CONTRIBUTED
PAPER
Energy Harvesting From
Human and Machine Motion
for Wireless Electronic Devices
Practical miniature devices are becoming available for harnessing kinetic energy as
a substitute for batteries in medical, and many other, low power applications.
By Paul D. Mitcheson, Member IEEE, Eric M. Yeatman, Senior Member IEEE,
G. Kondala Rao,
Student Member IEEE, Andrew S. Holmes, Member IEEE,and
Tim C. Green,
Senior Member IEEE
ABSTRACT
|
Energy harvesting generators are attractive as
inexhaustible replacements for batteries in low-power wireless
electronic devices and have received increasing research
interest in recent years. Ambient motion is one of the main
sources of energy for harvesting, and a wide range of motion-
powered energy harvesters have been proposed or demon-
strated, particularly at the microscale. This paper reviews the
principles and state-of-art in motion-driven miniature energy
harvesters and discusses trends, suitable applications, and
possible future developments.
KEYWORDS
|
Energy scavenging; micropower generator;
micropower supply; vibration-to-electric energy conversion
I. INTRODUCTION
Wireless power supplies have the same advantages for
electronic devices as do wireless communications: they
allow portability, and even for non-portable applications
they reduce installation costs by eliminating wiring. The
latter feature is particularly important w here sources of
wired power are not locally available. For this reason,
improved wireless power supplies are increasingly sought
after as electronic systems proliferate. Batteries in their
various forms have so far been the primary solution;
however, they frequently dominate the size, and sometimes
the cost, of the devices in question and introduce an
unwanted maintenance burden of replacement or rechar-
ging. Alternative power sources that overcome these
limitations are thus highly desirable. The possible ap-
proaches to this challenge are t o use local energy supplies
with higher capacity, to deliver power wirelessly from an
active source introduced fo r this purpose, or to extract
power from ambient sources in some way.
Improving the energy density (and other features such
as cost, n umber of charging cycles, and power density) of
batteries has been, and continues to be, a major research
field. Battery storage densities have increased substantially
in the last decades, with lithium-ion batteries in particular
now having typical capacities of about 160 W h/kg [1], [2],
i.e., about 1 kJ/cc. Hydrocarbon fuels, however, offer
energy densities more than an order o f magnitude above
even the theoretical potential of lithium -ion batteries, for
example, 8 kW h/kg for methanol [1]. Of course the use of
fuel requires a conversion mechanism (which will also
impact on the system volume). Small-scale converters in-
vestigated to date include miniature turbine engines [3],
[4] and a micro Stirling-engine [ 5], but the most re-
searched and most promising to date are micro fuel cells
[6], [7]. Fuel-based power sources naturally do not over-
cometherechargingrequirement of batteries, but rather
replace i t with a (less frequent) refuelling requirement.
Capacitors are ano ther possible finite energy store; how-
ever, although some advantage may be obtained by their
much higher power densities and cycle lifetimes compared
to batteries, their energy densities remain relatively low,
with theoretical limits around 10 W h/kg [8]. Conversely,
radioactive materials provide a possible power source with
low power density but long lifeti mes, and m iniature power
Manuscript received January 8, 2008; revised June 4, 2008. This work was
supported by the Engineering and Physical Sciences Research Council and the
European Commission.
The authors are with the Department of Electrical and Electronic Engineering,
Imperial College London, London, U.K. (e-mail; paul.mitcheson@imperial.ac.uk).
Digital Object Identifier: 10.1109/JPROC.2008.927494
Vol. 96, No. 9, September 2008 | Proceedings of the IEEE 14570018-9219/$25. 00
2008 IEEE
supplies based on these have also been demonstrated [9].
More comprehensive reviews of portable power sources
are presented by Roundy et al. in [2] and [10] and Fukunda
and Menz in [11].
