Energy Scavenging for Mobile and Wireless Electronics · laptop computers (a mature mobile...

18 PERVASIVEcomputing Published by the IEEE CS and IEEE ComSoc ■ 1536-1268/05/$20.00 © 2005 IEEE

E N E R G Y H A R V E S T I N G & C O N S E R V A T I O N

Energy Scavenging forMobile and WirelessElectronics

After Alessandro Volta invented thebattery in 1799, predating MichaelFaraday’s dynamo by 32 years, bat-teries provided the world’s first prac-tical electricity source until the

wiring of cities in the late 1800s relegated bat-teries to mobile applications. Despite vacuum-tube electronics’ weight and large associated bat-tery,1 people living in the early 1900s lugged such

enormous “portable” radios topicnics and other events off thepower grid. As electronics be-came smaller and required lesspower, batteries could growsmaller, enabling today’s wire-less and mobile applicationsexplosion. Although economicalbatteries are a prime agentbehind this expansion, they also

limit its penetration; ubiquitous computing’s dreamof wireless sensors everywhere is accompanied bythe nightmare of battery replacement and disposal.

Figure 1 depicts increases in the performance oflaptop computers (a mature mobile technology)on a logarithmic scale relative to a laptop from1990. As the graph indicates, battery energy isthe slowest trend in mobile computing. Althoughnew materials are revolutionizing the battery’sform factor, its energy density doesn’t scale expo-nentially; rather, battery capacity proceeds along

flattening S-curves. Overcoming this trendrequires moving to another energy source. Newfabrication technologies have recently resulted inmicro fuel cells aimed at recharging handheldswith power plants the size of a candy bar, andthey promise fuel cells on a chip for poweringwireless sensor nodes. Although research proto-types exist, laptop-sized plants (30–50 watt-hours) tend to be too big to directly power withmicrocells and too small for standard fuel cells,because their associated chemistry requires sig-nificant overhead in mass. More exotic emergingpower technologies exhibit characteristics thatforce them into niche applications. For example,radioactive decay can power batteries that lastfor decades, but they provide low current andinvolve complicated disposal. Furthermore,devices that burn fuel, such as microturbines andmicroengines, potentially pose issues with exhaust,heat, noise, thrust, or safety.

Ongoing power management developmentsenable battery-powered electronics to live longer.Such advances include dynamic optimization ofvoltage and clock rate, hybrid analog-digitaldesigns, and clever wake-up procedures that keepthe electronics mostly inactive. Exploiting renew-able energy resources in the device’s environment,however, offers a power source limited by thedevice’s physical survival rather than an adjunctenergy store. Energy harvesting’s true legacy dates

Energy harvesting has grown from long-established concepts into devicesfor powering ubiquitously deployed sensor networks and mobileelectronics. Systems can scavenge power from human activity or derivelimited energy from ambient heat, light, radio, or vibrations.

Joseph A. ParadisoMassachusetts Institute of Technology Media Laboratory

Thad StarnerGeorgia Institute of Technology,GVU Center

to the water wheel and windmill, andcredible approaches that scavenge energyfrom waste heat or vibration have beenaround for many decades. Nonetheless,the field has encountered renewed inter-est as low-power electronics, wirelessstandards, and miniaturization conspireto populate the world with sensor net-works and mobile devices. This articlepresents a whirlwind survey throughenergy harvesting, spanning historic andcurrent developments. For a more detailedtreatment emphasizing human-poweredsystems, see our recent book chapter onthe subject.2

Ergs from MaxwellWith copious radio transmitters scat-

tered throughout today’s urban land-scape, we might consider backgroundradio signals as a mobile-device powerreservoir. Electronic systems that harvestenergy from ambient-radiation sources,however, have extremely limited powerand generally require either a large col-lection area or close proximity to the radi-ating source. An RF power-scavenginganalysis crudely approximates the powerdensity a receiving antenna produces asE2/Z0, where Z0 is the radiation resis-tance of free space (377 ohms) and E isthe local electric field strength involts/meter.3 An electric field (E) of 1V/m thus yields only 0.26 µW/cm2, andfield strengths of even a few volts permeter are rare, except when close to apowerful transmitter. Although crystalradios derive their energy from broad-cast signals, the receiving antenna can berestrictively large unless the set is closeto a transmitter, and access to a goodground is usually required. Even so, acrystal set’s energy harvest is limited, typ-

ically on the order of tens of microwatts.Another approach, heralding Nicola

Tesla a century ago and William C.Brown 50 years later, involves broad-casting RF energy deliberately to powerremote devices. This practice is now com-monplace in passive radio-frequency-identification systems, which derive theirenergy inductively, capacitively, or radia-tively from the tag reader. Safety and USFederal Communications Commissionrestrictions severely limit availablepower—RFID tags generally consumebetween 1 and 100 µW.

