May 24, 2010

13:37

WSPC - Proceedings Trim Size: 9in x 6in

clawar2010

1

CHARACTERIZATION OF EMERGING ACTUATORS FOR EMPOWERING LEGGED ROBOTS J. PESTANA∗ , R.BOMB´IN, J.C. AREVALO and E. GARCIA Center for Automation and Robotics, CSIC-UPM, 28500 Arganda del Rey, Madrid, Spain ∗ E-mail: [email protected] http://www.iai.csic.es/dca Research in legged locomotion is directed to new service applications such as lower-limb exoskeletons and agile-locomotion robots. The main difficulty that is limiting their further development is the lack of appropriate actuation systems. The excessive material overhead of high power conventional technologies and the need of means to obtain gait energy recovery are some among the features which need to be improved. The aptitude of an actuation system to perform a certain task can be judged from a little set of parameters describing its rated and maximum capabilities. This paper presents experimental data of a variety of promising actuation technologies, and a final discussion on their suitability for empowering legged robots. Keywords: New actuators for service robots, actuator performance analysis

1. Introduction Service robotics is believed to have good market opportunities, since its applications, such as lower-limb exoskeletons1,2 and agile-locomotion robots,3 are of great interest for the general public. The international community is realizing a great financial effort to ensure the development of this research field. However, technical limitations inherent to conventional actuation technology are stalling the new recent achievements. The scientific community has already identified a set of parameters, characterizing an actuator that are to be improved and are “enabling” key factors for successfully empowering legged robots.4,5 In the past, some theoretical studies have been focused on obtaining the power delivery capabilities of different conventional6,7 and emerging8 actuation systems. This paper is an experimentally oriented prosecution of a theoretical previous work on the comparison of novel actuators.8

May 24, 2010

13:37

WSPC - Proceedings Trim Size: 9in x 6in

clawar2010

2

Measurement MSM Module Actuator

Fig. 1. (Left) Photograph of the General Actuator Test Bench, testing the Pneumatic artificial muscle (PAM). (Right) MSM Actuator Test Bench

2. Tested parameters and actuators The tested actuators are a selection of different actuation systems, some of them novel in operation technology, and some of them based on conventional technologies with some added new features. The selected actuators types and models, which operation principle and theoretical characteristics can be found in,8–10 are the following: (1) Series Elastic Actuators9 (SEAs) (a) Hydro-Elastic Actuator, Yobotics HEA-01 (b) Electric Series Elastic Actuator, Yobotics SEA-23-23 (2) Pneumatic Artificial Muscle8 (PAM) FESTO MAS-20-200N-AA (3) Magnetic shape memory actuator,10 Adaptamat MSMA Test Actuator 5/7N The set of studied parameters are the following: • Weight and specific power: required to deliver a sufficient impulse during the leg stance phase, • Specific force: required to provide a competitive robot overall payloadto-weight ratio of about 2, • Actuation response time under position or force control: it measures the time required by the actuated system to follow changes in the position and force references, • Maximum stroke-to-actuator length ratio: some available technologies have a prohibitive low stroke for large workspace applications.