Power can be actively delivered continuously, p eriod-
ically, or on demand, using far-field electromagnetic radia-
tion [12] or near-field coupling [13], [14]. Such power
supplies require the use of infrastructure in addition to the
powered device i tself, and of course the supplying source
must in turn be supplied with power. However, this can be
ausefulsolutionwhenthedevicetobepoweredisin-
accessible (e.g., implanted sensors) or when power is only
needed when information is extracted (e.g., passive RF
identification (RFID) tags).
Extracting power from ambient sources is generally
known as energ y harvesting, or energy scavenging. This
approach has recently attracted a great deal of interest
within both the academic community and industry, as a
potential inexhaustible source for low-power devices.
Generally energy harvesting suffers from low, variable,
and unpredictable levels of available power. However, the
large reductions in power consumption achieved in
electronics, along w ith the increasing numbers o f mo bile
and other autonomous devices, are continuously increasing
the attractiveness of harvesting techniques. Consequently,
the amount of research in the field, and the number of
publications appearing, have risen greatly. Special issues
on energy harvesting have appeared, for example, in IEEE
Pervasive Computing Magazine [15] and the Intelligent
Materials S ystems and Structures Journal [16].
The sources of energy available for harvesting are
essentially of four forms: light, radio-frequency (RF) elec-
tromagnetic radiation, thermal gradients, and motion, in-
cluding fluid flow. All have received attention, in varying
degrees. Solar cells are the most mature and commercially
established energy-harvesting solution [17]–[19], and are
of course exploited across a wide range of size scales and
power levels. While cost is a key parameter for large-scale
photovoltaic generation, at the small scale of portable
electronic devices this is less of an issue, and light avail-
ability is instead the key limitation. A wide range of work
has also been presented on small-scale thermoelectric
generation [20]–[23], and successful applications include
the Seiko Thermic watch.
1
Temperature d ifferences tend
to be small over the miniature size scale associated with
most harvesting applications, which leads to poor thermo-
dynamic efficiency, but useful power levels can be cap-
tured from differences as little as a few degrees celsius.
Ambient RF has also received some attention [13], [24],
[25], although availability of significant power levels is
again an issue [26], and efficient extraction using devices
much smaller than the radiation wavelength is another key
challenge. As an adjunct to the four main sources for
harvesting, fuel-based generation using ambient fluids as
fuel, specifically human bodily fluids, has a lso been
reported [27].
The relative advantages and disadvantages of the dif-
ferent sources for energy harvesting have been discussed
thoroughly by various authors [2], [10], [28]–[31]; co n-
sequently, the arguments will not be repeated here in
detail. The general opinion from the literature is that while
each application should be evaluated individually with re-
gards to finding the best energy-harvesting method, kinetic
energy in the form of motion or vibration is generally the
most versatile and ubiquitous ambient energy source
available. The purpose of this paper is to review the prin-
ciples, achievements, future potential, and possible appli-
cations of motion-based energy harvesting.
II. APPLICATIONS FOR MOTION-BASED
ENERGY HARVESTING
A. Wireless Sensor Networks
Traditionally, health care has concentrated upon short-
term treatment rather than long-term monitoring and
prevention [32]. However, many chronically ill p atients
could have a significant increase in quality of life and life
expectancy if certain biological signs could be continually
monitored and controlled during their daily lives. Three
examples illustrate the potential of this approach:
continually monitoring blood pressure in patients with
hypertension can significantly increase medication co m-
pliance [33]; real-time processing of electrocardiograph
traces can be very effective at revealing early stages of
heart disease [34]; and closed-loop control of insulin
administration for diabetic p atients would significantly
reduce the risk of hypoglycemia [35]. Monitoring can also
allow better targeting of medicines, reducing costs and
unwanted side-effects. In order to achieve these benefits,
many types of body-mounted or implanted medical devices
are desired [36].