Ambient light presents another oppor-tunity to scavenge power. The energyconversion efficiency of relatively inex-pensive crystalline silicon solar cell mod-ules is generally below 20 percent andcloser to 10 percent for flexible amor-phous silicon panels (standard solar cellsproduce roughly 100 mW/cm2 in brightsun and 100 µW/cm2 in a typically illu-minated office). Accordingly, the lack ofstrong, consistent sunlight constrainsapplications. Nonetheless, the estab-lished products that harness light as anenergy source span several orders ofmagnitude in power. These range fromsolar homes producing kilowatts on abright day, to solar battery chargers forcell phones that purport to produce up to2 W of power in direct sunlight, to a

PDA that runs off a panel of solar cellslining its case,4 to Citizen’s Eco-Drivewatch, which is powered by a solar cellhidden beneath a translucent dial. Re-searchers continually strive to refinesolar cell materials and technologies toincrease efficiency (today’s best researchdevices have reached 34 percent).

Thermoelectric conversionObjects or environments at different

temperatures offer the opportunity forenergy scavenging via heat transfer (theheat engine running in a geothermalpower station is a familiar large-scaleexample). The Carnot cycle—compris-ing adiabatic and isothermal opera-tions—provides the fundamental limitto the energy obtained from a tempera-ture difference. The Carnot efficiency is(TH – TL)/TH � � T/TH, where TL and TH

are the low and high temperatures(degrees Kelvin) across which the ther-mal generator operates. Accordingly,Carnot efficiencies are limited for small�T—for example, going from body tem-perature (37 degrees Celsius) to a coolroom (20 degrees Celsius) yields only 5.5percent. Mechanical products have beendesigned to run off the meager energythey harvest from ambient thermalcycling. For example, the ATMOS clockharnesses air pressure changes that result

JANUARY–MARCH 2005 PERVASIVEcomputing 19

1990 1994 1994Year

Disk capacityCPU speedAvailable RAMWireless transfer speedBattery energy density

1996 1998 2000 2002

Impr

ovem

ent m

ultip

le s

ince

199

0

1,000

100

10

Figure 1. Relative improvements inlaptop computing technology from1990–2003. The wireless-connectivitycurve considers only cellular standards inthe US, not including emerging, short-range 802.11 “hot spots.” The dip marksthe removal of the Metricom network.

20 PERVASIVEcomputing www.computer.org/pervasive

ENERGY HARVESTING & CONSERVATION

from dynamic thermal conditions.5 Re-search into novel thermoelectric andthermionic materials and the developmentof new kinds of devices, such as thermaldiodes and micron-gap thermophoto-voltaics, aspires to improve device perfor-mance. Energy conversion, however, lim-its available thermopile arrays, whichattain efficiencies well under 10 percent for200 degrees Celsius to 20 degrees Celsius,but below 1 percent for 40 degrees Celsiusto 20 degrees Celsius.6 Accordingly,thermoelectric generators can deliver sig-nificant power with high-temperaturesources (such as a hot exhaust pipe) butare much more limited for wearable appli-cations or temperate environments.

Nonetheless, companies have intro-duced a few wearable thermoelectricproducts over the last years. For exam-ple, the Seiko Thermic wristwatch (seeFigure 2) uses 10 thermoelectric mod-ules to generate sufficient microwatts torun its mechanical clock movement fromthe small thermal gradient provided bybody heat over ambient temperature.

Applied Digital Solutions’ Thermo Lifeis a thermoelectric generator measuring0.5 cm2 in area by 1.6 millimeter thick-ness. Comprising a dense array of Bi2Te3

thermopiles (most efficient at tempera-tures of 0 to 100 degrees Celsius) de-posited onto thin film, it can generate 10µA at 3 V (6 V open circuit) with only 5degrees Celsius of temperature difference(according to a personal communica-tion from Ingo Stark of Applied Digi-tal Solutions’ Thermo Life EnergyCorp.). Accordingly, Thermo Life gen-erators can power low-drain biosensorelectronics when in contact with the skin.These systems typically come with bat-teries that store extra energy producedduring periods of higher �T so they cancontinue to run during warmer, less effi-cient ambient temperatures.

Vibrational excitationFrom the subtle vibrations of floors

and walls that nearby machinery causesto the severe excitation of an automobilechassis or jet engine housing, mechanical

stimuli are common in many settings butcan vary widely in frequency and ampli-tude. Inventors have long designed sys-tems to harvest this energy, usually byexploiting the oscillation of a proof massresonantly tuned to the environment’sdominant mechanical frequency.

