UC Santa Cruz
UC Santa Cruz Previously Published Works
Title
An integrated design and fabrication strategy for entirely soft, autonomous robots.
Permalink
https://escholarship.org/uc/item/1182x4zm
Journal
Nature, 536(7617)
ISSN
0028-0836
Authors
Wehner, Michael
Truby, Ryan L
Fitzgerald, Daniel J
et al.
Publication Date
2016-08-01
DOI
10.1038/nature19100
Supplemental Material
https://escholarship.org/uc/item/1182x4zm#supplemental
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
25 AUGUST 2016 | VOL 536 | NATURE | 451
LETTER
doi:10.1038/nature19100
An integrated design and fabrication strategy for
entirely soft, autonomous robots
Michael Wehner
1,2
*, Ryan L. Truby
1,2
*, Daniel J. Fitzgerald
1,2
, Bobak Mosadegh
3,4
, George M. Whitesides
2,5
,
Jennifer A. Lewis
1,2
& Robert J. Wood
1,2
Soft robots possess many attributes that are difficult, if not
impossible, to achieve with conventional robots composed of rigid
materials
1,2
. Yet, despite recent advances, soft robots must still be
tethered to hard robotic control systems and power sources
3–10
.
New strategies for creating completely soft robots, including soft
analogues of these crucial components, are needed to realize their
full potential. Here we report the untethered operation of a robot
composed solely of soft materials. The robot is controlled with
microfluidic logic
11
that autonomously regulates fluid flow and,
hence, catalytic decomposition of an on-board monopropellant fuel
supply. Gas generated from the fuel decomposition inflates fluidic
networks downstream of the reaction sites, resulting in actuation
12
.
The body and microfluidic logic of the robot are fabricated using
moulding and soft lithography, respectively, and the pneumatic
actuator networks, on-board fuel reservoirs and catalytic reaction
chambers needed for movement are patterned within the body via a
multi-material, embedded 3D printing technique
13,14
. The fluidic and
elastomeric architectures required for function span several orders
of magnitude from the microscale to the macroscale. Our integrated
design and rapid fabrication approach enables the programmable
assembly of multiple materials within this architecture, laying the
foundation for completely soft, autonomous robots.
Soft robotics is a nascent field that aims to provide safer, more robust
robots that interact with humans and adapt to natural environments
better than do their rigid counterparts. Unlike conventional robots
composed of rigid materials, soft robots based on hydrogels
15,16
,
electroactive polymers
17
, granular media
18
and elastomers
5,19
exhibit elas-
tic moduli ranging from 10 kPa to 1 GPa (ref. 1), are physically resiliant
7,20
and have the ability to passively adapt to their environment
1,2,19
.
Moulded and laminated elastomers with embedded pneumatic net-
works are widely used materials in soft robotics
1,21,22
. Actuation of
these elastomeric composites occurs when interconnected channels
that make up the pneumatic network are inflated with incompressible
fluids or gases supplied via tethered pressure sources
1
. Robotic end
effectors with bioinspired
10
and rapid
6
actuation, deployable crawlers
3,7
and swimmers
8
with complex body motions, and robust jumpers
9,23
have been developed on the basis of this design strategy. However, in
each case, these robots are either tethered to or carry rigid systems for
power and control, yielding hybrid soft–rigid systems
4,7–9
.
Creating a new class of fully soft, autonomous robots
24
is a grand
challenge, because it requires soft analogues of the control and power
hardware currently used. Recently, monopropellant fuels have been
suggested as a promising fuel source for pneumatically actuated soft
robots
4,12
. Their rapid decomposition into gas upon exposure to a cata-
lyst offers a strategy for powering soft robotic systems that obviates the
need for batteries or external power sources. Here, we report a method
for creating a completely soft, pneumatic robot—the ‘octobot’—with
eight arms that are powered by monopropellant decomposition.
1
John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, USA.
2
Wyss Institute for Biologically Inspired Engineering, Harvard University,
Cambridge, Massachusetts 02138, USA.
