energy provision and storage for pervasive computing

28 PERVASIVE computing Published by the IEEE CS n 1536-1268/16/$33.00 © 2016 IEEE

Energy Provision and Storage for Pervasive Computing

A s Mark Weiser’s vision of ubiq-uitous computing continues to become reality, significant tech-nical challenges remain. Chief among them is achieving autono-

mous and long-term operation without using wires or “tethered” interfaces. From a com-

munications perspective, solu-tions have emerged quickly, including cellular, Wi-Fi, Blue-tooth, RFID, IEEE 802.15.4, and LoRa. However, there is growing interest in providing sustainable energy for per-vasive devices using wireless interfaces—to increase auton-omy, reduce clutter, reduce

maintenance requirements, and expand appli-cation potential. Given increasing acceptance in consumer electronics of the dedicated transmit-ter-receiver model, whereby power is intention-ally transferred from a source to one or more receivers (using Qi chargers, for example), it’s worth investigating the model’s applicability in complementary pervasive computing scenarios, such as applications using embedded sensors and actuators.

This raises interesting design questions in terms of the energy transmission’s range; physical size of the transmitters and receivers;

efficient management of on-board conversion, storage, and management; and handling of het-erogeneous mobility patterns in which receiv-ers might be in contact with sources for limited time periods (due to being mobile devices themselves, or because they use mobile energy sources). There are also cases in which both the source and receivers might be static or mobile (see Figure 1).1–3 The energy design space is thus growing, and wireless energy transfer is a promising candidate when recharging is neces-sary and feasible. Here, we discuss key factors that should be considered and incorporated into this emerging energy design space.

Energy and Mobility in Pervasive ComputingSmartphones are ubiquitous in modern society, typifying mobile devices with static wireless connectivity points and tethered recharging, lately supplemented by “wireless” charging (us-ing Qi chargers that exploit magnetic induction and resonance). This also holds for tablets and notebooks in their typical use. As the desktop continues its decline, a common characteris-tic requirement in pervasive computing is for devices to operate without a tethered energy source, particularly in challenging monitoring and control applications with deeply embedded devices.

Soon, devices requiring sufficient energy to operate maintenance-free for years will render today’s energy solutions insufficient, yet there are no tools or systematic methods for exploring the emerging energy design space. The authors discuss key factors such an energy design methodology should incorporate.

E n E r g y H a r v E S t i n g

David E. Boyle, Michail E. Kiziroglou, Paul D. Mitcheson, and Eric M. YeatmanImperial College London

octobEr–dEcEmbEr 2016 PERVASIVE computing 29

This requirement has traditionally been met by including an appropriate battery; however, a range of complemen-tary options has recently emerged that will help overcome many of the limita-tions of battery-only powering of devices.

Battery Capacity and Energy Harvesting ConsiderationsThinking about the power requirements of mobile devices, the range is typically 1–30 watts, with smartphones on the lower end, powerful notebooks on the upper end, and instantaneous power varying depending on the tasks per-formed. We tend to carry these devices around with us and accept regular charg-ing as a fact of life. Most of these devices are mass-market consumer products, and “battery life” is often used as a mar-keting tool. With the recent emergence of high-performance wearables, such as

smartwatches, it has become common to discuss how much utility can be had for so many minutes of charging.

Quantifying performance versus user acceptance in these terms is dif-ficult and subjective, but many com-panies and consumers accept that, for moderate use, one day between charge cycles is reasonable. Simultaneously, charging times are improving but are also difficult to comparatively analyze due to battery capacity variations. Typ-ical batteries can’t be charged in less than 30–60 minutes as a compromise between charging speed and rate of ca-pacity degradation. Today’s high-end devices typically charge from empty to full in 1–3 hours, and because they mostly use lithium-ion batteries, they have around 300–500 charge cycles (approximately two years) before they need to be replaced.

Emerging applications of distrib-uted and embedded wireless sensor networks (WSNs) and wireless sensor and actuator networks (WSANs) have many mixed-mobility scenarios and were initially considered constrained in practical utility by finite energy supplies (batteries). The lack of accep-tance of the associated maintenance requirements led to research efforts focused on energy efficiency across a spectrum of related themes—most notably, communications protocols and associated algorithms. WSN ap-plications are diverse in scope but typi-cally consist of static devices deployed across a sensing field, where data is transmitted across one or more hops toward a sink/gateway. Alternatively, data is collected via “muling” with a mobile sink—an example of delay-tol-erant networking (DTN).