Self-winding watches use the motionof the user’s body to wind their mecha-nisms. Abraham-Louis Perrelet createdthe first known self-winding pedometerwatch around 1770, although indica-tions suggest that others might havemade earlier versions in the 1600s.7

Widespread adoption of these systemsdidn’t occur until after the 1930s, whenwatch cases could be hermetically sealedto protect the mechanism from dust. Amodern self-winding wristwatch con-tains an approximately 2-gram rotary-proof mass mounted off-center on a spin-dle. As the user moves during the day, themass spins and winds the mechanism.

Figure 3 shows diagrams of two elec-tronic self-winding watch mechanisms. Inthe ETA Autoquartz, a proof mass winds

1.7 mm 2.14 mm

2.14 mm 2.36 mm

Thermoelectric module

Boosterintegrated circuit Arm

1.27 mm

(c)

(a)

(d)

(b)

Watch movementBattery

Adiabaticcase

Thermoelectricmodules

Heat flowHeat flow

Figure 2. The Seiko Thermic wristwatch: (a) the product; (b) a cross-sectional diagram; (c) thermoelectric modules; (d) a thermopilearray. Copyright by Seiko Instruments.

a spring that pulses a microgenerator atits optimal rate of 15,000 rotations perminute for 50 ms, yielding 6 mA bursts atgreater than 16 V that are integrated on astorage capacitor. Seiko’s AGS (AutomaticGenerating System) omits the intermediatespring and produces 5 µW on averagewhen the watch is worn and 1 mW whenthe watch is forcibly shaken. Similarly,shake-powered flashlights employ a mag-netic proof mass that passes through asolenoidal coil, bouncing against rubber

bumpers at each end.8 A typical device’sgenerating mechanism (see Figure 3c)weighs 150 grams and produces 200 mWwith a steady shake at its mechanical res-onance (roughly 200 cycles/minute).

Driven by strong interest in long-livedwireless sensing packages, many groupshave recently developed vibrational micro-generators. (Paul Mitcheson and his col-leagues present a review,9 and we also citeseveral devices in our book chapter.2) Thebasis for many are moving magnets or

coils, as in Figure 3. Although some havebeen fabricated at the scale of micro-electromechanical systems (MEMS),these tend to be larger structures, rang-ing from 1 to 75 cm3 (Ferro Solutions’Harvester) or more, exploiting vibrationsranging from 50 Hz to 5 KHz that inducemechanical oscillations between one-halfmicrometer to over one millimeter, andproducing powers that range from tensof microwatts (MEMS) to over a milli-watt (larger devices).


Gear train

RotorStator Coil

Oscillating weight

(c)

(a) (b)

Rotor Coil

Oscillatingweight gear

Transmission gear

Stator

Integrated circuit Quartz

Motor

Charge controlcircuit

Drive circuit

Secondarypower supply

Oscillatingweight

Oscillatingweight

Accumulator

Microgenerator

Microbarrel

Figure 3. Commercial inertial-power scavengers. Two mechanisms for self-winding electric watches—(a) the ETA Autoquartz designand (b) the Seiko AGS (Automatic Generating System) generator for the Kinetic series—and (c) an inertial solenoid generator froma commercial shake-driven light-emitting-diode flashlight. (diagrams courtesy of the Swatch Group and Seiko Epson)



Other approaches employ piezoelec-tric materials. The 1967 US Patent3,456,134 proposed using a small, tip-loaded piezoelectric cantilever for pow-ering bioelectric implants, claiming thata prototype device produced 150 µWwhen mechanically coupled to 80-Hzheartbeats.10 Eighteen years later, USPatent 4,510,484 proposed a similardevice for powering wireless sensors inautomobile tire hubs.11 Recently, ShadRoundy and his collaborators have devel-oped a compact piezoelectric generatormade from a pair of laminated PZT (leadzirconium titanate) strips to form a tip-loaded, cantilevered bimorph beam thatproduced nearly 100 µW when shakenat resonance.12 Piezoelectrics that arebonded to vibrating structural membersin bridges, buildings, or aircraft canachieve sufficient strain to generate usefulenergy. Researchers at Sandia NationalLaboratories and MicroStrain, for exam-ple, have recently developed such strain-based energy harvesters for powering

data-logging monitors and wireless sen-sor nodes mounted on large structures.13

Ocean Power Technologies exploitedstrips of piezoelectric foil to harness thepower of flowing water, in the form of18-inch-long “eels” that flap in turbulentflow like a flag in the wind.14 Dynamicstrain coupled from the tire of a movingcar has also powered a wireless tire con-dition monitor.15

Other microgenerators consist of acharged capacitor with moving plates.Unlike magnetic and piezoelectric gen-erators, such electrostatic generatorsneed to be “jump-started” with an ini-tial voltage before they can producepower. Because the force between thecapacitor’s electrodes increases with thesquare of the applied voltage, the poten-tial to do work (and thus generate elec-trical energy) correspondingly increasesas more charge is loaded onto the capac-itor, until charge leakage becomes sig-nificant or mechanical problems occurowing to excess force. Although someresearchers have developed larger elec-trostatic generators aiming to exploitlow-frequency vibrations in walls orwearable applications, electrostatic gen-erators tend to be small, often MEMS-scale devices, where they benefit signifi-cantly from being able to increase energydensity with applied voltage. Accord-ingly, developers design them to be drivenat frequencies ranging from hundreds ofhertz to several kilohertz, and, depend-ing on their excitation and power condi-tioning, they typically yield on the orderof 10 µW. Therefore, they’re intended tosupport extremely low-power applica-tions, perhaps sited on the same chip asthe generator.

