Communications of the ACM

tendrils, swapping hard for soft versions, the FAU team found there was an optimum combination for thrust: with both top and bottom being made of a material with floppiness similar to that of a mouse pad, being pulsed a little less than once a second.

In addition to the challenges roboti-
cists face with materials and power de-
livery, Josh Bongard, associate professor
at the University of Vermont, says con-
trol presents further problems. “Exploit-
ing the capabilities of soft robots is a
very non-intuitive thing for human engi-
neers to do. The mathematics that we’ve
developed over decades for designing
and controlling traditional robots made
up of rigid links simply doesn’t apply to
[these] systems. In short: it’s hard to de-
sign and control moving blobs.”
In contrast to the inverse kinematics
and closed-loop control that dominate
fixed-function robots, Bongard propos-
es harnessing evolutionary program-
ming coupled with machine learning
to develop novel control methods for
producing movement that take into ac-
count how plastic materials bend and
compress under force.

In these simulations, the design
starts with a basic shape made from
blocks with different levels of stiff-
ness and mobility. Evolutionary
algorithms gradually change the
properties of different blocks until
the robot is able to move. The algo-
rithms tune their response to the
way materials flex under strain in
different positions using machine-
learning algorithms such as neural
networks. Sometimes, the simulat-
ed machines use the equivalent of
body fat to help leverage the effects
of whatever type of motion the robot
adopts. Bongard says, “Our evolution-
ary algorithms often find very non-in-
tuitive designs, such as one that has a
hump on its back and uses it to throw
its weight forward to make movement
more efficient.”
The work tends to agree with re-
sults such as those from FAU: softer
robots perform better under water.
Those simulated on land tended to re-
quire stiffer structures. Bongard now
plans to take the work to physical ro-
bots in a collaboration with Kramer-
Bottiglio and her team on a project
backed by the National Science Foun-
dation. The simulations will help
create configurations for the robotic
skins and the underlying shapes the
skins are wrapped around.

Although soft robots have a long way to go before becoming autonomous enough to deliver on the promise of lighter, more functional machines, research and development is gradually bringing together the types of control and materials science that will make them work well on Earth and beyond.

Further Reading

Rus, D., and Tolley, M. T.

Design, Fabrication and Control of Soft Robots

Nature 521, Number 7553, pp467-475

(2015)

Booth, J. W, Shah, D., Case, J.C., White, E.L., Yuen, M.C., Cyr-Choiniere, O., and Kramer-Bottiglio, R.

OmniSkins: Robotic Skins That Turn Inanimate Objects into Multifunctional Robots,

Science Robotics, Vol 3, Issue 22, eaat1853 (2018)

Frame, J., Lopez, N., Curet, O., and Engeberg, E.D.

Thrust Force Characterization of Free-Swimming Soft Robotic Jellyfish,

Bioinspiration & Biomimetics, Volume 13, Number 6 (2018)

Corucci, F., Cheney, N., Giorgio-Serchi, F., Bongard, J., and Laschi, C.

Evolving Soft Locomotion in Aquatic and Terrestrial Environments: Effects of Material Properties and Environmental Transitions,

Soft Robotics, 5( 4): 475-495 (2018)

Chris Edwards is a Surrey, U.K.-based writer who reports on electronics, IT, and synthetic biology.

to electronic logic gates to implement components such as oscillators convey the peroxide to reaction chambers. Each chamber contains a platinum catalyst, which splits the molecule into water and oxygen. The resulting gas drives pneumatic actuators that move the limbs, although they can do little more than twitch.

Soft robots may overcome the power problem by operating in environments where gravity is not as big a problem as it is on land. The microgravity of space may be one obvious habitat for them, but robots made from elastomers and pumps already find the going much easier under water.

Former Florida Atlantic University (FAU) student Jennifer Frame chose the jellyfish as the biological model for her thesis. Named JenniFish, the robot uses stubby plastic tentacles harnessed to hydraulic pumps powered by a battery to mimic the pulsing action of the invertebrate’s body as it moves. Able to swim untethered in the ocean, it is flexible enough to squeeze through narrow orifices. In experiments, the robot would collide with the edges of a hole, then generate enough thrust to push its appendages to its side and squeeze through.

Another situation in which soft robots could perform well is in a different fluid environment: inside the human body. In work performed in Europe before moving to Stanford University in 2016, Stanford postdoctoral fellow Giada Gerboni (who works in surgical robotics) developed a soft endoscopic camera for surgery. Now Gerboni is focusing on soft robots that can enter the body, move around, and perform microsurgery operations without direct intervention from human surgeons. She describes it as a very flexible needle that can steer around parts of the body and into organs with minimal disruption.

In further work by the group led by FAU associate professor Erik Engeberg, the JenniFish has helped demonstrate how material choices form a key part of design for liquid environments. The algorithm needed to make the robot swim in different directions is relatively simple: it takes advantage of the way the polymer limbs are shaped to bend in on themselves when flexed. By changing the plastics used for the top and bottom surface of the JenniFish