Development of a novel test method for skin safetyverification of physical assistant robots*

Xuewei Mao, Yoji Yamada, Yasuhiro Akiyama, Shogo Okamoto and Kengo YoshidaDepartment of Mechanical Science and Engineering

Nagoya UniversityNagoya, Japan 464-8603

Email: xueweimaomao@gmail.com{yamada-yoji, akiyama-yasuhiro, okamoto-shogo}@mech.nagoya-u.ac.jp

yoshida.kengo@e.mbox.nagoya-u.ac.jp

Abstract—The study mainly concerns a novel blister gener-ation method for the establishment of a safety verification test.This test system focuses on the skin friction trauma, which hasbeen found in the operators of physical assistant robots due totheir misuses or overlong operating, and can even deteriorateinto serious dermatopathya without preventing and prompttreatment. In the present study, first, we prove that blisterscan be generated in the porcine skin. With the histologicalsection of the porcine skin getting micro blisters, the symbolsof porcine skin’s blister in the dermis-epidermis structure areidentified with those of human blisters, and summarized forapplying this substitutive material to help researchers avoidthe ethical controversy of adopting human subjects. Second,the characteristics of blister generation in the porcine skinwere further confirmed with a series of time-gradient tests.After a condition-matching calculation, these characteristicsare verified to play a role on extending that of human blister inprevious studies and we integrate our results with the previoushuman blister generation data and clarify an inherently safecondition region in the shear stress - time relationship. Third,with this novel method, we successfully complete a safetyvalidation test on a ”stand up and sit down” activity of auser wearing a physical assistant robot, which finally provesour findings are of practical usability in regard to a possiblephysical stress hazard.

I. INTRODUCTION

When physical assistant robots are worn by their usersduring the assistance, the close physical human-robot inter-action will help the robot realize a more effective transfer ofpower to the human musculoskeletal system [1], and assistthe user’s movements in a way similar to the natural one.However, it has been reported in some studies that the skinwounds such as friction blisters are likely to be generated inthe use of the robots’ assistance with such a close physicalinteraction, e.g. for a wearer suffering cirrhosis of the liver[2].

It is pointed out in ISO 13482 [3] that hazards due to thephysical stress may result in injuries, and its measurementis listed as one of the items that are requested for safetyverification and validation. Some studies make achievements

*This work was supported by NEDO (New Energy and IndustrialTechnology Development Organization)

on the safety issues of relatively slight injuries concerningthe human-robot coexisting system, such as the human paintolerance limit tested by a biomechanical method [4] ortruly static measurement [5], the safe contact realization withefficient human-robot system [6], and more general limit forstatic and dynamic human-robot contact [7]; however, theyonly focus on the normal stress issues and little attention hasbeen paid on the hazards of shear stress by the existing safetyassessment standard of machinery. By now, there has been noreports on the safety verification test method against the skinwounds mainly caused by shear stress, such as friction blis-ter. Therefore, in this study, we focus on establishing a safetyverification mainly related to the blisters generated by thecontinuous friction on the human-robot contacting surfacing.Considering the fact that the experimental subjects have tosuffer hurts during the friction blister test, a replacement ofthe human subject is needed to avoid this ethical controversy.Several special models, dynamic non-linear finite-elementmodel [8] and synthetic skin simulant platform approach [9],have already been built to simulate the skin blister generationprocess. They can help to estimate the influence of textiles,surface treatment, loading frequency and other factors onblister qualitatively. But if we want to establish a reliablesafety standard against friction blister, a new substitute muchcloser to the human condition is still needed for analyzingblister generation conditions quantitatively. In this study,porcine skin is chosen as the experimental subject because ofits similar physiological and biological property with humanbeing skin from epidermis to dermis [10]. Together with theremaining individuality of different experimental subjects,porcine skin can supply more reliable information than thoseexisting blister simulants.

Although several experiments were conducted to studythe friction genesis under various skin exposure conditions[11]–[13], none of them is similar to the actual operationsituation of a physical assistant robot. Thus, no previousstudies can help us set up a particular safety verification forit. Since the generation of friction blister can be affectedby various factors, such as interactive force combination[11], environmental condition [11], [12] and individual skincharacteristics [13], [14], it is necessary for us to firstlyrepeat the interactive force between a physical assistant

319978-1-4799-1808-9/15/$31.00 c©2015 IEEE

robot’s cuff and its wearer’s skin during an actual operation.By conducting tests on experimental subjects under therubbing conditions similar to the actual situation, we canpossibly provide a more accurate skin trauma which is alsoclose to the actual one.