Figure 1. Mobility scenarios in contemporary pervasive computing. The source and receivers can be static, mobile, or a combination of the two.1–3

Static Mobile

Mob

ile

Sour

ces

Receivers

Examples

- Wireless sensor nets with mobile sources and static nodes

- Wireless charging vehicle and renewable energy cycle (shown at right1)

- Drone or vehicle- based energy delivery

Examples

- WSN with buried or inaccessible devices, such as from mi6sense.org (shown at right)

- Wireless powering of static IoT devices; examples include Energous and WiTricity

Examples

- Smartphones, tabs, and so on using Qi

Examples

- Vehicle-to-vehicle charging scenarios, such as aircraft (shown at right2)

- Rendezvous-based methods (higher complexity)

Stat

ic

Service station Mobile wireless charging vehicle

Sensor node

Base station

- Inductively charging vehicles on the road (see example on right3)

30 PERVASIVE computing www.computer.org/pervasive

EnErgy HarvESting

Improvements at the silicon level—in addition to low-power RF and proces-sor designs, monolithic system-on-chip (SoC) integration, and energy-efficient communications protocols—have all facilitated good progress in making en-ergy harvesting (particularly from so-lar and airflow sources) a viable option in the past 10 years for WSN scenar-ios.4,5 Many examples have emerged that exploit hybrid harvesting and storage approaches, often using mul-tisource harvesting and multistorage configurations typically comprising supercapacitors and lithium ion bat-teries.4 The devices typical of these ap-plications require tens to hundreds of microwatts in active operation, while ultra-low power modes can facilitate “sleeping” in the tens of nanowatts range, often incorporating maximum power point tracking (MPPT) and techniques to optimize switching be-tween buffers or direct powering based on real-time available energy.4,6

Mobile Wireless EnergyRecent works propose using mobile charging vehicles to provide wireless energy to static in-field devices,1 such as in the EU FP7 Mobesens (mobile wa-ter quality sensor system) project. This is based on nonradiative wireless power transfer (WPT).7 Interestingly, the pop-ularization of drone technology could result in them being used as power de-livery vehicles.

Home- or office-based Internet of Things (IoT) devices are likely to be static and could benefit from mobile wireless energy, but they’ll most likely exploit similarly static sources. In this case, coils could be built into walls or under flooring for inductive WPT. WiTricity (http://witricity.com) is de-veloping products that do exactly this. For home or office IoT scenarios, it’s also possible to consider other wireless energy sources, such as electromag-netic waves (that is, radiative sources). A commercial example of the RF ap-proach is being developed by Energous (http://energous.com); its system delivers

wireless energy at a distance of ap-proximately 5 m from a transmitter to a receiver—maintaining charging while the receiver is in motion—for up to 12 devices simultaneously.

acoustic Energy transferAcoustic energy transfer is another possibility that can potentially fit a combination of mobile and static transmitter and receiver scenarios. uBeam (http://ubeam.com) is a con-troversial startup attempting to gen-eralize wireless energy infrastructure based on ultrasonic energy transmis-sion, despite widely publicized tech-nical challenges and improbability of success.8 There are numerous other efforts to demonstrate acoustic energy transfer, particularly where receivers might be otherwise inaccessible, such as subterranean pipelines, structural monitoring, or body implants.

For example, methods of ultrasonic and inductive power transfer (IPT) for powering implanted devices through body tissue have been compared, find-ing that ultrasonics are preferable for longer distances when small receivers are used (for example, over 1.5 cm for a 5 mm receiver).9 Recently, acoustic power transfer through solid structures was demonstrated and is a promising method in cases where beneficial geom-etries exist, such as beams, pipelines, or panel structures.10

Opportunistic HarvestingAnother area of interest is energy har-vesting from infrastructure transmit-ters, such as broadcast signals, cellular base stations, or even Wi-Fi routers, which create ambient RF energy for would-be scavengers. However, such sources create power levels insufficient to energize contemporary devices—ranging from a fraction of a nanowatt (as in a small antenna used near a Wi-Fi router) to tens of nanowatts per cm2 on average (as in GSM1800 from cellular base stations) in urban environments.11 Additionally, such RF harvesting requires tailored, highly

efficient omnidirectional rectennas (rectifying antennas), probably in ar-ray configurations. Static or mobile devices in “energy rich” areas could adopt this approach, although with significant variability in the available energy supply.