May 24, 2010

13:37

WSPC - Proceedings Trim Size: 9in x 6in

clawar2010

3

3. Experimental setup The need to measure mechanical displacements and efforts with different orders of magnitude is the reason to use two separate test benches, which are shown in Fig. 1. The data acquisition hardware used to implement the control of the actuators is a PXI-based industrial automation and acquisition system; model NI-PXI-1042-Q commercialized by National Instruments, programmed using NI LabVIEW Real-Time. The general actuator test bench, shown in Fig. 1 (left), has been used for the experimental analysis of the Yobotics SEA-23-23 and the FESTO MAS-20-200N-AA PAM actuators. It allows the transmission and measurement of movement and exchanged mechanical force and power between two actuators. Since Yobotics HEA-01, shown in Fig. 4(a), can impose larger forces in a force-controllable fashion, HEA is used as an actuator of antagonistic loads. Two elastic couplings mounted in the central axis add impact and vibration isolation to the test bench. The MSMA actuator has been tested using the low stroke actuator test bench, shown in Fig. 1(right). A detailed description of this test bech can be found in.10 4. Experiments The experiments described in this section are aimed at getting experimental insight into the real capabilities of the above mentioned actuators. 1. Series Elastic Yobotics HEA-01 Since its maximum force is the highest by far, HEA01 had to be tested using the arrangement shown in Fig. 2(a). The actuator lifts a weight, varying from 30 to 165 kg, at different speeds. Note that this actuator is asymetric due to two different piston effective areas. The tested “side” has theoretical maximum capabilities of 1.5 m/s and 2000 N. The maximum measured power, force and speed for high weight have been: 760 W, 1700 N and 0,8 m/s, Fig. 2(b). For low force tests the actuator can move faster with a maximum speed of about 1.2 m/s. 2. Series Elastic Yobotics SEA 23-23 In order to test this actuator we have used the antagonic HEA-01 exerting a load, as it is shown in Fig. 1(left). The antagonic load has been a constant load force with viscous damping, FHEA01 = Fload − b · vHEA01 . Where b is a viscous coefficient with value b = 607, 6 N/(m/s). Taking into account that the SEA 23-23 can exert a theoretical maximum force of 1250 N, Fload has ranged from 600 N to 1000 N. The SEA maximum measured speed has been about

13:37

WSPC - Proceedings Trim Size: 9in x 6in

clawar2010

0,8

800

0,7

700

0,6

600

0,5

500

04 0,4

400

ower [W] Power

4

Actual Actuator uator Speed [m/s]

300

0,3 Force - Speed Force - Power

200

0,2

100

0,1 900

1050

1200

1350

1500

1650

Actual Actuator Force [N]

(a)

(b)

1000

88

900

85

800

82

700 600 0,075

Force (N) Power (W) 0,105

Power (W)

Fig. 2. (a) Experimental setup used to test HEA-01 (b) The red line shows the maximum power developed by the actuator during the tests. The dashed blue line shows the values of speed and strength for which the actuator HEA-01 delivered the corresponding maximum power. Two experimental points in this graphic correspond to the same test if they are located in the same vertical straight line. The cross markers show speed and force ranges for which the developed power is close to its corresponding maximum value, specified in the red line

Force (N)

May 24, 2010

79 76 0,135

Speed (m/s)

(a)

(b)

Fig. 3. (a) Photo of the tested Electric Series Elastic Actuator. SEAs use an elastic element to exert force from the driving element, which in this case is a brushless DC motor, to the load. (b) Experimental data from SEA-23-23’s power experiments. Every experiment has been characterized with a set of average values of force, power and velocity. Calculations have been done based on the ESEA’s force and position measurements.

0.15 m/s. The main utility of the viscous damping is to avoid undesirable high values of HEA’s speed that could damage the SEA. Experimental data obtained, Fig. 3(b), shows taht SEA 23-23 can deliver its maximum power

13:37

WSPC - Proceedings Trim Size: 9in x 6in

clawar2010

5 Fluidic muscle − Charge with damping test 25 Load 0 60 N 120 N 180 N

20 15 Power (W)

May 24, 2010

10 5 0 −5

(a)

0

0.5

1 t(s)

1.5

2

(b)

Fig. 4. (a) Antagonistic-variable load actuator (HEA-01) (b) Power delivered by FESTO MAS-20-200N-AA during the experiments. Only positive lectures are taken in account, since negative values correspond to power delivered by the HEA-01. Performed calculations used the angular position and the torque acquired with the test bench’s torque sensor. Maximum power values are similar to expected theoretical values, from8 and the actuator’s datasheet