Implantable or wearable devices will only significantly
increase quality of life if they are unobtrusive to the patient
[37], [38] in terms of both use and maintenance. It is
especially important to eliminate maintenance for im-
plantable devices, for which replacement of the power
source in particular must be avoided [39]. While some
implanted sensors can be totally passive and used in con-
junction with active equipment when a measurement is
needed [40], and some active devices could be powered up
occasionally by wireless energy transfer, many require a
continuous source of electrical power [36]. Ideally, all im-
plantable medical devices would have a power-supply life-
time as long as the required operational lifetime, thus
keeping surgery, and cost, to a minimum. This vision of
unobtrusive, automated health care [41] using wearable
and implanted wireless medical devices is the main focus
of a new and fast-growing multidisciplinary research area,
that of the body sensor network (BSN) [42], [43]. In general,
1
http://www.seikowatches.com.
Mitcheson et al.:EnergyHarvestingFromHuman and Machine Motion
1458 Proceedings of the IEEE | Vol. 96, No. 9, September 2008
the tiny size of i nformation-processing and RF integrated
circuits means that batteries dominate the size of devices
that require long operating times [2], [35], [44], such as
BSN nodes. However, the continual evolution of solid-
state e lectronics, combined with new circuit design
techniques, has led to vast reductions in power consump-
tion,aswellassize,forcircuitsrequiredtoperformgiven
functions. This combination of low power requirements,
tight size constraints, and the need to eliminate mainte-
nance m akes BSN a particularly attractive application for
energy harvesting.
The BSN is a specific ins tance of a more gener al topic,
the wireless sensor network (WSN) [45], [46]. The general
wireless sensor network concept is that of deploying many
small, inconspicuous, self-contained sensor nodes, often
referred to as motes , into an environment to collect and
transmit information, and possibly provide localized actua-
tion. Other than medical applications, potential uses for
WSNs include structural monitorin g o f b uildings [47];
status monitoring of machinery; environmental monitor-
ing o f domestic environments to make them more
comfortable [48], [49]; military tracking [49]; security;
wearable computing; aircraft engine monitoring [50]; and
personal tracking and recovery systems [51]. As with BSNs,
many application areas will only be attractive for WSN use
if motes can be powered by an inexhaustible energy
source, such as harvested energy.
Fig. 1 shows a block diagram of the signal and proces-
sing elements of a wireless sensor mote capable of sending
the data to a remote location for processing. The minimum
power requirements of such a device can be estimated
using a mixture of currently available off-the-shelf tech-
nology, and devices which are the current state-of-the-art
in research. As an example, consider the following three
elements.
1) Sensor: The STLM20 tem perature sensor from ST
Micro [52] draws typically 12 Wquiescentpower
at 2.4 V supply voltage.
2) ADC: An ADC reported by Sauerbrey et al. [53] has
power dissipation below 1 W for 8 bit sampling
at 4 kS/s.
3) Transmitter: IMEC recently announced an IEEE
802.15.4a s tandard-compliant ultra-wide-band
transmitter [54] with a power consumption of
only 0.65 nJ per 16 chip burst operating at a low
duty cycle.
The required data rates for biomonitoring applications
tend to be quite low due to the relatively low rates of change
of the v ariables [44]. One o f the highest rates required is for
heartbeat monitoring, at around 100 samples/s. If this is
combined with a resolution of 10 bits, then the data rate is
1 kbps, which, if the transmitter power quoted can be
scaled to such low data rates, requires only 0.65 W. This
suggests a total power consumption for the sensor node of
10–20 W, or even 1–2 Worlessiftheothercomponents
are also duty cycled. There woul d be some extra overhead
for power-processing interface and timing circuitry, but it
is reasonable to estimate that the total device power
consumption could ultimately be reduced to a few W, at
least for this biosensor application. As discussed below,
this is within achievable levels for energy harvesters of
modest (below 1 cc) size, even when harvesting low-
frequency body motion. It should be noted that while the
power values quoted above are achievab le, cur rently
available wireless sensor nodes have substantially higher
levels of power consumption.