In this study, we develop a novel blister generationmethod for establishing a safety verification test systemfocusing on skin friction trauma. First, Section II describesthe entire experimental setup, which can control the rub-bing conditions applied on experimental subjects with highaccuracy, and reliably confirm the porcine skin’s epidermis-dermis structure after test. Second, Section III details theexperimental design and proposes an efficient way to testthe effects of various shear stress-time conditions under asituation that the porcine skin cannot generate macroscopicblisters on its surface. Based on the summarization of ourresults, the symbols of porcine skin’s blister in the epidermislayer are characterized and an inherently safe conditionregion in the stress-time relationship is clarified. Then,Section V demonstrates that this verified safe region canalso be applied on the safety verification for human users.This novel test method’s feasibility is finally proved by apreliminary application of a safety validation for a ”stand-up and sit-down” motion of a physical assistant robot’soperators.

II. EXPERIMENTAL MATERIAL AND METHOD

The porcine skin as experimental subject was excisedfrom the anterior part of a pig’s shank shortly after its death,and delivered to our lab within sixteen hours under 4 degreecentigrade. Before the experiments, the skin surface wasshaved by a scissor, cleaned by alcohol, and the excessivefat underneath was also cut off. In our study, the repre-sentative direction (shear direction) of force component isselected for emulating the skin surface rubbing phenomena.For conducting a horizontal reciprocating rubbing action,Discovery Hybrid Rheometer (DHR) (TA Instrument Co.)is used to exert oscillating rotational rubbing on the porcineskin with high calibration accuracy. In order to fix the skin onDHR and avoid the unexpected movement during rubbing,the porcine skin for the test was extended and then firmlyclamped between the plate with a sheet of sandpaper (40#)attached and a fixation frame as shown in Fig.1. After acertain period of rubbing under oscillating shear force, thestructure of the porcine skin will be examined by serialsectioning at intervals of 10 μm and the micro-section isanalyzed by haematoxylin and eosin (H & E) and the stainedsections are observed with a micro-scope sequentially.

III. EXPERIMENTAL DESIGN

The shear stress distribution of the porcine skin surfaceunder the stainless steel plate of DHR can be calculatedaccording to skin’s viscosity η , shear rate γ̇ , angular velocityΩ, skin’s thickness H and the distance from the detectedpoint to the center of the contact circle r [15]

τ(r) = ηγ̇(r) = ηΩrH

= ηΩH

· r (1)

Fig. 1. Diagram of porcine skin fixation and main part of experimentalapparatus

The torque of the rubbing head, T, can be measured byDHR and calculated with the contact area A

T =∫

dA · τ(r) · r = ηΩH

·∫

r2dA (2)

Therefore, the shear stress on the surface under rubbingcan be finally calculated as

τ(r) =2TπR4 · r (3)

where R is the radius of contact area.

Taking the individuality of porcine skin into consider-ation, only one trial for one shear stress-time point is notreliable enough to be used as verification data. Besides,fixing time gradient is too stiff to quickly approach thepossible time region for initial blister generation. Therefore,an improved experimental method was designed to solve theabove problems in a more efficient way.

First, since 3 is the least number which can help usidentify whether it is easier to generate blister under specificconditions, we tested 3 pieces of porcine skin at everystress-time point. Second, for each stress value, a seriesof time points were tested to confirm the blister generationconditions. During this process, we gradually modified therubbing time test by test. In order to avoid testing too manystress-time points which are too easy or too hard to generateblisters, the modification of rubbing time for the next testis based on the previous results. For example, if all 3 or2 of 3 samples get blisters after rubbing, the next step’srubbing time will be shorten by 100s or 50s; otherwise, itwill be lengthened by 100s or 50s when 0 or 1 of 3 samplesget blisters. Besides, some special principles for modifyingrubbing time were also applied to avoid repeated tests. Thedetailed experimental procedure flow (e.g. 10 time pointstested under one shear stress value) and its main parametersare shown in Fig. 2 and Table I, inside of which, Tbn andbk are set as 0 at first.