for a wide range of load forces. This value, 88 W, is low compared to 150 W, its expected theoretical value, obtained from9 and the DC motor datasheet. 3. Pneumatic artificial muscle FESTO MAS-20-200N-AA Power experiments on the PAM have been carried out by means of the assembly shown in Fig. 1(left). The HEA-01 is used to simulate a constant load with damping. The damping effect increase the experiment’s security to PAM, reliability and repeatability. Experimental data is shown in Fig. 4(b). 4. Magnetic shape memory actuator Adaptamat MSMA Test Actuator The MSMA actuator has been tested focusing on periodic actuation. Therefore, the magnetic core is energized with a periodic square pulse current. The pulse ”high”-time during tests was T = 40 ms, so that all the response characteristics could be measured. Further information about how the tests were carried out and about the actuator response can be found in.10 The measured maximum capabilities of this actuator are: 0.8 mm stroke, 6 N maximum force, 0.45 W maximum mechanical delivered power, 100-175 Hz maximum periodic working frequency and 4 ms response time. 5. Results and Conclusions This paper presents the experimental determination of significant characteristics of some emerging actuators for empowering robots. Tables 1, 2 and

May 24, 2010

13:37

WSPC - Proceedings Trim Size: 9in x 6in

clawar2010

Mechanical energy per cycle [mJ/cycle]

3

2.5

4

3

2

2 1.5 1 10 0 7

1

8

6

5

4

6

3

2

Load force [N]

1

4

Amplitude of the command tension pulse [V]

0.5

Mechanical energy per cycle [mJ/cycle]

6

3 3.5 3

2.5

2.5 2

2

1.5 1 1.5 0.5 6 6.5

5

6

4 5 2

(a)

0.7

1 0.8

0.6

0.6

0.5

0.4

0.4 9

0.2

0.3

8 7 5

5 4

Load force [N]

3

2

Amplitude of the command tension pulse [V]

1

0.2

6

Amplitude of the 4 command tension pulse [V]

0.1

14

16

Characteristic times [ms]

MSM Steady−state elongation [mm]

0.8

6

4.5

(b) 0.9

0 7

1

5.5

3

Load force [N]

14

12

12

Settling time

10 8 6 4 2

10

X: 8.5 Y: 5.008 Z: 0.7

8 X: 5.5 Y: 2.085 Z: 3.92

Magnet delay

6 4

Rise time

0 10 8

Amplitude of the command tension pulse [V]

6

(c)

4

8

6

4

2

2

0

Load force [N]

(d)

Fig. 5. The amplitude of the square pulse current can be estimated with a conversion constant of about 1 A/V from the amplitude of the command tension pulse. (a) (b) mechanical energy delivered per current pulse, (b) represents a second set of experiments in a high efficiency zone. (c) MSM steady-state elongation. (d) Characteristic response time of the actuator during the experiments. The authors defend that the actuator’s high settling time is due to a negative effect of the test bench load force, which is exerted by a spring.

Table 1.

HEA-01 SEA-23-23 MAS-20-200N-AA Adaptamat Test Act.

Experimental data of interest of the tested actuator technologies Weight (kg)

Total Weight (kg)

Lmax (cm)

Lwork (cm)

Max. Force (N)

Max. Power (W)

1,59 1,36 1,31 0,32

122 12,86 28,3 0,97

70,2 49,9 47 8,65(2)

15,9 16,3 4,43 0,8

1700 1250 206 6

760 90 23,8 0,45

3 briefly enumerate the results of the conducted experiments. Theoretical values, whenever not cited, have been obtained from previous publications.8 Specific power is an important parameter for design and conception in service robotics. As the experimental results show, this parameter is especially overestimated in theoretical studies. The cause is a too low expected value of the actuators weight, probably not considering power electronics

May 24, 2010

13:37

WSPC - Proceedings Trim Size: 9in x 6in

clawar2010

7 Table 2.

Specific power and force comparison

Specific power (W/kg) Exp. without source

Type HEA-01 SEA-23-23 MAS-20-200N-AA Adaptamat Test Act.