3
Dalio Institute of Cardiovascular Imaging, Weill Cornell Medicine and New York Presbyterian Hospital, New York, New York 10021, USA.
4
Department of
Radiology, Weill Cornell Medicine, New York, New York 10021, USA.
5
Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, USA.
* These authors contributed equally to this work.
Soft
controller
Fuel
reservoirs
Fuel inlets
Actuators
Vent
orifices
Pt reaction
chambers
a
b
c
d
g
e
f
Figure 1 | Fully soft, autonomous robot assembly. a, b, A microfluidic
soft controller is pre-fabricated (a) and loaded into a mould (b).
c–e, Matrix materials are poured into the mould (c) and fugitive and
catalytic inks are EMB3D printed (d, e). Scale bar in e, 10 mm. f, After
matrix curing, the printed fugitive inks auto-evacuate, yielding open
channels. Scale bar, 2 mm. g, The octobot is removed from the mould and
inverted to reveal a fully soft, autonomous robot that is controlled via the
embedded microfluidic soft controller and powered by monopropellant
decomposition. Scale bar, 10 mm. Fluorescent dyes have been added in
e and g to assist in visualization of internal features.
© 2016 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
452 | NATURE | VOL 536 | 25 AUGUST 2016
LETTERRESEARCH
To accomplish this, we use microfluidic logic
11
as a soft controller and
a multi-material, embedded 3D (EMB3D) printing method to fabricate
pneumatic networks within a moulded, elastomeric robot body. Our
hybrid assembly approach allows one to seamlessly integrate soft lithog-
raphy, moulding and 3D printing to rapidly and programmably fabri-
cate a range of materials and functional elements in the form factors
that are required for autonomous, untethered operation of a soft robot.
To fabricate an octobot, we first micro-mould
11,25
the soft controller
that houses the microfluidic logic necessary for controlling fuel decom-
position (Fig. 1a). The soft controller, which is protected temporarily
with a polyimide mask, is placed into a mould that is partially filled by
hyperelastic layers, which are needed for actuation (Fig. 1b). Matrix
materials are then poured into the mould (Fig. 1c) and the remain-
ing soft robot features are EMB3D printed into the moulded matrix
(Fig. 1d, e, Supplementary Video 1). After the matrix materials are
cross-linked, the aqueous inks ‘auto-evacuate’ at elevated temperature
10 mm s
–1
0.5 mm s
–1
3 786
5
4
29
a
b
c
de
f
Speed (mm s
–1
)
Trace width (μm)
10
3
10
2
10
1
Catalytic ink
Fugitive ink
Shear stress (Pa)
10
5
10
4
10
3
10
2
10
1
10
0
10
–1
10
–1
10
0
10
1
10
–1
10
0
10
1
10
2
10
3
Storage modulus, G′ (Pa)
Fugitive ink
Catalytic ink
Body matrix
Fuel reservoir matrix
1
Figure 2 | Multi-material, EMB3D printing. a, The octobot features
include (1) the body matrix, (2) the fuel reservoir matrix, (3) printed
fuel reservoir traces, (4) fugitive plugs in the soft controller, (5) printed
platinum reaction chambers, (6) printed pneumatic networks, (7) printed
vent orifices, (8) printed actuators and (9) moulded hyperelastic actuator
matrix. All printed features are composed of the fugitive ink except the
printed platinum reaction chambers (5), which are patterned using the
catalytic ink. b, The storage modulus, G′ , of the fugitive ink, catalytic ink,
body matrix and fuel reservoir matrix as a function of shear stress. The
plateau moduli of the inks are an order of magnitude higher than those
of the matrix materials. c, Trace widths of the fugitive and catalytic inks
printed at 450 kPa and 345 kPa, respectively, decrease with print speed.