Figure 4 shows two such MEMS-scalemicrogenerators. The device in Figure4a, which Jeffrey Lang and AnanthaChandrakasan’s team developed at

MIT,16 is a constant-gap variable capac-itor. The capacitor’s plates are realizedby the interdigitated tines at the upperleft that slide together and apart in thehorizontal plane with applied vibration.A portion of an 84-mg proof mass is vis-ible at the lower left, and the structureat the lower right is a flexure spring sus-pending the comb. The team designedthis device to produce 8 µW with half-millimeter comb displacement at 2.5kHz. Most microgenerators exhibit amechanical resonance (often in the kHzrange at MEMS scale) where vibrationcan be efficiently coupled. Because envi-ronmental excitation, especially fromhuman motion, can occur over a widespectrum of generally lower frequencies,some researchers have developed non-resonant or tunable microgenerators.Imperial College’s Peng Miao and hiscolleagues have recently realized such adevice (see Figure 4b) using a nonreso-nant snap-action restoring force on theproof mass instead of a continuousspring. Consisting of three stacked wafersand a gold-plated, 0.3-g proof mass, thecapacitor’s active area is 210 mm2, andthe plates are capable of a 320 µm maxi-mum displacement. Measurements haveyielded up to 300 nanojoules per mechan-ical cycle at 20 Hz (so, 6 µW).

A recent analysis indicated that we canexpect up to 4 µW/cm3 from vibrationalmicrogenerators (of order 1 cm3 in vol-ume) that typical human motion (5 mmmotion at 1 Hz) stimulates and up to 800µW/cm3 from machine-induced stimuli(2 nm motion at 2.5 kHz).9 As this sur-passes current devices’ performance bybetween one and three orders of magni-tude, researchers have plenty of room forimprovement.

Power from human inputAnother source of batteryless power is

Figure 4. Two electrostatic generators fabricated in microelectromechanical systems:(a) a portion of a constant-gap variable capacitor fabricated at MIT for a 1 cm2 deviceand (b) a 2 cm2 compressible-plate capacitor fabricated at Imperial College. (figure 4aphoto courtesy of Anantha Chandrakasan and Jeff Lang, MIT Microsystems TechnologyLab; figure 4b photo courtesy of Peng Miao)

(a)

(b)

deliberate human input. Inventors havefound many opportunities here, exploit-ing cranking, shaking, squeezing, spin-ning, pushing, pumping, and pulling topower their devices.2 Windup magnetic-generator-powered flashlights date to theearly 20th century,17,18 sprouting descen-dents such as windup cell phone chargersand radios. In a typical windup radio,such as those FreePlay makes, 60 turns(1 minute of cranking) stores 500 joulesof energy in a spring, which drives a mag-netic generator that’s 40 percent efficient,metering out enough power for up to anhour of play. By analogy, hand-chargingone of today’s 30–50 W-hr. laptop com-puters would require more than an hourof cranking (and involve a heavy andpotentially dangerous spring).

Some inventors have developed moreunusual methods of coupling energy intogenerators. At the MIT Media Lab, SaulGriffith’s Bettery—inspired by the Abo-riginal musical instrument, the bull-roarer—employs a 100–200 gram balltethered via a 0.3–0.5 m string to a hand-held generator. Revolving the ball at 1–2Hz produces 3–5 W, ample power for acell phone call.

Robert Adler designed a batteryless,wireless remote control for Zenith tele-visions in 1956 called the Space Com-mander (see Figure 5a). Its lineage is morethan a century old, in conveyances suchas the desk bell. It featured a set of but-tons that struck aluminum rods to pro-duce ultrasound, which, when decodedat the television, turned on the power,changed channels, or muted the vol-ume.19 Active infrared remote controlsreplaced the Space Commander morethan 25 years after Adler designed it.