Other examples of opportunistic har-vesting that necessitate static deploy-ment also exist, such as thermoelectric harvesting exploiting temperature dif-ferentials over time using phase-change materials for wireless sensor applica-tions—structural monitoring in air-craft is an example.12 Such harvesting approaches tend to be heavily optimized toward the targeted deployment sce-nario and the associated ambient avail-able energy sources, rather than being general-purpose solutions.

rendezvous ScenariosApplications based on rendezvous sce-narios between mobile sources and mobile devices/receivers are also possi-ble—particularly considering advances in drone technology and swarm coordi-nation and historical approaches to in-flight refueling in aviation.2 However, this is the least likely scenario to receive significant attention in the short- to medium-term for pervasive computing applications.

Static SourcesConsidering mobility patterns relevant to modern untethered pervasive com-puters, the only scenario that doesn’t require on-board energy storage is one in which the energy source and receiv-ers are static (or of limited mobility within the wireless energy transfer region of one or more transmitters), assuming no redundancy is required should the sources become unavail-able. Numerous recent examples of instantaneous WPT, such as power-ing light bulbs (60 W) and notebooks (12 W),13,14 are demonstrative of this approach. If the source fails, so does the device, but on-board storage costs are reduced or eradicated and the de-vice might be more compact.


Energy Design for Emerging Pervasive SystemsIn the following sections, our emphasis shifts from consumer electronics to-ward embedded monitoring and con-trol systems synonymous with WSN, cyberphysical systems, and IoT scenar-ios with high degrees of autonomy and ultra-long lived operational require-ments. In particular, we’re interested in where these systems intersect with the potential to exploit wireless energy transfer and harvesting. Nevertheless, heterogeneous mobility models, re-quirements, and characteristics are un-avoidable. Figure 2 provides a general overview of the active power of a num-ber of components and systems relevant to emerging pervasive computing sys-tems and their evolution over time.

Clearly, the power requirements vary over many orders of magnitude. The earliest generation of “mote” class devices comprised some combination of “cheap” energy-efficient sensors, an analog-to-digital converter, an RFIC, and a microcontroller. Active power is the sum of components’ active modes, and the primary energy cost is incurred during communication—typically in the tens of milliwatts range, irrespec-tive of chip manufacturers.

More recently, sleep or low-power modes of microcontrollers have been able to operate in the submicrowatt range—one example is the NXP LPC1100L series based on the ARM Cortex-M0 architecture. SoC imple-mentations, combining IEEE 802.15.4 radios and microcontrollers (such as TI-CC430 and NXP-JN5168), provide sim-ilar low-power modes and comparable RF power performance (approximately 15–17 mA in transmit and receive modes). There remains a convincing case to use duty-cycling techniques (where application requirements permit) to minimize net energy consumption.

Harnessing application-Level requirementsApplication-level performance and op-erational requirements are fundamental

considerations for energy design. Mini-mum performance and operational criteria are often stated in terms of availability, accessibility, throughput, reliability and security, and lifetime. Therefore, rather than being thought of as a constraint, “energy engineering” can be performed at design time for a given specific set of application-level requirements. This is particularly im-portant in the case of sensor-network application scenarios, where autono-mous field operation is required over long periods.

Consider an ad hoc WSN applica-tion tasked with periodically monitor-ing a phenomenon of interest, such as in long-term structural health moni-toring. An application designer might use aggressive duty cycles in each de-vice in the network to minimize energy use over time, while attempting to bal-ance this with meeting performance or quality-of-service requirements dictated by the application. Device operation in such a network can be modeled as a simple finite state machine, where av-erage power is the weighted average of the active- and sleep-phase power. The

active phase includes data acquisition, processing, and transmission, and the sleep phase includes the time spent in low-power modes and performing other tasks. Given a finite energy supply, it’s easy to estimate the lifetime of a sen-sor node by dividing its capacity rating by the average rate of consumption and taking into account self-discharge of the battery and environmental factors.

Such a model can be useful in de-termining when a sensor node might require recharging and could also inform the selection of suitable on-board storage at design time to satisfy an economically feasible maintenance schedule. However, a detailed model of the operational state machine and potential state space is first required to determine reasonable boundar-ies for the active phase. Factors such as position in the network affecting routing responsibilities, time-varying link qualities resulting in packet re-transmissions, and event detection or alarm threshold breaches make effectively modeling a device’s energy evolution a complex task. Modifying the energy model is also required for

Figure 2. Indicative power consumption of various components and systems. Note that the presented data is shown as a reference for evaluating overall power demand. The application objectives and performance levels of the systems presented here are very diverse, so a comparison of low-power electronics based on this graph should be avoided.