IV. EXPERIMENTAL RESULTS AND DISCUSSION

A. Feasibility of porcine skin as an experimental frictionblister model

After every rubbing test, the experimental subject wasfixed with 10% formalin for 8h, and then 20 % sucrose and

320 2015 IEEE International Conference on Rehabilitation Robotics (ICORR)

Fig. 2. Experimental procedure flow chart

30% sucrose for 8 hours sequentially and sectioned witha cryostat (Leica Microsystems Co.) at 10 μm. The skinstructure presence was then detected by H & E staining.The section of original porcine skin without any rubbingand the microscopic appearance of porcine skin after thefirst rubbing experiment are shown in Fig. 3 and 4.

It can be seen from Fig. 3 that the main epidermalstructure of porcine skin, including the stratum corneum,stratum granulosum, stratum spinosum, and epidermal basallayer, keeps intact before rubbing test. As supposed in theprevious experimental blister studies on normal human skin,friction force can be transmitted through ”stratum corneumand granulosum” of epidermis, degenerate stratum spinosum,and finally produce clefts between stratum granulosum andbasal layer. Once being filled up with fluid from dermis,those clefts will develop into friction blister rapidly [11],[12]. Compared with the section of original porcine skinshown in Fig. 3 and the section of the porcine skin after900s of rubbing under shear stress value 3.2×104Pa shownin Fig. 5, it can be observed clearly that there is severesplit generated in the stratum spinosum layer of the porcineskin which was rubbed under shear stress value 3.2×104Pafor 1200s (Fig. 4). According to Marion’s study [12], theblister cleft or cavity always appears at the same intra-epidermal layer, with the blister roof composed of thestratum corneum, stratum granulosum, and a segment oftraumatically degenerated stratum spinosum. Therefore, thecleft in Fig. 4 with intact stratum granulosum and stratum

TABLE I. MAIN PARAMETER OF IMPROVED EXPERIMENTAL

METHOD

k the order of trialTk the rubbing time of kth trialTbn the rubbing time of getting n samples with blisterbk the number of blister sample got in kth trialΔt time gradient

Testi estimation of initial blister generation time

Fig. 3. Appearance of original porcine skin(intact stratum spinosumindicated by arrow)

Fig. 4. Intra-epidermal clefts (indicated by arrow) after 1200s of rubbingunder shear stress value 3.2×104 Pa

Fig. 5. Intact stratum spinosum (indicated by arrow) after 900s of rubbingunder shear stress value 3.2×104 Pa

2015 IEEE International Conference on Rehabilitation Robotics (ICORR) 321

TABLE II. THE RESULTS OF EXPERIMENTS TESTED WITH THE

IMPROVED EXPERIMENTAL METHOD

Shear stress [104Pa] Rubbing time [s]

Number of

blister-samples

after rubbing

4.2

300 1

350 2

400 3

4.0

450 1

500 2

600 3

3.6

500 1

550 2

600 3

3.4

450 1

500 2

600 3

700 3

3.2

950 1

1000 2

1100 3

1200 3

3.0

1400 1

1600 2

2000 3

corneum, can be regarded as an evidence of the first stageof blister generation. Because of the biophysical propertychange in vitro compared with in vivo experiment, little fluidremains to fill those clefts and raise the stratum granulosumand corneum as obvious blister roof. Even though it isdifficult to get macroscopic blister on dead porcine skin withour test, the cleavage indicating blister generation is also ableto prove the effectiveness of this new experimental model inobtaining the friction blister generation time. This kind ofcleavage is regarded as the symbol of blister generation inthe following experiments.

B. Results of the Experiment Conducted with ImprovedMethod

The first shear stress value tested with improved experi-mental method is 4.2×104Pa and the first rubbing time forthis shear stress was set as 400s and all of the 3 samples gotblisters and one of them was heavily destroyed. Based onthe action selection principle in Section III, the next rubbingtime was set as 300s, which was 100s less than the lasttime value. This time, only one sample generated blisterswhile the other two samples’ epidermis still kept intact.Therefore, the next rubbing time was set as 350s, whichwas 50s longer than the last one. On the contrary, only onesample kept intact after the third step and the other two evenlost their epidermis. Until now, with only 3 tests, the timepoints around the most worth being tested points were bothtested, which further verified the efficiency and feasibility ofthe improved experimental method.