Joe Paradiso and Mark Feldmeiertook this theme further in 2001 by using

a piezoelectric element with a resonantlymatched transformer and conditioningelectronics that, when struck by a but-ton, generated approximately 1 mJ at 3 V per 15N push, enough power to runa digital encoder and a radio that cantransmit over 50 feet.20 This device (seeFigure 5b), built entirely with off-the-shelf components, enables placing com-pact digital controllers (such as a lightswitch) freely, without needing wiring orbatteries and their associated mainte-nance. In the US, a NASA-Langley spin-off is marketing the Lightning Switch,another self-powered piezoelectric radiobutton. The German company EnOceanis marketing self-powered radio trans-mitters, energized by a bistable piezo-electric cantilever that snaps whenpressed, conditioned by a switching reg-ulator. Their PTM100 produces about100 µJ per 8N push at 3.3 V.

Ambulatory power generationFrom resting to a fast sprint, the

human body expends roughly 0.1–1.5kW.2 The devices mentioned in this arti-cle, however, typically scavenge less thana milliwatt from human activity beforetheir presence becomes obvious orannoying. The most promising way toextract energy more innocuously frompeople is by tapping their gait. Humanstypically exert up to 130 percent of theirweight across their shoes at heel strikeand toe-off, and standard jogging sneak-ers’ cushioned soles can compress by upto a centimeter during a normal walk.For a 154-pound person, this indicatesthat about 7 W of power could be avail-able per foot at a 1-Hz stride from heelstrike alone.

Patents for electric shoes, based mostlyon shoe-integrated magnetic generators,

date back up to 80 years.21–23 Rotarygenerators’ mechanics, however, are dif-ficult to integrate reliably into standardfootwear, as evident in Figure 6, whichshows two MIT-built generator shoesfrom the late nineties.24 Although thesole-integrated generator in Figure 6bwas somewhat less cumbersome, bothdevices were fragile, as might be expectedwhen such complicated mechanics aremounted in shoe soles.

Other approaches that lack movingmechanics, such as piezoelectric or capac-itive generators, are more promising.James Antaki and his collaborators atthe University of Pittsburgh presented ashoe-mounted piezoelectric generator in1995 that they developed to power arti-ficial organs.25 Their device incorpo-rated two cylindrical tubes in the insole,each housing a PZT stack stimulated bya passive hydraulic pulser-amplifier. Theamplifier converted low-frequency foot-fall energy into an intense series of high-frequency impulses that drove the PZT at


Figure 5. Self-powered wireless pushbutton remote controls: (a) A 1956 Zenith SpaceCommander, a passive ultrasound transmitter, and (b) the MIT Media Lab’s self-powered wireless switch, which sends a digitally coded radio frequency stream.

(a)

(b)

its mechanical resonance. The hydraulicreservoir was differentially compressedduring heel strike and toe-off, so powerwas extracted across the entire gait.Although the prototype was somewhatbulky and heavy, the shoe contained theentire generator. Walking produced aver-age powers of 250–700 mW (dependingon the user’s gait and weight), and a sim-ulated jog produced over 2 W. MIT’sNesbitt Hagood has pursued a modern

version of this approach using a MEMS-fabricated, active hydraulic valve.26

Figure 7 shows a simple integration ofpiezoelectric elements beneath a stan-dard running sneaker’s removable insole.Joe Paradiso’s team at MIT built the shoein 1998 and refined it in 1999.24 Theshoe scavenged energy from heel strikeby flattening a clamshell made from twoback-to-back Thunder PZT/spring-steelunimorphs and from toe-off by bending

a bimorph stave made from 16 layers(two 8-layer laminates sandwiching a 1mm neutral core) of piezoelectric PVDF(polyvinylidine flouride) foil. Owing tolimited electromechanical conversionefficiency, the average power harvestedwas small (8.3 mW at the heel and 1.3mW at the toes during a standard walk).However, there was no interference withgait, and the piezoelectric elements wereeffectively hidden in the shoe. Each shoeproduced sufficient energy to transmit a12-bit ID code via an onboard radio tothe local area as the wearer walked.

Intrinsic material properties limit piezo-electric generators’ efficiency, whereasrunning capacitive generators at a highervoltage can improve their performance.To take advantage of this property, RonPelrine and his collaborators at SRIInternational have developed electro-static generators based around materi-



Figure 6. Magnetic generators in shoes at the MIT Media Lab: (a) A strap-on overshoe produced an average of 250 mW during astandard walk, powering a loud radio; (b) an assembly hosting twin motor-generators and step-up gears embedded directly into asneaker’s sole (without springs or flywheels for energy storage) produced 60 mW.

(c)

(a) (b)

(d)

InsoleSole

Polyvinylidine flouride stave

Metalmidplate

PZTbimorph

PZTunimorph

PZT unimorph

Figure 7. Integration of (a) a flexible PZTThunder clamshell and (b) a 16-layerpolyvinylidine flouride bimorph staveunder (c) the insole of a running shoe,resulting in (d) operational power-harvesting shoes with heel-mounted electronics for power conditioning,energy storage, an ID encoder, and a300-MHz radio transmitter.