1.E + 02

1.E + 01

1.E + 00

1.E – 01

Activ

e po

wer

/W

1.E – 02 ADCAD74761 MSPS

Radio Atmel AT86, 3 dBmCC430, 0 dBm

Smartphone SoC

Drones

Avyon MD4 – 1,000drone (2.7 kg)

Parrot AR.Drone 2drone (0.4 kg)

CheersonCX–10 C drone

(15 g)

QualcommSnapdragon 800 SoC

(max)

Michigan, –10 dBmUCLA, 1 GSPS

MIT, 1 GSPSMIT, 3 dBm

1.E – 03

1.E – 04

1.E – 051995 2000 2005 2010

ARM cortex M0/MHzProcessing core

Microcontroller

T1 MSP430/MHz

Year2015 2020


EnErgy HarvESting

time-varying yet predictable ambient energy harvesting sources.4

Energy System DesignHarvesting is increasingly included in designs (such as EnOcean; www.en-ocean.com), and sensor network de-vices’ energy subsystems are becom-ing more tightly engineered relative to the application and its deployment environment. Legacy philosophies, wherein devices comprising sen-sor networks are thought of as being cheap and disposable, are being dis-lodged by those that suggest that the resultant data is sufficiently important to collect over long time periods. This requires sustainable operation beyond the capability of many contemporary battery packages, while the replace-ment of batteries continues to be seen as an unacceptable overhead, is po-tentially environmentally unfriendly at scale, and (in some cases) is impos-sible due to device placement in inac-cessible locations. Most WSN-type ap-plications aim to achieve a duty cycle below 1 percent, which is essential to enable a long-lived application. Fur-thermore, given the ultra-low current draw in modern components, in many cases, the limiting factor in determin-ing a node’s lifetime will be the bat-tery’s self-discharge rate.

Approaches to system-level co-de-sign are beginning to include energy as a primary element of the design space, although this is largely restricted to re-search projects. Thus, models of rea-sonably predictable ambient energy sources are under development and im-provement. These are typically based on energy traces collected in the real world and subsequently introduced to simulation and emulation tools. Such efforts have led to useful advances, including improved predictability for short- and medium-term energy avail-ability,15 ambient energy models inte-grated into contemporary simulators16 and emulators,17 and improved co-de-signs tailored for specific application scenarios.4

As part of the design process, the am-bient available energy from potential harvesting sources is determined using surveys or estimated based on available data, such as prevailing weather condi-tions. Figure 3 shows the power density per surface area or volume for various harvesting solutions proposed over the years. Besides the wide range of power levels, these sources also range widely in temporal variability and predictabil-ity. There is a pressing need for more tools that enable realistic and consis-tent experimentation in which renew-able and time-varying sources of energy

are part of the design space. Whether ambient sources will be adequate for a given application will depend on the average power, variability, and predict-ability of the source, with respect to the application’s power requirement and its tolerance to temporary loss of service for individual nodes.

incorporating Wireless Power transferFor scenarios in which harvesting, teth-ered charging, or battery replacement are infeasible, recent advances in wireless power transfer provide an attractive al-ternative. Tracing back to Nikola Tesla’s experiments in the early 20th century, the idea behind WPT is mature and has found significant application in contempo-rary consumer electronics in the form of electromagnetic inductive coupling, usu-ally over very short distances (cm range or less). This involves generating a magnetic field by applying alternating current to a primary coil on the transmitter side, which induces voltage across the terminals of a secondary coil on the receiver side—acting as a transformer with the core removed. In 2010, the Wireless Power Consortium (http://wirelesspowerconsortium.com) in-troduced the safe and simple Qi standard for wireless charging of devices, at very short range, at up to 5 W.

However, there is significant perfor-mance degradation in terms of energy transfer when the distance between transmitter and receiver exceeds ap-proximately one coil diameter, due to the rapid reduction in magnetic field strength as the receiver moves away from the transmitting coil (which is in the cm or mm range for typical electron-ics devices due to size and packaging constraints). Degradation also occurs with misalignment, though this can be solved using magnets to hold devices almost precisely in place. Charging electric toothbrushes, smartphones, watches, and so on in this way is practi-cal in most cases, but is less than ideal when almost direct and aligned contact is impossible. In WSN applications, for example, distributed devices might

Figure 3. Power levels for recent approaches to ambient environmental energy harvesting. Indicative values are taken from the literature; power density is given for one unit of area (cm2) or volume (cm3) in the y-axis.

