Вы находитесь на странице: 1из 10

Robotics and Computer–Integrated Manufacturing 49 (2018) 88–97

Contents lists available at ScienceDirect

Robotics and Computer–Integrated Manufacturing


journal homepage: www.elsevier.com/locate/rcim

Review

A statistical review of industrial robotic grippers


Lionel Birglen a,∗, Thomas Schlicht b
a
Department of Mechanical Engineering, Polytechnique de Montréal, Montréal, Québec, Canada
b
Department of Mechanical Engineering, TU Darmstadt, Darmstadt, Germany

a r t i c l e i n f o a b s t r a c t

Keywords: With the recent introduction of ambitious industrial strategies such as Horizon 2020 and Industry 4.0, a massive
Gripper focus has been placed on the development of an efficient robotic workforce. Amongst all the operations robotic
Grasping systems can take care of, handling remains a preferred choice due to a combination of factors including its often
Market review
repetitive nature and low skill requirement. The associated demand for grasping tools has led to an ever increasing
Statistical analysis
market for manipulation end-of-arm tooling from which a handful of industry giants have emerged. Based on data
publicly accessible from the catalogs of several well-known companies, this paper aims at presenting a review
on the characteristics of pneumatic, parallel, two-finger, industrial grippers. Included in the specifications under
scrutiny in this paper are: stroke, force, weight, as well as a performance index referred to as the C-factor. This
last index is a combination of three of the aforementioned characteristics and has been proposed in the literature
as a measure of the efficiency that a gripper is capable of reaching. As will be shown, by analyzing hundreds of
specifications it appears that, indeed, the range of C-factors of the grippers built by one company can be often
consistently different from these of competitors. Furthermore, an important bias for certain typical specifications
can be observed in most of the grippers which seems at odd with the requirement of modern robotic systems.
This latter remark will open up a closing discussion proposed in the last part of this paper on the future evolution
of grippers based on emerging new products.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction gripper being able to seize only a limited set of objects. Ultimately, the
goal of all robot grippers remains however the same: holding and secur-
Industrial robotic grippers have a pivotal role in modern automa- ing the pose of a workpiece relatively to a (robotic) manipulator so the
tion since they constitute the end-of-arm of robotic manipulators and latter can move this workpiece from one place to the other where it is
thus, they are in direct contact with the workpiece to ndled. Grippers then released. The development of gripper technology has been closely
as end-of-arm tools have to perform their tasks under demanding re- connected to the development of robotics since the kinematics of robots
quirements in modern mass production because handling operations do strongly influence the requirements for the gripper [1]. Considering the
not directly increase the market or intrinsic values of the workpieces. wide range of robotic architectures one can find on the market it is
Therefore, grasping and manipulation should be achieved not only se- not surprising to see the plethora of commercially available grippers
curely but as fast as possible to reduce cycle times. An important issue that have been developed [2]. However, it is unclear how varied the
is for the output rate of the production lines where robots are used not specifications of all these grippers are and how their performance can
to be negatively impacted by the handling operations [1]. This require- be related or even quantified. The range and statistical values of the
ment for speed becomes a challenge when the workpieces do not have available specifications of available commercial products have not been
predictable or consistent properties, as for instance in food processing, previously listed to the best of the authors’ knowledge. The aim of this
goods packaging, recycling, or mining industries. The variability of the paper is to shed some light on this issue and propose an analysis of what
objects a robot or automated machine have to seize led manufacturers the market currently offers to its customers. While there exist a handful
of grippers to expand their catalogs to accommodate this issue and now, of references reviewing industrial grippers from a general point of view,
commercially available grippers come in innumerable shapes or sizes. e.g. [3–8], with a few exceptions such [9] most of them are rather old
Current grippers range from the minuscule to the gigantic. To concili- and consistently focus on specifically targeted applications/aspects, e.g.
ate all these different products and their various designs also emerged [10–12], or how the gripping motion is produced instead of what are the
the necessity of tool-changing, i.e. swapping one tool for another, each actual performances of these grippers in terms of force, stroke, weight,


Corresponding author.
E-mail address: lionel.birglen@polymtl.ca (L. Birglen).

http://dx.doi.org/10.1016/j.rcim.2017.05.007
Received 1 January 2017; Received in revised form 15 May 2017; Accepted 25 May 2017
0736-5845/© 2017 Elsevier Ltd. All rights reserved.
L. Birglen, T. Schlicht Robotics and Computer–Integrated Manufacturing 49 (2018) 88–97