B. Other Applications
Limited battery life is a significant inconvenience for
most portable electronic devices, so target applications for
energy harvesting are primarily limited by the feasibility of
harvesting in each case. This feasibility depends main ly on
four factors: the typical power consumption of the device;
the usage pattern; the device size (and thus the acceptable
harvester size); and the motion to which the device is
subjected (for m otion harvesting specifically). For exam-
ple, laptop computers are poor candidates for harvesting:
although they are relatively large, they have high power
consumption (10–40 W), and their typical usage patterns
comprise long periods (tens of minutes to hours) of con-
tinuous use, with idle periods mostly spent in a low-motion
environment. Even if harvesting is used to supplement
rather than replace batteries, the added battery life is likely
to be marginal at best for most users.
Mobile telephones (cell phones) are a somewhat more
attractive target, as they tend to be carried on the body for
much of the time, thus experiencing regular motion while
only being used (other than in low-power monitoring
mode) for relatively short periods. Of course the relative
amounts of motion and usage are highly dependent on the
user. The power l evels during calls are typically a few
watts, and this is likely to reduce to some extent with
advances in the relevant technologies. However, space is
very much at a premium in handsets, and energy-
harvesting power densities reported to date for body
motion sources, as reviewed below, are well below the
levels at which this application becomes feasible. For other
Fig. 1. Basic wireless sensor arrangement.
Mitcheson et al.:EnergyHarvestingFromHuman and Machine Motion
Vol. 96, No. 9, September 2008 | Proceedings of the IEEE 1459
handheld devices, such as mp3 players and personal
organizers, the considerations are similar to those for
phones, with some differences in power requirements and
usage patterns.
Thus, wireless sensors would appear to be the primary
application area for motion harvesting, at least in the short
term. However, niche or unexpected applications are
likely to appear as well. One that has already been suc-
cessfully exploited is the harvesting of mechanical power
in a finger-actuated light switch to power a transmitter
circuit that relays the switching signal to a remote lighting
module [55].
III. MOTION-DRIVEN ENERGY
HARVESTERS: OPERATING PRINCIPLES
A. Introduction
Motion-driven microgenerators f all into two catego-
ries: those that utilize direct application of force and those
that make use of inertial forces acting on a proof mass. The
operating principle of a direct-force generator is shown in
Fig. 2. In this case, the driving force f
dr
ðtÞ acts on a proof
mass m supported on a suspension with spring constant k,
with a damping element present to provide a force fð_zÞ
opposing the motion. If the damper is implemented usin g a
suitable transduction mechanism, then in opposing the
motion, energy is converted from mechanical to electrical
form. There are limits of Z
l
on the displacement of the
mass, imposed by device size. Direct force generators must
make mechanical contact with two structures that m ove
relative to each other, and can thus apply a force on the
damper.
The operating principle of inertial microgenerators is
shown in Fig. 3. Again a proof mass is supported on a
suspension, and its inertia results in a relative displace-
ment zðtÞ whentheframe,withabsolutedisplacement
yðtÞ, experiences acceleration. The range of zðtÞ is again
Z
l
. Energy is converted when work is done against the
damping force fð_zÞ, which opposes the relative mo tion.
Inertial generators require only one point of a ttachment to
a moving structure, which gives much more flexibility in
mounting than direct-force devices and allows a greater
degree of miniaturization.
In order to generate power, the damper must be im-
plemented by a suitable electromechanical transducer.
This can be done using one of the methods described
below.
B. Transduction Methods
In conventional, macroscale engineering, electrical
generators are overwhelmingly based on electromagnetic
transduction. In small-scale energy harvesting, two main
additional techniques are added. Electrostatic transduc-
tion, which is both impractical and i nefficient for large
machines, becomes much more practical at small size
scales and is well suited to microelectromechanical
(MEMS) implem entation. Piezoelectric transduction is
generally impractical for rotating systems but is well suited
to the reciprocating nature of the motions typically used
for harvesting (e.g., vibration).