Energy harvesting by source

Scavengeable energy sources

Pow

er d

ensi

ty (W

/(Are

avo

lum

e))

Hand

gene

rato

rHe

elst

rike

Outd

oor

light

Indo

orlig

ht

Vibr

atio

n 1

Vibr

atio

n 2

Ther

mal

1

Ther

mal

2

Ambi

entR

F

Airfl

ow

Acou

stic

(100

)Ac

oust

ic(7

5)

102

100

10–2

10–4

10–6

10–8

10–10


be inaccessible or arbitrarily oriented, making wireless transfer using simple inductive coupling impractical.

If the transmit and receive coils are operated at a frequency to maximize their Q-factors and thus minimize losses associated with the repeated energizing and discharging of the resonant tank components, reasonable efficiencies can be achieved at greater distances. André Kurs and his colleagues showed that a link efficiency of approximately 40 per-cent could be achieved over a distance of four times the coil diameter, which is in a region often called the “mid-range” region.13 The tuning out of the coil inductances with capacitors to both reduce the apparent power rating of the drive electronics and improve link effi-ciency has been long known, and most practical IPT systems have employed this for some time.18

Developing the principle, Kurs and his colleagues showed that for mid-range applications, it’s possible to si-multaneously power multiple devices when the devices have much smaller surface areas than the source. It has since been demonstrated over greater distances for powering wireless sen-sors using smaller receivers, achieving approximately 10 mW received at a 6 m standoff.19 We summarize the key fea-tures of IPT devices for different size scales in Table 1.

The benefits of IPT are numerous compared to other wireless energy transfer mechanisms, including short- to mid-range capability, high transfer efficiencies (difficult to achieve with ra-diative methods), and the ability to effi-ciently transfer power through a variety of materials. A greater understanding of the impact of external objects and

environments is required, where certain materials in proximity can shift reso-nant frequencies or reduce the Q-factor, ultimately affecting transfer efficiencies.

Health effects of magnetic and electric fields must be taken into con-sideration. Standards for exposure to electromagnetic waves are set by different bodies in different countries based on the same basic principles, intended to limit both tissue heating and muscle and nerve stimulation ef-fects. The World Health Organiza-tion set guidelines based on IEEE and International Commission on Non-Ionizing Radiation Protection (ICNIRP) recommendations. Abso-lute levels for ICNIRP and IEEE dif-fer but are both based on basic restric-tions: the maximum allowed in-body magnetic and electric field strengths as a function of frequency. Because these internal quantities are difficult to measure, a set of reference levels are described for outside-of-body measurements, which, if adhered to, allow basic restrictions to be met.

Typical specific absorption-rate limits are 10–20 W/kg, depending on where heating occurs, and less than 400 mW/kg across the entire body. Because the limit values are more gen-erous for low frequencies (in the 100 kHz range) than at higher frequencies used for IPT (for example, ISM bands at 6.78 and 13.56 MHz), high-power systems (for vehicle charging) tend to use lower frequencies than low power systems (which need to use higher fre-quencies to use small coils efficiently).

On-Board Storage SolutionsMost pervasive computing applications require local energy storage in a primary

capacity for battery-only cases or en-ergy buffering where harvesting or charging is possible. Lithium-ion bat-teries dominate in mobile electronics applications because of their high en-ergy density. In these applications, their relatively high cost and limited number of charge-discharge cycles are generally considered acceptable. However, as with other battery chemistries, they have low power densities, which can be a draw-back when the application has highly “peaky” power demand. Furthermore, this limits rapid recharging, which is needed for mobile power sources to wirelessly recharge multiple devices.

Supercapacitors provide an alterna-tive storage solution that looks increas-ingly attractive. Functionally occupy-ing a middle ground between batteries and conventional capacitors, they offer charging times of tens of seconds (see Table 2) but energy densities about 10 times less than lithium batteries (see Figure 4). As a long-term storage solu-tion, however, they fail because of their short self-discharge times of a few days. In a daily recharging scenario this is tolerable, but considering again the use of mobile power sources—such as for embedded sensors—what is really de-sirable is short charge times (a few min-utes or less), combined with long peri-ods (months) between recharges. There would appear to be no way of achiev-ing this with a single storage technology without dramatic improvements in sec-ondary battery design and fabrication.20

A hybrid solution, with supercapaci-tors buffering batteries, can offer this combination of quick charging and long operational life. Efficient energy-trans-fer circuits are available for this purpose; they drop dramatically in performance

TABLE 1 Key features of wireless power transfer devices for different size scales.