power, etc. Again, this review aims at filling that gap by providing actual 12; 4%
numbers for these specifications. 31; 11% Schunk
An interesting and maybe surprising point to note is that most com- 88; 30% phd_inc
panies which produce robots do not produce grippers themselves [2]. IPR-worldwide
49; 17% festo
There are of course a few overlaps here and there but as an indus-
try, gripper and robot manufacturers are generally distinct businesses. sommerautomac
Generally speaking, it seems that robot manufacturers and integrators afag
mostly rely on the products offered by gripper companies and only if 51; 18% 58; 20%
they cannot find a solution matching their requirements, a custom tool
is developed. Now, to see what is the current landscape of industrial
Fig. 1. Distribution of grippers by manufacturer.
gripper technology, one has to find a representative subset of the man-
ufacturers. However, therein lies the first obstacle: ranking these com-
panies by market volume or overturn is not possible. Indeed, little—if 4. two fingers,
any—data is available to the general public and sales’ numbers remain 5. out-of-the-shelf (no custom model),
sensitive and confidential figures. Furthermore, many companies pro- 6. no spring or magnetic assistance to the gripper force.
ducing grippers are still privately-owned and do not disclose any finan-
cial information. To circumvent this issue, the authors chose to have a An important point to notice is that the analyzed data is only based
look to the names of companies presenting their grippers at industrial on catalogs’ data and not sales figures for the grippers. Therefore, each
fairs and exhibitions such as: the Chicago Automate Fair, Tokyo IREX, gripper in the catalogs is equally considered while one model might ac-
and Hanover Messe. Recurring names were selected for this review. It tually be a bestseller product. This might appear as a weakness of this
was noticed that several of these companies are from Germany which the review but it does not necessarily introduce a significant bias in the re-
authors conjecture is due to the strong automobile production and geo- sults. Indeed, it appears reasonable to assume that the manufacturers
graphically close European market. Another point noticed is that there offer the most complete choice of grippers with the most commonly re-
is an impressive number of small and medium-sized enterprises (SME) quested features and therefore, a bestseller series of grippers is expected
but only a handful of global corporations. The list of companies we se- to have several declinations in the catalogs. Considering only catalogs’
lected was crosschecked with the online database Direct Industry using data does not ensure accuracy and the trustworthiness of manufacturers
the criteria presented in the subsequent Section and it was found out that is assumed when reporting specifications. Understandably, experimen-
with this methodology, almost all the companies offering a significant tally verifying each specification for each gripper is highly impractical
number of products were obtained. but this limitation must be acknowledged.
Processing the entire catalogs of all the gripper manufacturers was an After investigation, six manufacturers were chosen for this analy-
insurmountable task because of the very large number of products each sis, namely (in alphabetical order): AFAG, Festo, IPR, PHD, Schunk, and
one of them offers. Indeed, even disregarding all vacuum grippers and Sommer Automation (recently acquired by the Zimmer Group.) All these
special custom designs of limited interested outside their target market, companies provide the specifications of their grippers in catalogs pub-
e.g. for mechanical diggers or cranes, there are still literally thousands of licly available and a compilation was established in a digital database
products. Therefore, the scope of this review was further reduced down of all their models satisfying the aforementioned properties. Fig. 1 illus-
to a specific but very common design as will be discussed in Section 2. trates the share and number of grippers obtained for each manufacturer:
This will be followed by the presentation of statistical values of the spec- 12 for AFAG [13], 49 for Festo [14], 51 for IPR [15], 58 for PHD [16],
ifications pertaining to these grippers in Section 3. 31 for Sommer Automation [17], and 88 for Schunk [18]. The grand
Subsequently, after this presentation of the raw data, an actual com- total of all grippers considered in this review is therefore 289 and the
parison between manufacturers of all considered products can be un- complete list of models used is presented in Appendix A. The nominal
dertaken and the results is presented in Section 4. This comparison will pressure that was found for all grippers is 6 bars and thus, this value was
mainly focus on specifications such as the length of the stroke, gripping considered when establishing the generated force even if they could op-
force, weight of the gripper without fingers, and maximal length of the erate at other pressures.
finger that can be attached. Furthermore, the C-factor, i.e. a measure
of the energy that the gripper is outputting relatively to its mass, has 3. Data analysis
been computed for each of the grippers and will also be discussed in
Section 5. Finally, in Section 6, a discussion on the results and the fu- Once all the specifications are compiled, a statistical analysis of the
ture of industrial grippers is proposed. data can be conducted. Accordingly, an overview of the resulting an-
alyzed data is shown in Table 1 which lists minimal, maximal, mean,
2. Scope median and standard deviation of all the selected specifications of the
grippers. With these, one can define the ”most average” gripper of the
In this review, the authors chose to focus on a specific architecture sampled market to have a stroke of 20.8 mm, a force of 1,020 N, and a
of impactive tool, namely parallel grippers with two fingers and driven weight of 3.4 kg. However, if one compares these figures to the median
by pneumatic energy. The main reason for this is to narrow down the values, the results are noticeably different. Namely, the thusly defined
list of devices considered to a manageable size. Furthermore, this type median gripper has a stroke of only 9.5 mm, a gripping force of 320 N,
of grippers appeared from a preliminary investigation of manufactur- and a weight of 0.6 kg. A significant difference between mean and me-
ers’ catalogs to be the most common one, by far. In order to gather a dian values in a statistical analysis indicates that the repartition of the
meaningful list of specifications for the grippers fitting the scope of the data is not as simple as a bell curve. On closer inspection, this large dis-
review, the authors also chose to focus on the products readily avail- crepancy is due in this case to the fact that there is a significant bias
able from these catalogs, i.e. disregard special editions of more typical toward smaller values for many specifications and especially the stroke.
models such as oversized variants. As a summary, the properties of the Nevertheless, if small grippers fill the bulk of most of the manufacturers’
grippers considered in this review are: catalogs, very large products also exist but there are only few interme-
diate devices.
1. impactive, It should be noticed that in this analysis, the values indicated for the
2. driven by compressed air (pneumatic), stroke are actually for one finger only, following the standard designa-
3. a parallel finger movement, tion used in the catalogs, but the overall opening and closing ranges of

89
L. Birglen, T. Schlicht Robotics and Computer–Integrated Manufacturing 49 (2018) 88–97

Table 1
Statistical values of common specifications. 100%
90%

cumulated percentage
Average Median Std. dev. 80%
70%
stroke [mm] 20.78 9.55 35.10 60%
force [N] 1020.44 320 1938.01 50%
weight [kg] 3.41 0.59 8.43 40%
C-factor [J/kg] 6.91 5.68 4.98 30%
finger length [mm] 143.21 100 139.93 20%
closing time [s] 0.23 0.13 0.32 10%
air cons. [cm3/cyc.] 163.61 13.00 526.64 0%
power [W] 239.38 82.26 355.29 1 6 11 16 21 26 31 36 41 46
stroke size [mm]
Min Max Count

stroke [mm] 1 300 289


force [N] 6 15400 289 Fig. 3. Cumulated percentage of stroke lengths.
weight [kg] 0.01 70 289
C-factor [J/kg] 0.36 28.57 289
finger length [mm] 9 900 197
closing time [s] 0.01 2.30 270
air cons. [cm3/cyc.] 0.09 4600 170
power [W] 0.77 1840 170

Fig. 4. Histogram of gripping force in 100 N bins up to 4,000 N.

80%
cumulated percentage [%]

70%
Fig. 2. Histogram of stroke lengths up to 50 mm.
60%
50%
40%
the grippers are twice these numbers since two-finger grippers are con-
30%
sidered. The fact that most grippers offer very small strokes can be ex-
20%
plained by two reasons. First, in theory only an infinitely small stroke is
10%
required to grasp a known workpiece. Indeed, the fingers of the gripper
0%
can be designed with an offset leaving a gap that is exactly the size of the
1
51
101

201

301

401

501
551

651

751

851

951
151

251

351

451

601

701

801

901

1001
1051
workpiece when closed. Then, opening these fingers of only an infinites-
imal amount would release this part. Of course, in practice, positioning Force [N]
tolerance, precision, as well as structural compliance dictate that a cer-
tain margin of safety should be taken into account and a small but finite
Fig. 5. Cumulated percentage of force.
stroke be used. Second, in mass production where many of these grippers
are implemented, the cycle times must be reduced as much as possible
to increase production output rates. Since these grippers are typically The surface of the contact between the fingers and the workpiece
not precisely controlled in position but rather commanded with simple can typically be on either side of the fingers (referred to as internal
fully open / fully close instructions, small strokes typically yield shorter and external grasps) but this has no influence on this specification, only
times to grasp and release workpieces. These two factors explain why the mechanical design of the fingers is changed to accommodate the
the median gripper has a significantly smaller stroke then the average desired grip. Sometimes, the available force is slightly different though.
one: a lot of grippers are small with tiny strokes. Since robots have typically a good accuracy and excellent repeatability,
the grippers rarely have to compensate an imprecise positioning of the
3.1. Stroke workpiece, although exceptions exist, cf. Section 6.