(a) (b)


als called electroactive polymers ordielectric elastomers. Made from com-ponents such as silicone rubber or softacrylics, these materials are extremelycompliant—a displacement of 2–6 mmcan easily drive them to 50–100 percentarea strain, depending on the generator’sconfiguration. So, they’re ideal substi-tutes for a running shoe’s rubbery heel.They can also be efficient, with a practi-cal device achieving energy densities of0.2 J/g and calculations indicating a pos-sibility of approaching 1.5 J/g.

The SRI team has built an elastomergenerator into a boot heel (see Figure 8).When the heel presses down, the bellowscompress, applying pressure to the elas-tomer membrane. The membrane bal-loons into the holes in the frame, pro-ducing strain, and when voltage is appliedacross the electrodes, it produces power.They’ve achieved an energy output of 0.8J per step with this boot with a heel com-pression of only 3 mm, yielding 800 mWof power per shoe at a pace of two stepsper second.27 Benchtop testing indicatedthat the material will last for at least100,000 cycles. However, the teambelieves that improved packaging anddesign can increase the lifetime beyond 1million cycles—enough to meet commer-cial footwear’s required lifetime. As morecompression is feasible in a commercialshoe, they anticipate being able to extract1 W of power, allowing for a 50 percentvoltage conversion (from several kilo-volts to 3 V) and storage efficiency.

Matthew Laibowitz and Joe Paradisohave recently explored a means of har-vesting translational energy and localnavigational intelligence for mobile sen-sor networks. Inspired by nature’sprocess of phoresis, they’ve looked at theway small organisms (such as ticks,nematodes, remoras, and burrs) hitchrides on other animals.28 In what theyterm parasitic mobility, these nodesrequire no complicated, massive, orpower-hungry systems to move aroundand navigate—they need only a means ofdetecting and attaching to hosts andlocating themselves as they move. Simu-lations have indicated that although per-formance depends on appropriate hosts’density and behavior along with thenode’s requirements for detecting hostsand attaching, this technique is at leastan order of magnitude more energy effi-cient than driving a robotic body directlyto a destination. They’ve built proof-of-concept hardware to explore four para-sitic mobility modes. Active attachmentnodes, like nature’s ticks, launch them-selves at a nearby host and detach whenthey’re either at their goal location or aremoving in the wrong direction. Passivenodes, like nature’s burrs, stick to thenext object that they come in contactwith (quasipassive nodes can shake offwhen they want to leave the host). Value-

added nodes are symbiotic objects thatpeople carry (such as pens)—the sensorand communications suite is appendedto the device’s user-attracting functionsand travels with it. Although the currentactive and passive devices are too largeto innocuously attach to people and ani-mals, they’re amenable to vehicular hosts.

Although many different tech-niques are available to harvestenergy from various environ-ments to power electronics,

the amount of available raw energy (forexample, sunlight, vibration, heat) andthe surface area or net mass that thedevice permits limit the power yield inpervasive computing’s everyday habi-tated settings. With the exception ofheel-strike harvesting in electric shoes orsolar cells in bright light, available pow-ers generally hover at mW or µW levels(see Table 1). Nonetheless, researchersare striving to marry more miserly powermanagement techniques with electron-ics that consume less energy, enablingembedded devices to conduct more usefuloperations with the limited power thatthey can commonly scavenge. Accord-ingly, energy harvesting is an area of rapiddevelopment, and the day approaches

Figure 8. An electrostatic generator basedon compression of a charged dielectricelastomer during heel strike: (a) a prototype implementation in a boot and(b) detail of the generator, showing thebellows on the bottom and the retainingframe on top. (photos courtesy SRI International)

(a) (b)

when component life rather than batterycharge will limit low-duty-cycle sensorsystems.

REFERENCES1. M.B. Schiffer, The Portable Radio in Amer-

ican Life, Univ. of Arizona Press, 1992.

2. T. Starner and J.A. Paradiso, “Human-Generated Power for Mobile Electronics,”Low-Power Electronics Design, C. Piguet,ed., CRC Press, 2004, chapter 45, pp. 1–35.

3. E.M. Yeatman, “Advances in PowerSources for Wireless Sensor Nodes,” Proc.Int’l Workshop Wearable and ImplantableBody Sensor Networks, Imperial College,2004, pp. 20–21; www.doc.ic.ac.uk/vip/bsn_2004/program/index.html.

4. H. Schmidhuber and C. Hebling, “First

Experiences and Measurements with a SolarPowered Personal Digital Assistant (PDA),”Proc. 17th European Photovoltaic SolarEnergy Conf., ETA-Florence and WIP-Munich, 2001, pp. 658–662.

