introduction.tex 7.68 KB
Newer Older
Tran Huy Vu's avatar
Tran Huy Vu committed
1 2

U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
4 5 6 7 8 9 10 11 12
%Embedded sensors, deployed on small form-factor devices, have transformed
%our ability to pervasively observe the state of the physical world,
%including the monitoring of human activities, ambient context and machinery
%state.  As obvious examples, (a) inertial or physiological sensors in
%wearable devices (such as smartwatches and smart-necklaces) have been used
%to monitor an individual's eating behavior~\cite{XXX}, smoking~\cite{XXX} or
%stress levels~\cite{XXX}; (b) vibration, audio or light sensors have been
%used on innovative IoT platforms to detect operating conditions and
%anomalies in factories, city neighborhoods and critical infrastructure.

U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
\emph{Energy} remains perhaps the greatest challenge in the pervasive
15 16
deployment of sensing systems, such as wearable devices for human activities (e.g. 
eating behaviour~\cite{thomaz2015}, smoking~\cite{parate2014}, or stress
Tran Huy Vu's avatar
Tran Huy Vu committed
levels~\cite{ertin2011}) or embedded devices used for environmental sensing~(e.g.,~\cite{campbell2014}).  In particular,
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
18 19 20 21 22 23 24 25 26 27 28
sensors such as accelerometers or gyroscopes simply consume too much energy
to operate continuously without either a dedicated power source or a large
battery.  However, using battery power introduces two distinct
disadvantages: (i) frequent recharging may simply be cumbersome or
impractical--e.g., wearable-based health monitoring may be much more
palatable if an embedded device may be worn for months without needing to be
taken off and recharged; (ii) perhaps not as widely appreciated,
high-density storage batteries can \emph{leak}, causing corrosion and other
serious hazards, especially when the sensors are deployed in volume and out
of sight (e.g., embedded inside factory equipment in industrial IoT

U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
To overcome this disadvantages, many solutions using renewable energy
31 32
harvesting capabilities have been proposed--such as ambient light~\cite{hande2007}, temperature gradients~\cite{campbell2014b} 
and kinetic energy~\cite{ryokai2014}.  Each such technique is innovative, but has its own limitations--e.g.,
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
33 34
ambient light cannot be used for sensors mounted in poorly lit or occluded
locations (e.g., in a dark warehouse or on occluded body locations).

U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
36 37 38 39 40 41
In this paper, we demonstrate the practical feasibility of using WiFi
packets from a commodity WiFi AP (access point) to power a wearable sensor
device.  The widespread coverage of WiFi deployments makes this an
intriguing, orthogonal alternative for pervasive energy harvesting. 
Wireless charging, itself, is not novel, but current solutions require
either close proximity (3-5cm) to the transmitting power source (e.g., the
Qi~\cite{liu2015} standard used by modern high-end phones), or can only charge
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
ultra-low power passive RFID tags~\cite{yeager2008} at longer ranges.  More
44 45
recently, PoWiFi~\cite{talla2015powering} demonstrated the use of WiFi, using multiple channels simultaneously, to power an ultra-low
power wearable (with a temperature or camera sensor), with low duty cycles.  

47 48
\emph{Our key scientific contributions are two-fold}: we show (a) how to increase
the harvested WiFi power (via directional WiFi transmissions) to much higher levels (O(100$\mu$W)), even on a single channel, on an embedded
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
49 50
device, at a much greater distance ($\sim$3-4meters from the transmitter)
than had been previously possible.  This enables many more use cases, in
industrial IoT, smart homes,  etc.; and (b) that, with novel triggered-sensing
Tran Huy Vu's avatar
Tran Huy Vu committed
techniques (that further extend the energy-driven intermittent sensing paradigm articulated in~\cite{hester2017b}), our solution can be used to collect useful gesture-related data from a batteryless, wearable sensing device, using an embedded accelerometer.
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
53 54 55 56