The median stroke of all the grippers is only of 9.5 mm, while the 3.2. Force
average value is more than twice that at 20.8 mm. This latter value is
actually greater than 77.2% of all these of the 289 grippers considered As Table 1 shows, the range of gripping force in the gathered data
and 90% of them have a stroke smaller than 50 mm, as illustrated in goes from 6 N up to an impressive 15,400 N and the average is ap-
Figs. 2 and 3. Larger values of stroke are uncommon (but exist) and prox. 1,020 N. An histogram of the most common values in presented
spread sporadically until the maximal value found at 300 mm. Also, in Fig. 4. On the other hand, the median grasping force is only 320 N.
it should be emphasized again that these numbers must be doubled for Similarly to the discussion about stroke size, this indicates that there is a
the full difference between an opened and closed gripper. Therefore, the significant number of grippers in the lower-end of the spectrum. Indeed,
median stroke of approx. 10 mm means a total opening of 20 mm is this is confirmed by looking at cumulative percentages as shown in the
available between the fingers. That obviously does not mean that the Fig. 5: almost 70% of the grippers provide less than 750 N. Interestingly,
largest object the gripper can seize if of 200 mm since the fingers can this number is close to the maximal grip strength of males as reported by
be offset and leave a gap even when fully closed. NASA (729 N, right-hand, 95th percentile), see [19]. While this value

90
L. Birglen, T. Schlicht Robotics and Computer–Integrated Manufacturing 49 (2018) 88–97

The figures that were used for the statistics presented in Table 1 were
computed with a reduced number of grippers because not all manufac-
turers list this value in their specifications. From the gathered data, the
longest finger length is 900 mm while the shortest is 9 mm. Again,
the limitation on the finger length mainly stems from the magnitude of
the moment the gripper can physically resist before risking at best, a di-
minished service life or, at worst, mechanical failure. Parallel grippers
indeed often rely on precise motion guides which can be fragile.

3.5. Power

Fig. 6. Histogram of gripper weight up to 10 kg.


The power of each gripper was established by computing the air con-
sumption per cycle multiplied by the nominal pressure and divided by
the time for the cycle. A cycle was defined as fully closing the gripper
90% from a fully opened position and then going back to the fully opened
80%
configuration. Similarly to the finger length, not all manufacturers read-
cumulated percentage

70%
60%
ily provide values for the air consumption or cycle time. The values pre-
50% sented in Table 1 were computed with a limited sample of only 170 grip-
40% pers. The range of the power that was found is quite large: from 0.77 to
30% 1,840 W. Exactly, and understandably, as with force and stroke, power
20%
is mainly spread between lower numbers and a few outliers with large
10%
0%
values. This is illustrated in Fig. 8 where a clear high density of grippers
0.1 0.6 1.1 1.6 2.1 2.6 can be seen below 20 W. Due to this concentration, the median power
weight [kg] is at 82.3 W while the average is 239.4 W and this latter value defines
a threshold under which almost 69% of the grippers fall.
Fig. 7. Cumulated percentage of weight.

4. Comparisons
might seem large, it is rather modest for industrial purposes. Indeed,
parts manipulated by industrial robots can be quite heavier and a large 4.1. C-Factor
grasping force may be required to prevent slippage. It should also be
remembered that only a fraction of the grasping force, proportional to As defined in [1] the C-factor of a gripper can be computed as the
the friction coefficient, might be available to secure the workpiece and ratio of the force it produces over its weight and multiply this ratio by
counteract gravity depending on the design of the fingers. Additionally, the stroke. The value obtained is arguably a measure of the efficiency of
a common technique to seize large parts is to use several grippers at- the gripper and can be used for comparison between different products,
tached to a custom frame which itself is rigidly attached to the robot. designs, etc. The unit of the C-factor is J/kg and thus, it can indeed be
This also has the advantage of distributing the required grasp force to interpreted as the output energy the gripper is capable of producing per
all the grippers and therefore, to decrease local stress on both the work- unit of weight. As illustrated in Fig. 9, Schunk, PHD, and Festo all reach
piece and the grippers. This issue is especially critical when the center high values for the C-factors of their products1 . On the other hand, IPR,
of gravity of the part to handle is far from the gripper. The resulting Afag, and to a lesser extent, Sommer Automatic, all show typically lower
moment from the weight and internal forces can then be significant and C-factors. In the case of Afag, this can be explained by the fact that this
most manufacturer specify maximal values of the moment their gripper company offers grippers with smaller sizes only and, as it seems, it is
can tolerate around the different axes of a local frame. more difficult for small grippers to be C-factor efficient. Indeed, if one
looks at the definition of this index, it appears to favor large force and
3.3. Weight large stroke although a relative normalization is realized by dividing by
the weight.
The weight of the grippers is another property which is of great im- This figure is not a value judgement about the products of each com-
portance for robotic applications since it directly impacts the total mass panies since other practical aspects such as reliability might explain the
that can be manipulated. The lightest and heaviest grippers that were low values of the C-factor. The C-factor could be used to compare prod-
found weighted 8 g (!) and 70 kg respectively. The median gripper ucts with similar characteristics fitting the application at hand but com-
weights 0.6 kg and 61% of the grippers are lighter than 1 kg as il- paring the C-factor of a minuscule gripper for white room usage from
lustrated in Figs. 6 and 7. Up to the average weight of 3.41 kg can be one company to a very large one for foundry metalwork from another
found already 81% of all grippers. This clearly indicates that most grip- makes no sense. Fig. 9 however illustrates the range of values the C-
pers are very light, as expected. Indeed, for a constant maximal payload factors usually take in the industry.
of the robot, the weight of the gripper has to be subtracted from the Most of the grippers considered in this review have a C-factor be-
maximal weight of the workpiece that can be manipulated. Even within low 15 J/kg and going above this threshold seems to require special
the permitted payload of the robot, a lighter gripper will result in faster attention. Overall, the minimal C-factor in the data is 0.36 J/kg and
motions thereby decreasing cycle times. It is therefore natural that man- the best one 28.57 J/kg with an average of 6.91 J/kg and a median
ufacturers tend to favor lighter designs. of 5.68 J/kg, cf. Table 1. An interesting fact to notice is that with the
C-factor, the median and average values are not significantly different,
3.4. Finger length at least not as much as many of the previously presented characteristics.