Rotating electromagnetic generators are in common
use from power levels of a few watts (brushless dc domestic
wind turbine systems) to several hundred megawatts
(synchronous machines in power plants). It is possible to
implement the damper of a microgenerator using the same
principle, i.e., that described by Faraday’s law of induction,
as illustrated in F ig. 4. A change of magnetic flux linkage
withacoilinducesavoltagevðtÞ in the coil, driving a
current iðtÞ in the circuit. The combined force fðtÞ on the
moving charges in the magnetic field acts to oppose the
relative motion, as described by Lenz’s law. The mechanical
work done against the opposing force is converted to heat in
the resistance of the circuit and to stored energy in the
magnetic field associated with the circuit inductance. Some
key practical issues for electromagnetic energy harvesters
are as follows: strong damping forces require rapid flux
changes, which are difficult to achieve in small geometries
Fig. 2. Generic model of direct-force generator.
Fig. 3. Generic model of inertial generator.
Mitcheson et al.:EnergyHarvestingFromHuman and Machine Motion
1460 Proceedings of the IEEE | Vol. 96, No. 9, September 2008
or at low frequency; the number of coil turns achievable in
aMEMSorothermicroscaledevicewillbelimited,
resulting in low output voltages; a nd integrati on of
permanent magnets, and ferromagnetic materials for the
flux path, is likely to be required.
In electrostatic generators, mechanical forces are
employed to do work against the attraction of oppositely
charged parts; in effect, such devices are mechanically
variable capacitors whose plates are separated by the
movement of the source. They have two fundamental
modes of operation: switched and continuous [56]. In the
switched type, the transducer and the circuitry is recon-
figured, through the operation of switches, at different
parts of the generation cycle. Switched transducers can
further be split into two main types: fixed charge and fixed
potential. The first is illustrated in F ig. 5(a). For a paral-
lel p late structure with a variable separation and constant
overlap (i.e., the horizontal component of _zðtÞ is zero)
and with a negligible fringing field, the field strength is
proportional to the (constant) charge, and thus the
energy density of the electric field is independent of plate
separation. As the electrode separation increases [by
doing mechanical work against the attractive force fðtÞ],
addit ional potential energy is stored in the increase d
volume of electric field. If instead the plates are moved
relative to each other laterally (i.e., the vertical compo-
nent of _zðtÞ is zero), mechanical work is done against the
fringing field. There is an increase in stored electrical
energy because the electric field strength increases with
the reduction in plate overlap, and the energy d ensity of
the field (proportional to the square of field strength)
increases faster than its volume decreases.
Constant voltage operation is illustrated in Fig. 5(b). If
the plate separation is increased with a fixed overlap, the
electric field strength falls, causing charge to be pushed off
the plates into an external circuit as a current iðtÞ.Ifthe
plates are moved with constant separation and changing
overlap, the field strength stays constant but current is
again forced to flow into the source because the volume of
the field decreases. In both cases, the mechanical work
done is converted into additional electrical potential
energy in the voltage source.
For both modes, since the charge equals the
capacitance times the potential ðQ ¼ CVÞ,andstored
energy is
1
=
2
CV
2
, the electrostatic force is found to be h alf
thevoltagesquaredtimestherateofchangeof
capacitance, i .e.,
F ¼
1
2
V
2
dC=dz (1)
formotioninthez-direction. Thus a constant force is
obtained for normal m otion in the constant charge case
Fig. 5. Principle of operation of the electrostatic transducer:
(a) constant charge and (b) constant voltage.
Fig. 4. Principle of operation of the electromagnetic transducer.
Mitcheson et al.:EnergyHarvestingFromHuman and Machine Motion
Vol. 96, No. 9, September 2008 | Proceedings of the IEEE 1461
![Fig. 12. Comparison of minimum Q factors with electromagnetic and piezoelectric cube devices of volume 0.1 cc (from [70]).](/figures/fig-12-comparison-of-minimum-q-factors-with-electromagnetic-3spy1a2b.webp)
![Fig. 13. Energy-harvesting shoe (from [77]).](/figures/fig-13-energy-harvesting-shoe-from-77-7j5yq2r4.webp)
![Fig. 19. In-plane generator by Roundy et al. (from [2] with kind permission of Springer Science and Business Media).](/figures/fig-19-in-plane-generator-by-roundy-et-al-from-2-with-kind-es2p9phe.webp)