5. J. Lebet, Living on Air—History of theATMOS Clock, Jaeger-LeCoultre, 1997.

6. J. Stevens, “Optimized Thermal Design ofSmall ∆T Thermoelectric Generators,” Proc.34th Intersociety Energy Conversion Eng.Conf., Soc. of Automotive Engineers, 1999,paper 1999-01-2564; www2.msstate.edu/~stevens/iecec.pdf.

7. A. Chapuis and E. Jaquet, The History ofthe Self-Winding Watch, Roto-Sadag, 1956.

8. S.R. Vetorino et al., Renewable EnergyFlashlight, US patent 6,220,719, to AppliedInnovative Technologies, Inc., Patent andTrademark Office, 2001.

9. P.D. Mitcheson et al., “Architectures forVibration-Driven Micropower Genera-

tors,” J. Microelectromechanical Systems,vol. 13, no. 3, 2004, pp. 429–440.

10. Piezoelectric Energy Converter for Elec-tronic Implants, US patent 3,456,134,Patent and Trademark Office, 1969.

11. D. Snyder, Piezoelectric Reed Power Sup-ply for Use in Abnormal Tire ConditionWarning Systems, US patent 4,510,484, toImperial Clevite, Inc., Patent and Trade-mark Office, 1985.

12. S. Roundy, P.K. Wright, and J. Rabaey, “AStudy of Low Level Vibrations as a PowerSource for Wireless Sensor Nodes,” Com-puter Communications, vol. 26, no. 11,2003, pp. 1131–1144.

13. D.L. Churchill et al., “Strain Energy Har-vesting for Wireless Sensor Networks,”Smart Structures and Materials 2003: Mod-eling, Signal Processing, and Control, Proc.SPIE, vol. 5005, Int’l Soc. Optical Eng.,2003, pp. 319–327; www.microstrain.com/white/pdf/strainenergyharvesting.pdf.



TABLE 1Energy-harvesting opportunities and demonstrated capabilities.

Energy source PerformanceA Notes

Ambient radio frequency < 1 µW/cm2 Unless near a transmitter3

Ambient light 100 mW/cm2 (directed toward bright sun) Common polycrystalline solar cells are 16%–17% efficient, 100 µW/cm2 (illuminated office) while standard monocrystalline cells approach 20%.

Although the numbers at left could vary widely with a given environment’s light level, they’re typical for the garden-variety solar cell Radio Shack sells (part 276-124).

Thermoelectric 60 µW/cm2 Quoted for a Thermo Life generator at �T = 5�CB; typical thermoelectric generators � 1% efficient for �T < 40�C.6

Vibrational 4 µW/cm3 (human motion—Hz) Predictions for 1 cm3 generators.9 Highly dependent onmicrogenerators 800 µW/cm3 (machines—kHz) excitation (power tends to be proportional to �3 and y0

2, where � is the driving frequency and y0 is the input displacement), and larger structures can achieve higher power densities. The shake-driven flashlight of Figure 3, for example, delivers 2 mW/cm3 at 3 Hz.

Ambient airflow 1 mW/cm2 Demonstrated in microelectromechanical turbine at 30 liters/min.29

Push buttons 50 µJ/N Quoted at 3 V DC for the MIT Media Lab Device.20

Hand generators 30 W/kg Quoted for Nissho Engineering’s Tug Power (vs. 1.3 W/kg

for a shake-driven flashlight).2

Heel strike 7 W potentially available (1 cm deflection Demonstrated systems: 800 mW with dielectric elastomerat 70 kg per 1 Hz walk) heel,26 250–700 mW with hydraulic piezoelectric actuator

shoes,24 10 mW with piezoelectric insole.25

AThese numbers depend heavily on the ambient excitation and harvesting technologies. By comparison, lithium-ion batteries can yield up to 0.52W-hr./cm3 (0.18 W-hr./g), and the Toshiba DMFC (direct methanol mini fuel cell) achieves 0.27 W-hr./cm3. The theoretical energy available frommethanol is 4.8 W-hr./cm3 (6.1 W-hr./g).BQuoted by Ingo Stark of Applied Digital Solutions’ Thermo Life Energy Corp.

JANUARY–MARCH 2005

14. G. Taylor and S. Kammann, “Experimentaland Theoretical Behavior of Piezoelec-tric/Electrostrictive ‘Eel’ Structures,” DARPA

Energy Harvesting Program Rev., R.Nowak, ed., 2000; www.darpa.mil/dso/trans/energy/briefing.html.

15. Vehicular Mounted Piezoelectric Genera-tor, US patent 4,504,761, Patent and Trade-mark Office, 1985.

16. S. Meninger et al., “Vibration-to-ElectricEnergy Conversion,” IEEE Trans. VeryLarge Scale Integration (VLSI) Systems, vol.9, no. 1, 2001, pp. 64–76.