The finger length is defined as the distance from the gripper palm to
the farthest location where a gripping force can be transmitted to the 1
It should be disclaimed that this performance index has been proposed by authors
workpiece in a direction perpendicular to the closing/opening motion. affiliated to the Schunk company.

91
L. Birglen, T. Schlicht Robotics and Computer–Integrated Manufacturing 49 (2018) 88–97

Histogram of Power - for each value up to 1000 W (160 of 170)


7

Frequency vs. Power - for each value unl 20 W Power - cumulated


6 7 percentage
6
100,0%

cumulated percentage [%]


5
5 80,0%

Count
4
60,0%
3
2 40,0%
4 20,0%
Count

1
0 0,0%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 1 101 201 301 401 501 601 701 801 901
3 Power [W] Power [W]

0
1 51 101 151 201 251 301 351 401 451 501 551 601 651 701 751 801 851 901 951
Power [W]

Fig. 8. Distribution of the power values.

0,00 5,00 10,00 15,00 20,00 25,00 30,00

Schunk phd_inc IPR-worldwide Festo sommerautomac afag

Fig. 9. C-Factor distribution by manufacturer.

18000,00
(0.59 kg) where the dense cloud of data points is visible. With Schunk,
line 1
16000,00
there also seems to be two trend curves, although not as much linear.
14000,00
The first one is close to a logarithmic curve (curve #1) and reaches
12000,00
10,000 N for a gripper with a weight of 35 kg. The second one, closer
force [N]

10000,00
line 2 to a straight line (curve #2) goes up to 2,000 N. This actually can be
8000,00
explained for this second trend if one looks more closely to the data: it
6000,00
mainly comes from a range of grippers with similar properties where the
4000,00

2000,00
stroke is variable (namely the PFH 150 - PFH 300 series) [5]. Two simi-
0,00
lar lines can also be seen for Festo and PhD with maximal values around
0 10 20 30 40
weight
50 60 70 80
3,000 N. A few grippers from IPR have a significantly heavier weight
Schunk pdh_inc IPR festo sommerautomac afag when compared to the other manufacturers. In this manufacturer’s cat-
alog, these grippers are described as a rugged design series (IPW) [7]
Fig. 10. Force vs weight by manufacturer. and the fact that emphasis is put on this aspect might also explain why
their C-factors is not as high as these of the other manufacturers. These
grippers probably can take a lot more abuse but this property is difficult
4.2. Force vs weight to establish quantitatively.
In Fig. 11 it is peculiar to notice that the grippers from Festo seem
Beside the C-factor, it is also interesting to compare the manufac- to more or less follow two horizontal lines: one at approx. 500 N and
turers by plotting the force vs weight of each grippers as illustrated in the second at about 150 N (dashed lines #3 and #4.) Grippers from IPR
Fig. 10. As expected, weight generally increases with the force and re- also lie on an approximately horizontal line, i.e. with an almost constant
ciprocally. While it might be difficult to read clear trends in this cloud of force at about 300 N. This indicates that the grippers have comparable
points, certain tendencies can be observed from the plots. For instance, force but very different weights and points to a standard value of force
in the case of Sommer Automatic, the plot shows 2 almost linear trend produced by grippers with very different strokes and thus, weight. It is
lines (lines #1-2 in Fig. 10). The first, steeper one, is on the left hand remarkable that PHD has a glut of grippers in the smaller range of values
side and goes until the highest force available from this manufacturer at which underlines that their grippers provide the most force compared
14,900 N (the gripper weights 14 kg). The second line, with a gentler to their weight in this smaller range. Finally, almost all grippers from
slope, ends at only 6,600 N with a weight of also 14 kg. These two Afag also coexist in this latter range since their maximal force values
trends are of course not perfect as shown in Fig. 11, a zoom around the does not go beyond 300 N.
range of values up to the median force (320 N) and the median weight

92
L. Birglen, T. Schlicht Robotics and Computer–Integrated Manufacturing 49 (2018) 88–97

curve 2
600,00

line 3
500,00

curve 1
400,00
force [N]

300,00

200,00

line 4
100,00

0,00
0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8 2
weight

Schunk pdh_inc IPR festo sommerautomac afag

Fig. 11. Force (600 N) vs weight (2 kg) by manufacturer.

350 PHD offer very similar distributions which show that they typically offer
line 5
300 products competing with each other over all their catalog range. On the
other hand, IPR and Festo have significantly different distributions. The
250
former seems to be specialized in both ends of the force spectrum with
stroke [mm]