Error bars indicate the standard deviation for n = 3 measurements.
d, Optical images of channel cross-sections printed at speeds of 0.5 mm s
−1
and 10 mm s
−1
, which demonstrate that trace dimensions can be changed
on-the-fly. Scale bars, 100 µ m. e, f, Reaction chambers printed with the
catalytic inks contain a platinum-laden plug, as shown in a cross-section
(e; scale bar, 500 µ m) and a scanning electron micrograph (f; scale bar, 25 µ m).
T
c
Fuel
reservoirs
Upstream
check valves
Soft
controller
Oscillator
Fuel
inlets
Pinch
valves
Downstream
check valves
Outlets
Reaction
chambers
a
b
Reaction
chamber
From
reservoir
From
oscillator
Actuator
1
Time
Pressure
2
3
4
5
TimeTime
Flow
rate
Voltage
Vent
orifices
Oscillator
Fuel
reservoirs
Reaction
chambers
Actuators
Pt
Pt
H
2
O
2
H
2
O
2
Time
Pressure
Figure 3 | Octobot control logic. Discrete sides are shown in red and blue
for clarity. a, A system of check valves and switch valves within the soft
controller regulates fluid flow into and through the system. The letters
of ‘VERITAS’, each with a height of 500 µ m, are patterned into the soft
controller as an indication of scale. b, A schematic (top) and qualitative
electrical analogy (bottom) of the octobot system; check valves, fuel
tanks, oscillator, reaction chambers, actuators and vent orifices are akin to
diodes, supply capacitors, electrical oscillator, amplifiers, capacitors and
pull-down resistors, respectively. c, Conceptual curves show key variables
as a function of time. (1) Nominal pressure drives fuel through system
at a decreasing rate. (2) Pinch valves in the oscillator convert upstream
flow into alternating flow between red and blue channels. Flow rate and
switching frequency are functions of upstream pressure and downstream
impedance. (3) When upstream pressure is too low, oscillation is not
possible, so both sides flow at a reduced rate. (4) Catalyst decomposes fuel,
yielding pressurized gas, which flows downstream to the actuators and
the vent orifices concurrently. (5) Actuators deform (θ, actuator tip angle)
as the pressure changes. Vents must be sufficiently small to allow full
actuation, yet sufficiently large to allow timely venting.
© 2016 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
25 AUGUST 2016 | VOL 536 | NATURE | 453
LETTER RESEARCH
as water evaporates and diffuses through the matrix, leaving behind
an open network of channels that are interfaced with the soft control-
ler (Fig. 1f). Octobot fabrication is completed upon removal of excess
matrix material (Fig. 1g). A more detailed description of this multi-step
assembly process is provided in Extended Data Fig. 1.
By combining micro-moulding with EMB3D printing, we rapidly
patterned the required mesofluidic networks by extruding a fugitive
ink, which is subsequently removed via auto-evacuation, through a
fine nozzle that is embedded within the uncured elastomer matrix. To
self-heal crevices that form within the ‘body’ matrix as the nozzle is
translated during the printing process, we created a new elastomeric
material containing fumed silica nanoparticles that exhibits thixotropic
behaviour
26
(Extended Data Fig. 2a). When completely restructured
or at rest, this matrix behaves like a Herschel–Bulkley fluid; that is, it
exhibits both shear-thinning behaviour (Extended Data Fig. 2b) and a
shear yield stress (Extended Data Fig. 2c). These properties ensure that
the extruded inks remain in place within the matrix
13,14
. However, upon
yielding, the body matrix readily flows (Extended Data Fig. 2c) into any
crevices formed. The body matrix restructures with time, ultimately
recovering its original viscoelasticity (Extended Data Fig. 3), which
ensures that EMB3D printing can be repeated later in the same matrix
region. We also created a ‘fuel reservoir’ elastomeric matrix, into which
fuel reservoir channels are printed. Both the body and fuel reservoir
matrices are cross-linked within the mould after printing is completed.