Five more shear stress values were also tested with thesame improved experimental method. The results of thisseries of experiment were summarized by Table II.

C. Calculation of Inherently Safe Condition Thresholdagainst Blister Generation on Porcine Skin

Since we tested 3 samples for every shear stress-timepoint, the points generating 1, 2 or 3 blister-samples can beregarded as the conditions which can cause blisters in 33.3%,66.7% or 100% of experimental subjects respectively. Sup-posing that the rubbing time is normally distributed as blistergeneration conditions for a certain shear stress, a rubbingtime point, which can only cause blister in less than 0.13%of subjects, can be calculated according to the ”3σ -rule”,and such a time point can be regarded as an inherently safepoint against blister generation under a certain shear stressvalue. All the safe points for the tested shear stress valueswere calculated respectively and summarized in Table III.

V. IDENTIFICATION OF BLISTER GENERATIONEXPERIMENTAL CONDITIONS FOR SAFETY

VERIFICATION

In order to confirm the feasibility of porcine skin, thecharacteristics of blister generation should match with thoseobtained in the previous human friction blister test, whichwas conducted by Naylor [11]. In our experimental model,we set shear stress as a criterion of the interactive forcebetween the porcine skin and the rubbing head. In the caseof human skin test reported in the Naylor’s study, onlythe compression and average friction force were describedin detail. Therefore, a condition-matching calculation isnecessary for us to convert the friction force in Naylor’sstudy to the corresponding shear stress for comparing theirblister generation characteristic.

A. Matching the Experimental Conditions

Since the rubbing head applied in the Naylor’s study isa small hemisphere [16], the tested shear stress value can becalculated approximately based on Hertz theory [17]

τ(r) =3μP2πa2

√a2 − r2

a(6)

where P is the total load compressing the skin surface, μis friction coefficient, and a is the radius of contact circle,which can be calculated by the following formula

a = (3PR4E∗ )

13 (7)

where R is the relative curvature, which can be calculatedby the principal curvature of the two contact surfaces, R1and R2

1R=

1R1

+1

R2(8)

TABLE III. INHERENTLY SAFE CONDITION AGAINST FRICTION

BLISTER ON PORCINE SKIN

Stress [104Pa] 4.2 4 3.6 3.4 3.2 3

Time [s] 227.6 329.5 427.6 329.5 829.5 918.5

322 2015 IEEE International Conference on Rehabilitation Robotics (ICORR)

E∗ can be calculated by the elasticity modulus of the twosurfaces in contact with each other, E1 and E2, and Poisson’sratio, ν1 and ν2

1E∗ =

1−ν21

E1+

1−ν22

E2(9)

The rubbing head in Naylor’s study was made by poly-thene, whose elasticity modulus is much higher than that ofhuman skin, and thus its contribution to E∗ can be ignored.Considering that the tested human skin area is approximatelythe middle third of the anterior surface of the tibia [11],the elasticity modulus and Poisson’s ratio of human skinare assumed to be 130kPa and 0.45, respectively [18]. Afterthe above calculation, we can finally obtain the relationshipbetween the rubbing time required to produce blisters andthe corresponding shear stress in Naylor’s study.

B. Analysis of Porcine Skin’s Feasibility as a Reliable Modelfor Safety Verification Test

We calculate to draw a regression curve from the in-herently safe time points, regard it as an inherently safethreshold, and combine it with the human subject’s blistergeneration conditions to produce a new shear stress-timecharacteristic for blister generation, which is shown in Fig.6. It can be seen that the threshold is located in the safetyregion left of and below the points representing humanblister generation. We can claim from the results that it isfeasible to utilize porcine skin as a reliable safe verificationtest model to confirm that a certain operation action issafe enough to avoid generating friction blisters. However,more appropriate calculation for the inherently safe timepoints and drawing a threshold line may require furtherconsideration because there is one point which is not soconsistent with the threshold line.