200
little intermediate values, if any, while the latter offers on the contrary
150 a large range of grippers at these intermediate values. The force ranges
100
of these two companies look almost complementary except at the very
low values, where actually all manufacturers heavily compete. Schunk
50
and PHD have most of their grippers in a category where forces do not
0
0 10 20 30 40 50 60 70 80 exceed 50 N and only a few products after this threshold value. Con-
weight
Schunk pdh_inc IPR festo sommerautomac afag
versely, IPR does not have any gripper with less than 50 N of force
which seems to be an exception. The peak of the graph is in the cat-
egory of gripper with a force between 150 and 200 N. Afag was not
Fig. 12. Stroke vs weight by manufacturer.
considered in this diagram because their grippers actually do not pro-
duce forces above 300 N for the nominal input pressure. Therefore, its
4.3. Stroke vs weight distribution is only shown in Fig. 15 where pie charts are used to show
the force range for this manufacturer and compare it with the ones of
Fig. 12 shows another comparison between two specifications, Schunk and IPR. Please note that the slight difference between the per-
namely stroke and weight, for each company. When looking at all the centages in Fig. 14 and Fig. 15 is due to the fact that the maximal values
values, it can be seen that Schunk, PHD and Festo all have gripper se- considered for the pie charts is 3,000 N. For example, the percentage of
ries where these two numbers are almost linearly correlated (line #5). IPR gripper up to 200 N is 18% in Fig. 14 while it is 21% in Fig. 15 since
PHD’s and Festo’s go up to a 150 mm stroke whereas Schunk reaches the number of grippers considered is reduced from 51 to 43 (which still
the double of that value at 300 mm. Again, IPR grippers are signifi- represents 84% of their grippers).
cantly heavier, assumably due to their improved ruggedness, and thus,
their grippers stay in the lower right corner of the graph. 5. Performance metrics
A zoom to the part where most values lie is presented in Fig. 13
where the limits on each axes are set to a stroke of 20 mm and a The previously discussed comparison between grippers based on the
maximal weight of 2 kg. These values encompass the great majority C-factor gives some interesting insight about performances and can be
of grippers found in the manufacturers’ catalogs. This zoom shows an used to evaluate a product against typical values on the market. For
expected spread of the values for certain stroke sizes. The different val- instance, the C-factor of the median gripper from the collected data (i.e.
ues of weights for the same stroke indicate that the associated company using the median values of force, stroke, and weight) is:
offers this stroke in different series. PHD is shown to have a generally 320 𝑁 × 9.55 𝑚𝑚
smaller stroke with a comparable weight of the gripper as compared to 𝐶𝑚𝑒𝑑𝑖𝑎𝑛 = = 5.18 𝐽 ∕𝑘𝑔 (1)
0.59 𝑘𝑔
other manufacturers. Despite this, PHD still offer grippers which lie in
the upper left area of the graph which indicate a good ratio of stroke to While the C-factor of the virtual average gripper would be:
weight. It can also be seen that Schunk appears to offer several grippers 1020.44 𝑁 × 20.78 𝑚𝑚
𝐶𝑎𝑣𝑒𝑟𝑎𝑔𝑒 = = 6.21 𝐽 ∕𝑘𝑔 (2)
with large strokes (greater than 10 mm) and very light weights [20]. 3.41 𝑘𝑔
These values are consistent with these presented in Table 1 where the
4.4. Force percentage C-factors are computed as both the average and median of individual
C-factors of all grippers. One limitation of C-factors is that cycle times
Finally, Fig. 14 shows the distribution of the amount of grippers are not taken into account and two similar grippers with very different
sorted by force categories (each of 50 N) up to 1,000 N. Schunk and speed would not be distinguished using this metric. This issue could

93
L. Birglen, T. Schlicht Robotics and Computer–Integrated Manufacturing 49 (2018) 88–97

20

18

16

14

12
stroke [mm]

10

0
0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8 2
weight

Schunk pdh_inc IPR festo sommerautomac afag

Fig. 13. Stroke (20 mm) vs weight (2 kg) by manufacturer.

20%
18%
Schunk
16%
14%
phd_inc
12% IPR-worldwide
10% festo
8% sommerautomac
6%
4%
2%
0%
50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
force [N]

Fig. 14. Percentage vs force by manufacturer.

Schunk - IPR-worldwide - afag -


(Force [N]; Amount; Percent) (Force [N]; Amount; Percent) (Force [N]; Amount; Percent)
2700; 1; 1% 2150; 1; 2% 3000; 0; 0% 3000; 0; 0%
2600; 2; 2% 3000; 0; 0% 1800; 1; 2% 2600; 1; 2% 50; 0; 0%
1800; 1; 1%
1750; 1; 2%
1650; 2; 2% 100; 3; 7% 300; 1; 8%
2200; 4; 1600; 1; 2%
1600; 1; 1% 5% 50; 13; 15% 1550; 1; 2%
1450; 1; 1%
1300; 1; 1%
150; 5; 12%
1200; 2; 2%
1050; 2; 5% 50; 4; 33%
1100; 2; 2% 250; 2; 17%

1050; 1; 1%
900; 1; 1% 100; 8; 10% 1000; 2; 5%
850; 1; 1%
800; 1; 1%
200; 0; 0%
750; 1; 1% 950; 2; 5%
700; 2; 2%
200; 9; 21%
650; 3; 4% 150; 10; 12%
850; 1; 2%
750; 1; 2% 150; 2; 17%
600; 1; 1%

550; 2; 2% 700; 4; 9%
500; 1; 1% 200; 4; 5%
450; 2; 2%
400; 6; 7% 350; 2;
250; 6; 7% 250; 1; 2% 100; 3; 25%
5% 300; 5; 12%

350; 2; 2%
300; 2; 2%

Fig. 15. Distribution of force by 50 N steps for Schunk, IPR, and AFAG.

however be solved very easily by dividing the original C-factor by the which would measure the output power-to-weight ratio (notice the unit
closing time, i.e.: is W/kg) of a gripper. As an example, the median and average C⋆ are:
⋆ 5.18 𝐽 ∕𝑘𝑔
𝐶𝑚𝑒𝑑𝑖𝑎𝑛 = = 39.8 𝑊 ∕𝑘𝑔 (4)
0.13 𝑠
𝐶 − factor [ ]
𝐶⋆ = 𝑊 ∕kg (3) ⋆ 6.21 𝐽 ∕𝑘𝑔
closing time 𝐶𝑎𝑣𝑒𝑟𝑎𝑔𝑒 = = 27 𝑊 ∕𝑘𝑔. (5)
0.23 𝑠
94
L. Birglen, T. Schlicht Robotics and Computer–Integrated Manufacturing 49 (2018) 88–97