To create the fuel reservoirs, catalytic reaction chambers, actuator
networks and vent orifices, two hydrogel-based inks (fugitive and
catalytic) are EMB3D printed within the moulded matrix materials
(Fig. 2a). These printed features are interfaced with each other as well
as with the soft controller through the use of ‘fugitive plugs’ introduced
at the inlets of the controller before filling the mould with the matrix
materials. The fugitive ink is composed of an aqueous, poly(ethylene
oxide)-b-poly(propylene oxide)-b-poly(ethylene oxide) triblock copol-
ymer (Pluronic F127) gel
13,27
. The catalytic ink contains platinum
particles (Supplementary Video 2) suspended in a mixture of Pluronic
F127-diacrylate (F127-DA) and poly(ethylene glycol) diacrylate
(PEG-DA) that is photo-cross-linked after printing. The rheological
properties of both inks are specifically tailored for EMB3D printing
13,14
(Fig. 2b, Extended Data Fig. 4). The printed features produced from both
inks can be changed ‘on-the-fly’ by varying the print speed (Fig. 2c).
Typically, this fugitive ink must be removed or evacuated after print-
ing to yield open channels
13,27
. However, we find that fugitive ink
composed of pure Pluronic F127 can be auto-evacuated by heating
the printed features within the cross-linked, silicone-based matrices
at 90 °C (ref. 28; Fig. 2d, Extended Data Fig. 5). As water evaporation
State 2
Red flow
State 1
Blue flow
T
Inflate
12 21
d
T = 0
a
Top
Remove and invert
h
Side
iii
iii
bc
h = 1,500 μm
h = 1,250 μm
h = 1,000 μm
h = 750 μm
80
60
40
20
0
Pressure (bar)
0 0.2 0.4 0.6
Displacement
angle, T (°)
ii
iii
Mould
Figure 4 | Octobot actuation. a, Two-bladder actuator design in which
traces (i) are printed in contact with the hyperelastic layer (ii) inside the
body matrix material (iii) and differences in modulus result in bending
upon inflation. The thickness, h, of the hyperelastic layer is modified
to change the characteristics of the actuator. In this example, the body
matrix material (iii) has a height of 800µ m. b, Top, the actuator tip angle,
θ, changes upon inflation. Scale bar, 10 mm. Bottom, mean displacement
angle, θ, taken from three representative actuators during five inflation
cycles as a function of inflation pressure, for varying hyperelastic layer
heights, h (in µ m). Error bars, denoted by the shaded regions, indicate
the 95% confidence interval. c, The oscillator of the soft controller
causes an octobot to alternate between blue and red actuation states.
The monopropellant fuel is dyed to show states. Scale bar, 5 mm. d, Stills
from top-down (top; Supplementary Video 5) and face-on (bottom;
Supplementary Video 6) operation videos show an octobot autonomously
alternating between blue (‘1’) and red (‘2’) actuation states. Scale bars, 10 mm.
© 2016 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
454 | NATURE | VOL 536 | 25 AUGUST 2016
LETTERRESEARCH
ensues, the triblock copolymer species either form a thin coating at the
matrix-open channel interface or may partially diffuse into the matrix.
The fugitive plugs within the inlets of the soft controller also undergo
this auto-evacuation process, facilitating connectivity between the
microfluidic logic and all printed mesofluidic components (Extended
Data Fig. 6). By contrast, the catalytic ink is cross-linked in place after
printing, yielding a platinum-laden plug within the matrix (Fig. 2e).
To achieve the desired autonomous function, we incorporated a soft,
microfluidic controller within the octobot (Fig. 3a). The control system
is roughly divided into four sections: upstream (liquid fuel storage),
oscillator (liquid fuel regulation), reaction chamber (decomposition
into pressurized gas) and downstream (gas distribution for actuation
and venting). Upstream, 0.5 ml of fuel is infused via a syringe pump
into each of two fuel reservoirs printed into the hyperelastic matrix.
Upstream check valves in the soft controller prevent fuel from flow
-
ing back out the fuel inlets. The fuel reservoirs expand elastically to
a pressure of approximately 50 kPa, forcing fuel into the oscillator.