478 66,2 18,2 1,41

Exp. with source 6,22 7,00 0,841 0,464

Theoretical 1039 1359 200 -

Spcific force (N/kg) Exp. without source

Exp. with source

Theoretical

1069 919 157 18,8

13,93 97,2 7,28 6,19

2000 1000 5000 21,9

Note: First column uses only actuators weight and the second the total actuators weight, which includes power source weight.

and supply systems. Regarding to this problem, the power-to-weight ratio can be improved, in all four cases, using multiple actuators with the same power source. Strain has also been overestimated in theoretical studies. There are still some comments to be done regarding individual actuators: • HEA-01: it would be really interesting to compare the reliability of different force measurements in hydraulic actuators. For example, spring deflection (SEA type), strain gauge and internal pressure sensors • SEA-23-23: a more complete analysis of the actuator is required, in order to identify the element that is causing the power looses from 150 W to 90 W. Replacing it for a more energy efficient element will Table 3.

Settling time and strain comparison Settling time (ms)

Strain (%)

Type

Experimental

Theoretical

Experimental

Theoretical

HEA-01 SEA-23-23 MAS-20-200N-AA

16∗ 35∗ 750 tp : 450 4∗∗

50∗9 35∗9 500

22,2 33,3 9,43

70 70 25

1,5

0,92(4)

1,1-4,6(5-20)

Adaptamat Test Act.

Note: The strain theoretical values are above experimental data. PAM’s strain is increased when using bigger artificial muscles. The Adaptamat actuator uses two MSMA samples with a length of 20 mm. This is the reason why the values between parentheses are reported to this length. ∗ The settling time of the indicated technologies has been obtained under force control. ∗∗ As stated above, the values of the measured settling time may be affected by type of load force exerted by the test bench, which is developed by a spring

May 24, 2010

13:37

WSPC - Proceedings Trim Size: 9in x 6in

clawar2010

8

consequently increase the maximum power delivered by the actuator. • MAS-20-200N-AA: it is necessary to get a pressure pneumatic servovalve to tests its settling time in closed-loop control. Also, using a bigger PAM may increase its power-to-weight ratio. • Adaptamat MSMA Test Actuator 5/7N: the experimentally measured power efficiency of this actuator was about 0,5% to at most 1% which is too low.10 This technology requires further research in power units, magnetic cores, and the development of new less demanding MSMA. 6. Acknowledgments This work has been founded by AECID under grant PCI-Iberoam´erica D/026706/09 References 1. A. Zoss, H. Kazerooni and A. Chu, On the mechanical design of the berkeley lower extremity exoskeleton (BLEEX), in IEEE Proceedings on International Conference on Intelligent Robots and Systems (IROS), 2005. 2. H. Herr, C. J. Walsh and K. Pasch, An autonomous, underactuated exoskeleton for load-carrying augmentation, in 2006 IEEE/RSJInternational Conference on Intelligent Robots and Systems, (Beijing, China, 2006). 3. M. Raibert, K. Blankespoor, G. Nelson and R. Playter, Bigdog, the roughterrain quaduped robot, Proceedings of the 17th International Federation of Automation Control. (2008). 4. K. J. Waldron, J. Estremera, P. J. Csonka and S. Singh, Analyzing bounding and galloping using simple models, Journal of mechanisms and robotics 1 (2009). 5. A. M. Dollar and H. Herr, Lower extremity exoskeletons and active orthoses: Challenges and state-of-the-art, Ieee Transactions on Robotics 24, 144 (2008). 6. S. . Hirose, K. . Ikuta and Y. . Umetani, Development of shape-memory alloy actuators. performance assessment and introduction of a new composing approach, Advanced Robotics 3, 3 (1988). 7. I. W. Hunter, J. M. Hollerbach and J. Ballantyne, A comparative analysis of actuator technologies for robotics, Robotics Review 2 (1991). 8. E. Garcia, H. Montes and P. G. de Santos, Emerging actuators for agile locomotion, in 12th International Conference on Climbing and Walking Robots and the Support Technologies for Mobile Machines, (Istanbul, Turkey, 2009). 9. D. W. Robinson, thesis design and analysis of series elasticity in closed-loop actuator force control, PhD thesis, Massachusetts Institute of Technology, 2000. 10. J. Pestana, R. Bomb´ın and E. Garcia, Characterization of magnetic shape memory alloys (msma) oriented to periodic actuation, in 12th International Conference on New Actuators. 6th International Exhibition on Smart Actuators and Drive Systems, (Bremen, Germany, 2010).