It should be pointed out that the cycle time is the sum of both the closing these larger robots also grew and catalogs now offer products with very
and opening times of the gripper and that these times could be slightly large strokes and forces.
different. However, the difference between average and median closing Machine tending for instance with drilling, milling or turning opera-
vs opening times in the data is small (approx. 5% at most) and can thus tions is also an area where gripping capabilities are important. For these
be neglected. tasks however, the chosen gripper must be versatile enough to be adapt-
Using the C-factor and the previously listed ratios as performance able to the many pieces that can be machined which also requires larger
metrics to compare grippers is one of many possible choice. Specifically, strokes and possibly, forces. According to our data, the median gripper
as with any other mechanical devices, the efficiency of the grippers is force is 320 N which is quite small and assuming a safety factor of 2
an important aspects. The latter being defined as the ratio between the (as recommended in many catalogs), the maximal weight that can be se-
output and input powers of the system, it is however difficult to estimate curely seized is limited. As previously stated, the nominal pressure con-
its value based on the data from the specifications only. Indeed, if the sidered here when compiling the data was 6 bars. However, it must be
output power is easy to estimate knowing the force produced by the noted that the working range of air pressure that the grippers require to
gripper and a rough idea of the velocity (stroke divided by closing or function varies considerably. The minimal operating pressure that was
opening time depending on the case), to compute the input force on found out in catalois a mere 0.4 bar while most grippers usually require
the gripper cylinder at a known pressure, the area of this cylinder is at least 4 bars. On the other hand, the maximal air pressure tolerated by
necessary. Sadly, this data is usually not listed. A common practice and the grippers is typically more limited and most of the specifications list a
a solution to this issue can be to use an estimate for the friction force2 value of 8 bars as the upper limit (the greatest value found was 10 bars.)
but unless this estimation results from an experimental identification Similarly, operating temperatures are quite comparable in all the data
(which is impossible when considering hundreds of grippers as in this with a range of typically 0°C to 80°C. A few manufacturers offer models
paper), the numbers obtained with this method are highly subjective that can work under higher temperatures but they are uncommon.
and cannot be used to reliably compare grippers. The demand to grasp many different workpieces clearly influences
Another metric of great interest in the industry is the estimated To- the gripper market and the continuous penetration of robotics in many
tal Cost of Ownership (TCO) [21]. This performance index is defined industrial sectors and also more and more in office/domestic environ-
as the sum of the initial purchase cost plus the yearly associated costs ments has forced manufacturers to develop new products. A particular
(maintenance, etc.) multiplied by the years of service. While this metric aspect often emerging now concerns versatility: grippers are expected to
is obviously very relevant as it clearly spells out the overall cost of using securely handle pieces of various shapes, weights, materials, etc. To in-
one gripper, estimating its fair value is quite difficult and requires addi- crease the flexibility of grippers, many technological avenues are possi-
tional information than what is listed in the specifications. For instance, ble from offering more evolved mechatronic architectures with complex
estimating the years of service as well as the maintenance costs is not kinematics as well as sensing apparatus and control to, more simply, in-
an information provided by the grippers’ manufacturers as they depend creasing the strokes of the grippers. These solutions however conflict
on the customer and the application (mean time before failure is even with the requirement of a fast grip and release of the object to handle.
rarely clearly stated). The development of robotic hands, often mimicking the human hand,
Performance indices such as the C-factor allow for an unbiased com- is one of the possible future evolution of robotic grippers. If precursor
parison of the actual performance of grippers but on the other hand, designs such as the Utah/MIT and Stanford/JPL Hands [22,23] have
it does not and cannot paint a full picture of product selection. Price been known for decades, they had little commercial success. Theoreti-
is of course a major concern for the industry, especially when a large cal works about these hands on topics such as mechanical synthesis, e.g.
number of grippers are required to equip a new production line. Fur- [24], have been of interest also for decades but only recently commer-
thermore, other aspects which are far more difficult to quantify may cially available products have been released. This recent development is
also be of major interest such as reliability (cf. the discussion about the mainly due to the constant progress in miniaturization and affordability
TCO), reputation, quality of customer services to name but a few. In the of both actuation and sensing system such as with the Schunk Anthro-
end and since the authors wished to present a metric that used impar- pomorphic Hand (SVH) [25] or the Shadow Robotics hand [26]. In re-
tial and publicly accessible data, the best option found was to rely on cent years, an intermediate solution between industrial robotic grippers
the C-factor and its slightly modified version discussed above. But it is and multi degrees-of-freedom hand has also gained a lot of attention,
worth noting that there are limits to the interpretation one can draw namely adaptive/underactuated grasping [27]. These graspers combine
from using solely these numbers. a single actuator to a more complex kinematic structure which is de-
signed to mechanically envelop the object to be grasped without any
sensing or control (hence, adaptive). This single actuator drives mul-
6. Discussion tiple output joints simultaneously (hence, underactuated). Commercial
products based on this principle now exist, most noticeably from Bar-
Based on the data compiled and presented in this paper and, again, rett Technology, Robotiq, and RightHand Robotics. All these adaptive
assuming that the most common grippers sold by the manufacturers cor- graspers are however driven by electrical energy and indeed, the market
respond to specifications where one can find the most products, it seems share of electrical grippers is obviously growing as well since electrical
that lightweight grippers with a small stroke and limited force are very drives and motors are becoming smaller, lighter, and more powerful.
popular. The reason for this is conjectured to be that the automation Electrical grippers have clear advantages over pneumatic grippers: they
of handling tasks often takes place in modern mass production which are more convenient to supply with energy, the closing force and finger
typically deals with smaller size parts of limited inherent values. If the position are easier to control, and they also are typically quieter. In con-
seized workpiece grows, its weight and monetary value typically also trast, the force of a pneumatic gripper cannot be controlled on the fly as
tends to increase as well and then, manual operations can still be cost easily and, normally, they do not allow for stopping at intermediate lo-
effective. However, this tendency is receding in the industry in recent cations between the fully open and fully closed positions which prevents
years with the introduction of ever larger robotic systems. Serial robots prepositioning the fingers close to the expected part size. It should be
can now be found with a rated payload above 1,000 kg, e.g. the FANUC nuanced however that pneumatic energy might not disappear soon and
M-2000iA and Kuka KR-1000 Titan series. The grippers combined with innovative solutions using air pressure to drive adaptive grippers exist,
both commercially, e.g. the Versaball from Empire Robotics, the Mul-
tiChoiceGripper from Festo, the grippers of Soft Robotics Inc, or from
2
see for instance: http://www.hydraulicspneumatics.com/blog/how-calculate academia, e.g. [28,29].
- hydraulic- cylinder- efficiency.