The oscillator includes a system of pinch and check valves based on
prior designs
11
, which convert pressurized fuel inflow into alternat-
ing fuel outflow. With one channel temporarily occluded, fuel from
the other channel flows from the outlets of the soft controller into
the platinum-laden reaction chambers, where it rapidly decomposes.
The resulting pressurized gas, which is prevented from returning to
the soft controller via downstream check valves, flows into one of the
downstream mesofluidic networks consisting of four actuators and one
orifice. The supplied pressure deflects the actuators and exhausts to
atmosphere through the vent orifice. Therefore, for robust actuation
and timely venting, a balance must be reached between supply gas flow,
actuation pressure and exhaust rate. These subcomponents operate on
the basis of the interaction and timing of the local pressures, which is
conceptually similar to an electrical oscillator (Fig. 3b). Upon successful
venting, the fuel flow into one reaction chamber stops and flow to the
other begins, initiating a similar sequence in the other downstream
catalytic chamber and actuator network (Fig. 3c).
To provide an on-board power source, we use 50wt% aqueous hydro-
gen peroxide as the fuel, owing to its high energy density (1.44 kJ g
−1
as
compared to 0.1–0.2 kJ g
−1
for batteries) and its benign decomposition
by-products. As the fuel decomposes in the presence of the platinum cat-
alyst, the following reaction occurs: 2H
2
O
2
(l)→ 2H
2
O(l,g) + O
2
(g).
This reaction results in volumetric expansion by a factor of approxi-
mately 240 (at ambient pressure)
12
. At our operating pressure of 50 kPa,
an expansion to 160 times the original volume is expected. Although
higher fuel concentrations would provide increased expansion and
energy density, concentrations above 50wt% drastically increase the
decomposition temperature, resulting in combustion within the printed
catalytic reaction chambers (Supplementary Videos 3 and 4). Because
this monopropellant liquid fuel can be handled in small volumes and
decomposes at the point of use, we can use microfluidic logic to directly
handle the fuel, eliminating the need for external valves
7
to control gas
at high pressure and flow rate.
The geometry of the microfluidic soft controller is designed to oper-
ate at a fuel flow rate of about 40 µ l min
−1
, thereby yielding pressurized
gas at a rate of about 6.4 ml min
−1
(ref. 11). Under these operating con-
ditions, the theoretical run time of 12.5 min could be achieved using
a system with a fuel capacity of 1 ml. The actuators, which consist of
printed bladders in contact with a lower-modulus, hyperelastic elasto-
mer layer (Fig. 4a), are designed to inflate asymmetrically to generate
angular displacement. Their maximum working pressure and dis-
placement are tuned on the basis of the thickness of the hyperelastic
layer (Fig. 4b, Extended Data Fig. 7). If this layer is too thin, then it
ruptures prematurely. However, the working pressure increases with
thickness. As a compromise, we selected a layer thickness of 1,000 µ m
because it affords consistent performance at the lowest working pres-
sure. In parallel with the actuators, we tailored the diameter of the vent
orifices by modulating print speed. Orifices roughly 75 µ m in width
allowed proper actuator displacement with timely subsequent venting.
The ability to rapidly pattern and adjust the geometry of these features
on-the-fly via EMB3D printing allowed us to iterate through more
than 30 designs and nearly 300 octobots to converge on an appropriate
system-level architecture.
Through this iterative process, we created octobots with embedded
components that work together in concert to alternate between the
red and blue actuation states shown in Fig. 4c. The resulting octobots
operated autonomously (Fig. 4d, Extended Data Fig. 8, Supplementary
Videos 5 and 6), cycling between actuation states for four to eight min-
utes. Although this is less than the predicted theoretical run time, the
soft controller alternates actuation states as expected. We believe that
downstream impedances arising from decomposition–actuation–
venting cycles, as well as the decreasing flow rate of fuel into the soft
controller with time, are responsible for the departure from theo-
retical performance
11
. These issues can be addressed by integrating
more sophisticated microfluidic circuits, such as universal logic gates,
or components with ‘gain’ that enable advanced control schemes
(see Methods for an extended discussion).
