Signal Processing - September 2017 - 10

MIT

FIGURE 2. A prototype device that MIT researchers claim could lead to external robot sensors that
can be inexpensively printed in bulk quantities.

According to Subramanian Sunda-
ram, an MIT graduate student in electri-
cal engineering and computer science,
who led the project, organisms in nature
are internally connected with a dense
sensorimotor network that allows them
to sense vast amounts of data and, more
importantly, process and respond to the
most relevant signals. "So we want to
replicate the 'sense-process-respond'
pipeline we see in a simple creature-
the golden tortoise beetle," he says.
The goldbug senses external mechani-
cal disturbances and changes the color/
transmission of its reflective shell.
"Replicating functions like these are ex-
tremely relevant for robots; a majority of
contemporary robots operate without a
high density of sensors on their external
surfaces," Sundaram explains. "In that
sense, they are blind."
Collaborating with Sundaram on
the project is his advisor, Wojciech Ma-
tusik, an associate professor of electrical
engineering and computer science, and
Marc Baldo, a professor of electrical en-
gineering and computer science and di-
rector of MIT's Research Laboratory of
Electronics. Other participants include
Pitchaya Sitthi-Amorn, a former post-
doctoral researcher in Matusik's lab;
Ziwen Jiang, an undergraduate electri-
cal engineering and computer science
student; and David Kim, a technical as-
sistant in Matusik's Computational Fab-
rication Group.
Printable electronics-flexible circuitry
deposited onto a plastic substrate-has
10

been a major research area for decades.
The MIT project, however, marks the
first demonstration of printed electron-
ics and printed substrates combined.
"We use a custom-built multimate-
rial 3-D printer to print the entire
composite," Sundaram says. "We show
devices with up to six materials print-
ed together."
The researchers' prototype device
(Figure 2) is approximately T-shaped,
featuring a short, wide base and an elon-
gated crossbar with a strip of silver run-
ning across its length. The base includes
a pair of printed transistors and a circle
of semiconducting polymer-dubbed a
pixel by the researchers-that changes
its color when the crossbar stretch-
es, modifying the silver strip's elec-
trical resistance.
Signal processing plays a key role
in the project. "Overall, we work on
printing low-level signal processing
elements," Sundaram says. The main
signal processing element in the printed
composite is a single stage common
source amplifier with a diode con-
nected load. "We use it for threshold-
ing as well," Sundaram says. "However,
at the high level, we believe that signal
selection is an equally important prob-
lem when a vast number of sensors
are present."
Mechanical strain is measured by
printed strain sensors connected in a re-
sistive ladder configuration. "The signal
is amplified using our single stage am-
plifier, and then the output is displayed
IEEE SIGNAL PROCESSING MAGAZINE

September 2017

using an electrochromic pixel," Sunda-
ram says.
According to a paper published by
Sundaram and his coresearchers, the
monolithic integration of sensing, pro-
cessing and response mechanisms allows
transducing signals across mechanical,
electrical and optical domains using low-
power organic processors and sensors
that can be powered by 1.5 V. Control-
ling multidomain properties with uni-
form resolution and without any external
processing should enable advances in
biologically inspired autonomous multi-
functional systems with increased local
signal processing efficiencies and levels
of self-sufficiency currently only seen
in nature.
"We believe that arranging multiple
electrical devices in freeform 3-D is still a
challenge," Sundaram says. "Constructing
the supporting polymer composite along
with the electronic devices is useful for
application in robotics and, more broadly,
flexible and nonplanar electronics."

Into the air
Signal processing lies at the heart of a
novel new approach to robotic unmanned
aerial vehicles (UAVs) developed by a
pair of doctoral students at the National
University of Singapore's Unmanned
System Research Group.
U-Lion is a reconfigurable hybrid
UAV offering both vertical takeoff and
landing (VTOL) and cruise flying capa-
bilities (Figure 3). VTOL allows U-Lion
to take off and land in small, tight spaces
lacking a conventional runway. Cruise
flying permits the aircraft to perform
long-range and duration missions. "With
the ability to fully open and retract the
wings, its flying performance is opti-
mized for both flying modes," says
coresearcher Kangli Wang, who worked
on the project with fellow doctoral stu-
dent Yijie Ke and advisor Ben M. Chen.
"We have utilized advanced signal pro-
cessing, modeling as well as control
techniques to enable U-Lion with fully
autonomous flying capabilities."
Hybrid UAV technology is a hot
research topic both academically and
commercially. Several companies, such
as Google (Project Wing), have devoted
significant resources to developing hybrid

Table of Contents for the Digital Edition of Signal Processing - September 2017