95
L. Birglen, T. Schlicht Robotics and Computer–Integrated Manufacturing 49 (2018) 88–97

A point of interest for future grippers also highlighting the require- Appendix A. List of grippers used
ment for flexibility lies in the increasing number of collaborative robots
sold around the world. These robots, as their name indicates, work in
SCHUNK PSH 32 FESTO MPG801N
close proximity and sometimes in direct contact with humans. There are
DPG-plus 100-1 PSH 42 DHPS 10 MPG802N
many reasons for this, from the avoidance of having to secure the im- DPG-plus 125-1 PSH 52 DHPS 16 MPG803N
mediate environment of the robots (which is costly) to public/societal DPG-plus 160-1 RH 801 KP DHPS 20 MPG804N
pressure for robotic labor not to completely replace humans’. End-of- DPG-plus 200-1 RH 806 KP DHPS 25 MPG806N
arm tooling for collaborative robots is still in its infancy and if a recent DPG-plus 40 RH 9010 DHPS 35 MPG808N
DPG-plus 50-1 RH 901K-St10 DHPS 6 MPG810N
norm for collaborative robots has been released [30], there are no guide-
DPG-plus 64-1 RH 905 HGP 16 MPG812N
lines currently available for these tools to ensure safety. Nevertheless, DPG-plus 80-1 RH 907 HGP 25
from a technical perspective, if a human and robot collaborate for han- JGP 100 RH 910K HGPC 12 IPR
dling tasks there are implications for the robotic grippers. Indeed, there JGP 125 RH 918 HGPC 16 CHP-302-30
JGP 160-1 RH 925 HGPC 20 CHP-302-60
are major differences between a human and a mechanical conveyor or
JGP 40 RH 940 HGPD 16 CHP-303-40
an automated transport system providing a part to the robot. Beside the JGP 50-1 SPG HGPD 20 CHP-303-80
obvious requirement for the safety of the human operator that could JGP 64-1 HGPD 25 CHP-304-100
limit the grasping force and speed, the location of the human cannot be JGP 80 PHD HGPD 35 CHP-304-50
precisely known and its dexterity is also inferior to the one of a machine. KGG 100-40 190/191 Style 1 – 19065 HGPD 40 CHP-306-120
KGG 140-60 190/191 Style 1 – 19075 HGPD 50 CHP-306-60
Therefore, when a human wants to hand over a workpiece to a robot,
KGG 220 190/191 Style 1 – 19085 HGPD 63 IPW-120
either the robot gripper needs to have a large stroke to accommodate lo- KGG 280 190/191 Style 1 – 19095 HGPD 80 IPW-160
cation uncertainties or use a sensing system to adaptively plan the grasp, KGG 60-20 190/191 Style 2 – 19066 HGPL 14 IPW-200
e.g. a camera system. A combination of both solution is probably desir- KGG 70-24 190/191 Style 2 – 19067 HGPL 14 IRP-13-ISO-31.5
KGG 80-30 190/191 Style 2 – 19076 HGPL 14 IRP-14-ISO-40
able to ensure robustness. Furthermore, collaborative robots typically
LGP 10 190/191 Style 2 – 19077 HGPL 14 IRP-15-ISO-50
have a small payload since safety dictates that they cannot apply large LGP 16 190/191 Style 2 – 19086 HGPL 25 IRP-16-ISO-50
forces to their environment. The combination of both large stroke and LGP 20 190/191 Style 2 – 19087 HGPL 25 IRP-17-ISO-80
low weight appears to be uncommon if not missing in the current state LGP 25 190/191 Style 2 – 19096 HGPL 25 IRP-18-ISO-100
of the market. Adding to that further important issues such as flexibility LGP 40 190/191 Style 2 – 19097 HGPL 25 IRP-19/20-ISO-125
LGP 8 5300 – 15362 HGPL 40 IRP-19/40-ISO-125
and affordability only accentuate this shortcoming.
MPG 10 5300 – 15372 HGPL 40 IRP-20-ISO-160
MPG 12 5300 – 15382 HGPL 40 IRP-21/37-ISO-250
MPG 16 5300 – 15392 HGPL 40 IRP-21/64-ISO-250
MPG 20 GRA – 10 HGPL 40 LPG-160-P
MPG 25 GRA – 16 HGPL 63 RP-05
7. Conclusion
MPG 32 GRA – 20 HGPL 63 RP-10
MPG 40 GRA – 6 HGPL 63 RP-100
As a result of the increasing automation of handling tasks, the mar- MPG 50 GRDx31 HGPP 10 RP-11
ket for the associated end-of-arm tooling is now an essential part of MPG 64 GRDx41 HGPP 12 RP-12
the robotic ecosystem. Impactive grippers are one of the most common MPG 80 GRDx51 HGPP 16 RP-15
PFH 150 GRDx52 HGPP 20 RP-17
product of this market and many different designs exist. In this review, PFH 200 GRDx61 HGPP 25 RP-18
the authors focused on pneumatically driven, parallel grippers with two PFH 250 GRDx62 HGPP 32 RP-19
jaws and compiled the specifications of hundreds of products as listed in PFH 30 GRF – 19 × 4.5 HGPT 16 RP-25
the catalogs of selected manufacturers. After a statistical analysis of all PFH 300 GRF – 25 × 6.5 HGPT 20 RP-35
PFH 40 GRF – 28 × 10 HGPT 25 RP-40
the data thereby obtained, it appeared that if the number of grippers on
PFH 50 GRF – 28 × 6 HGPT 35 RP-43
the market is large, most of them share similar characteristics such as PGB 100 GRF – 32 × 13 HGPT 40 RP-50
a small stroke and limited force. Efficiency as measured by the C-factor PGB 125 GRF – 32 × 8 HGPT 50 RPG-120-P
on the other hand is quite distinct from one manufacturer to the other. PGB 64 GRK – 35 × 12 HGPT 63 RPG-80-P
The C-factor of the grippers studied here range from 0.36 to 28.57, a PGB 80 GRK – 35 × 6.5 HGPT 80 RPL-1
PGF 100 GRK – 46 × 16 RPL-2
difference of almost two orders of magnitude. An important limitation PGF 125 GRK – 46 × 8 SOMMER AUTOM. RPL-3
of the C-Factor though is that it does not take into account the cycle PGF 50 GRK – 58 × 10.5 GP100-B-99 RPL-4
time of the gripper. This issue can be very simply solved by dividing the PGF 64 GRK – 58 × 20 GP12-B RPW-250
C-Factor by the latter and thereby quantifying the power-to-mass ratio PGF 80 GRK – 75 × 12.5 GP19-B RPW-375-1
PGM 120 GRK – 75 × 26 GP30-B RPW-375-2
of the gripper, defined here as the C⋆ -Factor. The large difference be-
PGM 140 GRL 14 × 13 GP403N-C RPW-500-1
tween average and median values of many specifications also highlights PGM 29 GRL 14 × 7 GP404N-C RPW-500-2
the significant tilt toward the lower end of the spectrum value. This PGM 38 GRL 16 × 20 GP404S-C RPW-625-1
is where the glut of the products coexist and almost all manufacturers PGM 50 GRL 16 × 26 GP406N-C RPW-625-2
compete. Finally, a comparison between these available products and PGM 60 GRRx2-x-63 × 150 GP406S-C RPW-750
PGM 82 GRRx2-x-63 × 200 GP408N-C
the requirements of emerging robotic systems also points to the need PGM 96 GRRx2-x-63 × 250 GP408S-C AFAG
for the development of new products with greater versatility and lower PGN-plus 100-1 GRRx2-x-63 × 350 GP410N-C GMK 20/P
weight. PGN-plus 125-1 GRS – 27 × 4.5 GP410S-C GMK 20/P-01
PGN-plus 160-1 GRS – 27 × 7 GP412N-C GMK 32/P
PGN-plus 200-1 GRS – 28 × 10 GP412S-C GMQ 12/P
PGN-plus 240-1 GRS – 28 × 6 GP416N-C GMQ 20/P
PGN-plus 300-1 GRS – 32 × 13 GP416S-C GMQ 20/P-01
Acknowledgments PGN-plus 380-1 GRS – 32 × 8 GP420N-C GMQ 32/P
PGN-plus 40 GRS – 50 × 128 GP420S-C PG12
PGN-plus 50 GRS – 50 × 19 GP430N-C PG16
This research did not receive any specific grant from funding agen-
PGN-plus 64-1 GRS – 63 × 32 GP430S-C PG20
cies in the public, commercial, or not-for-profit sectors. PGN-plus 80-1 GRS – 63 × 44 GP45-B UG 20 NN
PSH 32 GP75-B UG 25 NN

