haptic feedback, force augmentation exoskeletons, dexterous manipulators, humanoid robotics, biomimetic systems, rehabilitation robotics, telepresence and teleoperation procedures, and robotics and automation systems for the food industry.