VI. PRELIMINARY APPLICATION OF PORCINE SKIN FORSAFETY VALIDATION WITH MANIPULATOR

After confirming the feasibility of using porcine skinfor a safety verification test, we conducted an experimentvalidating a cuff attachment in the use of a wearable robot.

Fig. 6. Combination of porcine skin’s inherently safe conditions and humansubject’s blister generation conditions

We focused on a human subject’s ”stand-up and sit-down”motion which is considered to exhibit a longest level ofrelative displacement between the cuff and the skin surfaceof the subject. According to the previous study of ourgroup [19], the interaction force between the robot’s cuffand the user’s thigh skin can be exerted repeatedly by amanipulator with high accuracy at the endtip of which acuff with a contact sensing function is mounted. We choosethis apparatus to realize a safety validation test.

A. Experimental Setup for Preliminary Application

The experimental setup of this application is shown inFig. 7, where the urethane is used to simulate the cuffmaterial of a physical assistant robot, the polyurethane gel(hardness is 0 by use of ASKER Durometer Type C) ispasted on the dummy leg. The porcine skin, after the samepreprocessing introduced in Section II, is pasted and fixedon the gel layer with tape and band.

B. Estimation of Experimental Results

During the series of experiments, three pieces of porcineskin were tested for safety validation of a cuff-human skinsurface system, and every test lasted for 1800s. Since thesmallest contact area between urethane and porcine skinis about 12×10−4m2 and the largest shear force appliedon porcine skin is 27.7N, the highest shear stress appliedon porcine skin is calculated approximately as 2.3×104Pa.According to the inherently safety region shown in Fig. 6,no blister can be generated on porcine skin inside of 3350sunder the highest shear stress. Therefore, the porcine skinshould keep intact after this rubbing test.

C. Experimental Results for Safety Validation

The safety validation test were repeated three times.After every test, three samples were respectively taken fromthe higher, middle and lower parts where the porcine skinwas contacted with the urethane-made cuff shown by thered rectangular part in Fig.8. Every sample taken from thepurple rectangular part in Fig.8 contains a small additionalarea where the skin wasn’t be rubbed during the test. Onemore sample was also taken from the uncontacted part shownby the green circle in Fig.8, which is located far from therubbing area. All of these samples were processed withthe same method introduced in Section V, but not a single

Fig. 7. Experimental setup of preliminary application with porcine skin’ssafe verification model

2015 IEEE International Conference on Rehabilitation Robotics (ICORR) 323

Fig. 8. Sample position for safety validation

blister was found inside of them, which indicates that 1800sof repeated operation of ”stand-up and sit-down” are safeenough under the special conditions [19]. The actual testresult is in agreement with our estimation of the safety regionfor a larger number of repeating times that was not evidentbased upon the original blister generation conditions in theNaylor’s study.

VII. CONCLUSIONS AND FUTURE WORK

In the study, a novel blister generation method wasdeveloped for establishing a safety verification test, whichmainly focuses on the skin trauma for the user of physicalassistant robots during their operation. By a series of rubbingexperiments on the porcine skin with Discovery HybridRheometer, we first confirmed and summarized the possibil-ity and special characteristics of friction blister generationon the porcine skin. Besides, an improved experimentalmethod was designed to conduct a series of repeated testswhich can rapidly approach to the time range really worthtesting and clarify the initial blister generation time moreaccurately. After a calculation matching the human skin’sblister generation conditions with ours, it was revealed thatthe inherently safe threshold for the porcine skin’s blistergeneration was located far from the hazardous conditionsfor human skin’s blister generation, which further provesthe feasibility of our experimental system. With this novelmethod, we conducted a safety validation test on a ”stand upand sit down” motion of a user wearing a physical assistantrobot. The consistency between the estimation and actualresults of this test finally proves our findings are practicalenough for safety validation of physical assistant robots witha possible physical stress hazard.

Since now the way of estimating the inherently safeconditions in our study is only based on an generallyaccepted assumption that the rubbing time acting as a kindof blister generation conditions is normally distributed, inthe future, the calculation method for the inherently safethreshold will be reconsidered.