96
L. Birglen, T. Schlicht Robotics and Computer–Integrated Manufacturing 49 (2018) 88–97

References [17] Sommer Automatic Inc, Catalogs (one per model), 2015, (http://sommerautomatic.
com/web/catalogue/en/products/standardkomponenten/greifer/pneumatisch/
[1] A. Wolf, R. Steinmann, H. Schunk, Grippers in Motion: The Fascination of Automated 2- backen- parallelgreifer.html) [Online; accessed October 06, 2015].
Handling Tasks, Springer, Berlin Heidelberg, 2006. [18] Schunk GmbH, Gripping Moduls - Part 1–5, 2015, (http://www.ca.schunk.com/
[2] G. Monkman, S. Hesse, R. Steinmann, H. Schunk, Robot Grippers, Wiley, 2007. schunk/schunk_websites/service/catalogue_order_check.html) [Online; accessed
[3] G. Lundström, Industrial robot grippers, Ind. Rob. Int. J. 1 (2) (1974) 72–82. October 05, 2015].
[4] F.Y. Chen, Gripping mechanisms for industrial robots, Mech. Mach. Theory 17 (5) [19] J. Stokes, MSFC-STD-512A Man/System Requirements for Weightless Environments,
(1982) 299–311. Technical Report, NASA-MSFC, 1976.
[5] N.P. Belfiore, E. Pennestri, An atlas of linkage-type robotic grippers, Mech. Mach. [20] C. Parlitz, Hardware for Industrial Gripping at SCHUNK GmbH & Co. KG, Springer
Theory 32 (7) (1997) 811–833. London, London, 2013, pp. 363–384.
[6] C. Xiong, H. Ding, Y. Xiong, Fundamentals of Robotic Grasping and Fixturing, Series [21] A. Dietrich, Electric actuators vs. pneumatic cylinders: a comparison based on total
on manufacturing systems and technology, World Scientific, 2007. cost of ownership, Tolomatic Inc. Technical Report (9900-4044-00) (2015) 1–10.
[7] P.O. Hugo, Industrial Grippers: State-of-the-Art and Main Design Characteristics, [22] S. Jacobsen, J. Wood, D. Knutti, K. Biggers, The Utah/MIT dexterous hand: work in
Springer London, London, pp. 107–131. progress, Int. J. Rob. Res. 3 (4) (1984) 21–50.
[8] K. Tai, A.-R. El-Sayed, M. Shahriari, M. Biglarbegian, S. Mahmud, State of the [23] J.K. Salisbury, J.J. Craig, Articulated hands: force control and kinematic issues, Int.
art robotic grippers and applications, Robotics 5 (2) (2016) 11, doi:10.3390/ J. Rob. Res. 1 (1) (1982) 4–17.
robotics5020011. [24] C. Tischler, A. Samuel, K. Hunt, Kinematic chains for robot handsi. orderly number-
[9] G. Carbone, Grasping in Robotics, Springer London, London, 2013. synthesis, Mech. Mach. Theory 30 (8) (1995) 1193–1215. http://dx.doi.org/10.
[10] P. Chua, T. Ilschner, D. Caldwell, Robotic manipulation of food products a review, 1016/0094- 114X(95)00043- X.
Ind. Rob. Int. J. 30 (4) (2003) 345–354, doi:10.1108/01439910310479612. [25] Schunk GmbH, Servo-Electric 5-Finger Hand, 2017, (https://de.schunk.com/de_en/
[11] I. Chelpanov, S. Kolpashnikov, Problems with the mechanics of industrial robot gripping-systems/series/svh/) [Online; accessed March 22, 2017].
grippers, Mech. Mach. Theory 18 (4) (1983) 295–299. http://dx.doi.org/10.1016/ [26] M. Reichel, Transformation of shadow dextrous hand and shadow finger test unit
0094- 114X(83)90122- 2. from prototype to product for intelligent manipulation and grasping, in: Intelligent
[12] M. Ceccarelli, Notes for a History of Grasping Devices, Springer London, London, Manipulation and Grasping, Inter Conf., Genova, Italy, 70, 2004.
2013, pp. 3–16. [27] L. Birglen, T. Laliberté, C. Gosselin, Underactuated Robotic Hands, Springer, New-Y-
[13] Afag, Gripper Moduls, 2014, (http://www.afag.com/fileadmin/user_upload/afag/ ork, 2008.
Handlingmodule_Katalog/05_PG- UG- GM_2014- 01.pdf) [Online; accessed October [28] L. Birglen, Enhancing versatility and safety of industrial grippers with adaptive
07, 2015]. robotic fingers, in: 2015 IEEE/RSJ International Conference on Intelligent Robots
[14] Festo, Catalogs (one per model), 2015, (https://www.festo.com/cat/en-ca_ca/ and Systems (IROS), 2015, pp. 2911–2916, doi:10.1109/IROS.2015.7353778.
products_010804) [Online; accessed October 19, 2015]. [29] K.C. Galloway, K.P. Becker, B. Phillips, J. Kirby, S. Licht, D. Tchernov, R.J. Wood,
[15] IPR, Intelligente Peripherien fur Roboter GmbH, Catalogs (one per model), 2015, D.F. Gruber, Soft robotic grippers for biological sampling on deep reefs, Soft Rob. 3
(http://www.iprworldwide.com/en/products/standard-components/grippers/ (1) (2016) 23–33.
parallel- grippers/2- jaw- parallel- grippers.html) [Online; accessed October 19, [30] Robots and robotic devices – Collaborative robots, Standard ISO 15066:2016, Inter-
2015]. national Organization for Standardization, Geneva, CH, 2016.
[16] PHD Inc., Catalogs (one per model), 2015, (https://www.phdinc.com/products/
category/?product=grippers) [Online; accessed October 05, 2015].

97

Вам также может понравиться