We have demonstrated the untethered operation of a robot com-
posed solely of soft materials. The coupling of monopropellant fuels
and microfluidic logic allowed us to power, control and realize auton-
omous operation of these pneumatically actuated systems. Through
our hybrid assembly approach, we both constructed the robot body
and embedded the necessary components for fuel storage, catalytic
decomposition and actuation to enable system-level function in a rapid
manner. The octobot is a minimal system designed to demonstrate
our integrated design and fabrication strategy, which may serve as a
foundation for a new generation of completely soft, autonomous robots.
Online Content Methods, along with any additional Extended Data display items and
Source Data, are available in the online version of the paper; references unique to
these sections appear only in the online paper.
Received 29 March; accepted 7 July 2016.
1. Rus, D. & Tolley, M. T. Design, fabrication and control of soft robots. Nature 521,
467–475 (2015).
2. Wang, L. & Iida, F. Deformation in soft-matter robotics: a categorization and
quantitative characterization. IEEE Robot. Autom. Mag. 22, 125–139 (2015).
3. Shepherd, R. F. et al. Multigait soft robot. Proc. Natl Acad. Sci. USA 108,
20400–20403 (2011).
4. Onal, C. D., Chen, X., Whitesides, G. M. & Rus, D. Soft mobile robots with
on-board chemical pressure generation. In 15th International Symposium on
Robotics Research (ISRR 2011) 1–16 (2011).
5. Lin, H.-T., Leisk, G. G. & Trimmer, B. GoQBot: a caterpillar-inspired soft-bodied
rolling robot. Bioinspir. Biomim. 6, 026007 (2011).
6. Mosadegh, B. et al. Pneumatic networks for soft robotics that actuate rapidly.
Adv. Funct. Mater. 24, 2163–2170 (2014).
7. Tolley, M. T. et al. A resilient, untethered soft robot. Soft Robot. 1, 213–223
(2014).
8. Marchese, A. D., Onal, C. D. & Rus, D. Autonomous soft robotic sh capable of
escape maneuvers using uidic elastomer actuators. Soft Robot. 1, 75–87
(2014).
9. Bartlett, N. W. et al. A 3D-printed, functionally graded soft robot powered by
combustion. Science 349, 161–165 (2015).
10. Peele, B. N., Wallin, T. J., Zhao, H. & Shepherd, R. F. 3D printing antagonistic
systems of articial muscle using projection stereolithography. Bioinspir.
Biomim. 10, 055003 (2015).
11. Mosadegh, B. et al. Integrated elastomeric components for autonomous
regulation of sequential and oscillatory ow switching in microuidic devices.
Nat. Phys. 6, 433–437 (2010).
12. Wehner, M. et al. Pneumatic energy sources for autonomous and wearable soft
robotics. Soft Robot. 1, 263–274 (2014).
13. Wu, W., Deconinck, A. & Lewis, J. A. Omnidirectional printing of
3D microvascular networks. Adv. Mater. 23, H178–H183 (2011).
14. Muth, J. T. et al. Embedded 3D printing of strain sensors within highly
stretchable elastomers. Adv. Mater. 26, 6307–6312 (2014).
15. Palleau, E., Morales, D., Dickey, M. D. & Velev, O. D. Reversible patterning and
actuation of hydrogels by electrically assisted ionoprinting. Nat. Commun. 4,
2257 (2013).
16. Ionov, L. Biomimetic hydrogel-based actuating systems. Adv. Funct. Mater. 23,
4555–4570 (2013).
17. Anderson, I. A., Gisby, T. A., McKay, T. G., O’Brien, B. M. & Calius, E. P.
Multi-functional dielectric elastomer articial muscles for soft and smart
machines. J. Appl. Phys. 112, 041101 (2012).
18. Brown, E. et al. Universal robotic gripper based on the jamming of granular
material. Proc. Natl Acad. Sci. USA 107, 18809–18814 (2010).
© 2016 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.