17. Self Contained Generating and LightingUnit, US patent 1,184,056, Patent andTrademark Office, 1916.

18. Magneto Flash Light, US patent 1,472,335,Patent and Trademark Office, 1923.

19. R. Adler, P. Desmares, and J. Spracklen,“Ultrasonic Remote Control for HomeReceivers,” IEEE Trans. Consumer Elec-tronics, vol. 28, no. 1, 1982, pp. 123–128.

20. J. Paradiso and M. Feldmeier, “A Compact,Wireless, Self-Powered Pushbutton Con-troller,” Ubicomp 2001: Ubiquitous Com-puting, LNCS 2201, Springer-Verlag, 2001,pp. 299–304.

21. Electric Shoe, US patent 1,506,282, Patentand Trademark Office, 1924.

22. Shoe with Internal Foot Warmer, US patent4,674,199, Patent and Trademark Office,1987.

23. Dynamoelectric Shoes, US patent 5,495,682,Patent and Trademark Office, 1996.

24. N.S. Shenck and J.A. Paradiso, “EnergyScavenging with Shoe-Mounted Piezo-electrics,” IEEE Micro, vol. 21, no. 3, 2001,pp. 30–42.

25. J.F. Antaki et al., “A Gait-Powered Autol-ogous Battery Charging System for Artifi-cial Organs,” ASAIO J., vol. 41, no. 3,1995, pp. M588–M595.

26. O. Yaglioglu, “Modeling and Design Con-siderations for a Micro-Hydraulic Piezo-electric Power Generator,” master’s thesis,Dept. Electrical Eng. and Computer Science,Massachusetts Inst. of Technology, 2002.

27. R.D. Kornbluh et al., “Electroelastomers:Applications of Dielectric Elastomer Trans-ducers for Actuation, Generation, andSmart Structures,” Smart Structures andMaterials 2002: Industrial and Commercial

Applications of Smart Structures Tech-nologies, Proc. SPIE, vol. 4698, Int’l Soc.Optical Eng., 2002, pp. 254–270.

28. M. Laibowitz and J.A. Paradiso, “ParasiticMobility for Pervasive Sensor Networks,”to be published in Proc. 3rd Ann. Conf. Per-vasive Computing (Pervasive 2005),Springer-Verlag, 2005.

29. A.S. Holmes et al., “Axial-Flow Microtur-bine with Electromagnetic Generator:Design, CFD Simulation, and PrototypeDemonstration,” Proc. 17th IEEE Int’lMicro Electro Mechanical Systems Conf.(MEMS 04), IEEE Press, 2004, pp.568–571.

For more information on this or any other comput-ing topic, please visit our Digital Library at www.computer.org/publications/dlib.

the AUTHORS

Joseph A. Paradiso is theSony Career DevelopmentAssociate Professor of MediaArts and Sciences and directorof the Responsive Environ-ments Group at the Massa-chusetts Institute of Tech-nology Media Laboratory.

His research interests include sensor systems forhuman-computer interfaces and ubiquitous com-puting. He received his PhD in physics from MITas a K.T. Compton Fellow. He’s a member of theIEEE, ACM, AIAA, APS, OSA, Sigma Xi, Tau BetaPi, and Eta Kappa Nu. Contact him at MIT MediaLaboratory, 20 Ames St. E15-327, Cambridge,MA 02139; [email protected]; www.media.mit.edu/~joep.

Thad Starner is an assistantprofessor of computing atGeorgia Tech's GVU Center,where he directs the contex-tual-computing group. Hisresearch interests includemobile computing, human-computer interaction, and

computational perception. He received his PhDfrom the Massachusetts Institute of TechnologyMedia Laboratory. He’s a member of the IEEE,ACM, and, because of this article, the NationalAssociation of Watch and Clock Collectors. Con-tact him at 85 5th St. NW, College of Computing,Georgia Tech, Atlanta, GA 30332-0760; [email protected]; www.cc.gatech.edu/~thad.

Stayinthesoftware

game!

Stayinthesoftware

game!

Don’t miss this year’s special issues, including:

POSTMODERN SOFTWARE DESIGN

•ASPECT-ORIENTED

PROGRAMMING•

PROJECT MANAGEMENT•

INCORPORATING COTS INTO DEVELOPMENT

IEEE

To subscribe or renew, visit

www.computer.org/subscribe

IEEE Software magazine is building

the community of leading software

practitioners by delivering practi-

cal, peer-reviewed information to

developers and managers who want

to keep up with today’s rapid tech-

nology change.

Energy Scavenging for Mobile and Wireless Electronics · laptop computers (a mature mobile...

Documents

Transcript of Energy Scavenging for Mobile and Wireless Electronics · laptop computers (a mature mobile...