Our solution, called \names, uses beam-formed transmissions, by a
multi-antenna AP, of WiFi ``power packets'' (transmissions performed
explicitly to transfer RF energy) to deliver bursts of directed WiFi energy
Tran Huy Vu's avatar
Tran Huy Vu committed
57 58 59
to a client device.  To point the beam towards the client, \name utilizes AoA
(angle-of-arrival) estimation techniques~\cite{xiong2013arraytrack}. 
These AP-side techniques are paired with novel energy-conserving features on the wearable
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
60 61 62 63 64 65 66 67
device, which activates its communication and sensing components
intelligently and selectively, to help capture only key events.  While the
core ideas were articulated in our preliminary work~\cite{tran2017}, this
paper presents a detailed design, implementation and empirical validation of

\noindent \textbf{Key Contributions:} To our knowledge, we are the first to
design and empirically demonstrate a working prototype (called \names) that
uses RF transmissions, on a single channel, from a \emph{realistically-distant} WiFi AP to power a
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
batteryless, wrist-mounted wearable sensor device (which is collecting and
transmitting significant accelerometer data generated by regular human movement).  To achieve this goal, we make
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
the following key contributions:
72 73 74


U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
75 76 77 78
\item \emph{Use of Beamformed WiFi Transmissions for Power Delivery:}
Through empirical experiments, it is clear that the harvested power, from a
conventional omni-directionally transmitting WiFi AP, is too low for
practical use: around $1-3\mu$W at distances of 3-4 meter.  To tackle this
problem, we extend prior work to beamform the WiFi transmissions
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
80 81 82 83 84 85 86 87 88 89 90 91
to spatially concentrate the transmitted power.  Via experimental studies,
we show that even with real-world errors in direction estimation and
beamforming, directional transmissions with an 8-antenna array result in
dramatically higher levels of harvested energy ($\sim 700-800 \mu$W), even
when the embedded device is 3-4 meters distant from the WiFi AP.

\item \emph{Design \& Implementation of an Intermittently-Triggered Wearable
Sensor:} We built a wrist-worn, \names-compatible wearable device, which
utilizes  WiFi harvesting to power an inertial sensor used in various
gesture-tracking applications.  Such a wearable device, worn by a mobile
user, gives rise to two challenges: (i) the WiFi AP must be able to track
the wearable's changing location, without requiring constant active
92 93
transmissions from the wearable, and (ii) the peak power  overhead of the wearable
system, including the accelerometer and the RF frontend, is over 40 mW-- while low, this is much
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
94 95 96 97 98 99 100
higher than the harvested power of 700-800$\mu$W to permit continuous
sensing.  To tackle both these challenges, the wearable employs a simple
kinetic energy harvester-based trigger to first detect \emph{significant
motion} of the wearable device.  Such significant motion triggers both (i)
the transmission of ``ping'' packets by the wearable, which allows the AP to
determine the wearable's new AoA, and (ii) the activation of the
accelerometer sensor, during the likely occurrence of meaningful gestures. 
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
In addition, the wearable utilizes a super capacitor to store the harvested
U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
102 103 104 105 106 107
RF energy, and smoothen out transient fluctuations in power supply and

\item \emph{Demonstrate Feasibility of \name in a Real Office Environment:}
We utilize a series of quasi-controlled studies, in an office environment,
to demonstrate the overall effectiveness of the proposed \name approach. 
Tran Huy Vu's avatar
Tran Huy Vu committed
Our microbenchmarks shows a huge potential use of the \name architecture for battery-less device.  Though our user study shows that the application of \name to wearable devices is much more challenging, we do notice one case that the system provides sufficient energy for the wearable to work. In other cases, we do notice that our AP can transmit tens of \micro W to the device.
109 110

U-RAJESH-SIS\rajesh's avatar
U-RAJESH-SIS\rajesh committed
111 112 113

We believe that our work lays the foundational principles of a practical
WiFi-based energy harvesting mechanism for future embedded sensing devices
114 115
that have application beyond just static sensors (e.g., in commercial or industrial sites) to
include wearable devices.
116 117 118 119