ACKNOWLEDGMENT

The authors would like to thank Dr. Cota Nabeshima, Cy-berdyne Inc., for discussing the possible injury in the frame-work of contact safety associated with physical assistantrobots. The authors are also grateful to Prof. Susumu Hara,

Nagoya University, for his helpful advice of data analysisimprovement, and Mr. Ishiguro Kenji, Nagoya University,for his technical support for human-robot data collection.

REFERENCES

[1] Tommaso Lenzi, Nicola Vitiello and Stefano Marco Maria De Rossi,et al., Measuring human-robot interaction on wearable robots: Adistributed approach, Mechatron., vol. 21, no. 6, pp. 1123-1131, Sep.2011.

[2] O. Kyouko, Current status of the introduction of lower extremityfunctional training with the robot suit HAL in our hospital, (inJapanese) Japanese Journal of Pressure Ulcers, vol.16, no. 3, pp.289, Aug. 2014.

[3] ISO/FDIS 13482 Robots and robotics devices - safety requirementsfor non-industrial robots - non-medical personal care robot, 2011.

[4] Y. Yamada, K. Suita, H. Ikeda, N. Sugimoto, H. Miura and H.Nakamura, Evaluation of pain tolerance based on a biomechanicalmethod for human- robot coexistece, (in Japanese) Transactions ofthe JSME, vol. 63, no.612, pp.2814-2819 , Aug. 1997.

[5] T. Saito and H. Ikeda, Measuring system and analytical method ofpain tolerance of mechanical stimulus for safe design of human-collaborative robot, SIAS2005.

[6] Y. Yamada, Y. Hirasawa, S. Huang, Y. Umetani and K. Suita,Human-robot contact in the safeguarding space, IEEE/ASME Trans.Mechatron., vol.2, no. 4, pp. 230-236, Dec. 1997.

[7] ISO/CDTS 15066 Robots and robotic devices - safety requirementsfor industrial robots - collaborative operation, 2013.

[8] Malcolm Xing, Ning Pan, Wen Zhong, Skin friction blistering:computer model, Skin Res. Technol., vol. 13, pp. 310-316, 2007.

[9] C. Guerra and C. J. Schwartz, Investigation of the influence of textilesand surface treatments on blistering using a novel simulant, Skin Res.Technol., vol. 18, no. 1, pp. 94-100, Feb. 2012.

[10] Barry Goldstein, Joan Sanders, Skin response to repetitive mechan-ical stress: a new experimental model in pig, Arch. Phys. Med.Rehabil., vol.79, no.3, pp. 265-272, Mar. 1998.

[11] P. F. D. NAYLOR, Experimental Friction Blisters, Br. J. Dermatol.,vol.67, no.10, pp.327-342, Oct. 1955.

[12] Marion B. Sulzberger, Thomas A. Cortese, Leonard Fishman, HushS. Wiley, Studies on Blisters Produced by Friction I. Results of LinearRubbing and Twisting Techics, J. Invest. Dermatol., vol. 47, no. 5,pp. 456-465, May. 1966.

[13] Akers, Colonel William A, Measurements of friction injuries in man,Am. J. Ind. Med., vol. 8, no. 4-5, pp. 473-481, 1985.

[14] Knapik, Joseph J., et al., Friction blisters, Sports Med., vol. 20, no.3,pp.136-147, Sep. 1995.

[15] Sokolnikoff I S, Specht R D., Mathematical theory of elasticity. NewYork: McGraw-Hill, 1956.

[16] P. F. D. NAYLOR, The skin surface and friction, Br. J. Dermatol.,vol. 67, no. 7, pp. 239-248, 1955.

[17] K.L.Johnson, Contact Mechanics. New York: Cambridge UniversityPress, 2003.

[18] F.T.Mak, GeorgeH.W.Liu, S.Y.Lee, Biomechanical assessment ofbelow-knee residual limb tissue, J. Rehabil. Res. Dev., vol. 31, no.3, pp. 188-198, Aug. 1994.

[19] K. Yoshida, Y. Yamada, Y. Akiyama, S. Hara, and S. Okamoto,Development of a safety validation test equipment for severityestimation of wounds caused by physical assistant robots, presentedat the 20th Robotics Symposia, Karuizawa,Nagano, Mar. 2015.

324 2015 IEEE International Conference on Rehabilitation Robotics (ICORR)