Charger size range

Required proximity

Required alignment Power range

Resonance sensibility

Safety requirements

2–10 mm < 10 mm ± 1 mm < 10 W ± 1% Packaging for safety

20–100 mm < 100 mm ± 10 mm < 10 W ± 1% Packaging for safety and user training

Ø > 100 mm < 1 m ± 100 mm < 100 W ± 1% Electromagnetic shielding, user training, and protective equipment


EnErgy HarvEsting

when the input (that is, the supercapaci-tor) voltage falls below a few hundred mV. However, if the battery voltage is, say, 3–4 V, then a residue of even 0.5 V in the capacitor represents only a small fraction of its energy. Furthermore, if a lifetime of several months or more is achieved between charges, the battery limitation of hundreds of charge-dis-charge cycles need not be a concern.

To decide whether the quick, infre-quent recharge scenario is practical, the next consideration is whether a sufficient power rate can be delivered. If we tar-get one five-minute charge per year, the ratio is 105, meaning that the charging power must be 105 times the device av-erage power consumption. If the average power is 0.1 mW, charge power needed is thus 10 W. For a wireless source of a size compatible with an autonomous

carrier, especially an airborne one, transfer of 10 W is probably near the upper limit. This suggests that a 10 W charging rate will require high transfer efficiency—that is, a proximity between charger and receiver of a few coil diam-eters or less.14 If this isn’t achievable—for example, if the wireless device is too deeply embedded in some structure, or the delivery vehicle can’t practically get close enough—rapid charging becomes infeasible. In such a case, buffering the battery with a supercapacitor probably won’t offer an advantage.

S tate-of-the-art sleep modes of electronics and pervasive com-puting devices are often below the self-discharging rates of

batteries and supercapacitors. A typical

lithium-ion battery self-discharging at 0.1 percent per day means that a 0.1C power system (C is the capacity rating of the battery specifying the maximum safe discharge rate), duty-cycled at less than 0.1 percent, loses one third of its energy to leakage. Therefore, the self-discharge rate of a storage medium can be consid-ered as a lower limit of beneficial duty cycling. Lithium primary batteries’ self-discharge rates are significantly less than those of rechargeable chemistries (1 per-cent of nominal capacity per annum is typical of off-the-shelf packages) and are still one of the best solutions for applica-tions that lend themselves to duty-cycled methods. The charging speed of superca-pacitors is limited by the state of the art of WPT systems, as opposed to the bat-tery case, where the limiting factor is the battery charging speed. Coupling these limitations with size, orientation, and proximity constraints, IPT isn’t likely to be a widespread solution to energy provi-sion in many application contexts.

The energy design space remains in-formal and subjective. There is no dis-cernible methodological approach, and there are insufficient analyses in the lit-erature concerning emerging battery and supercapacitor technologies’ use in contemporary applications beyond well-known solar-supercapacitor sys-tems. Nevertheless, with the increas-ing need for long-lived autonomous operation of devices and the emergence of new harvesters, deliberate transmis-sion, and improved storage technolo-gies, we argue that it’s time to develop a systematic approach to energy engi-neering for the next generation of per-vasive computing systems.

Figure 4. Indicative energy and power densities of popular commercially available battery and supercapacitor models.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.01 0.1 1 10

Ener

gy d

ensi

ty (k

J/g)

Power density (W/g)

Tecate PC5-5 Skeleton 3.2 kF

Yunasko 5.2 kF

MaxwellBCAP0310

PowerstreamLIR2450

Powerstream1 × 44 × 61 mm

Nokia BL-5C

LG BL-T9Li-Po

Parrot Ar.DroneLi-Po

Batteries Supercapacitors

TABLE 2 Features of 10 grams of state-of-the-art energy storage.

Feature Battery Supercapacitor

Charge time 1–3 hours 10–100 seconds

Capacity (kJ) 2–10 kJ 0.02–2 kJ

Self-discharge time to 50% 2 years 4–15 days

Powering an active 3 W SoC 5–30 minutes 6 seconds to 10 minutes

Powering an active 10 mW MCU 2–10 days 30 minutes to 2 days

Powering 10 μW of sleep mode Self-discharge dominated Self-discharge dominated


REFERENcEs 1. L. Xie et al., “Making Sensor Networks

Immortal: An Energy-Renewal Approach with Wireless Power Transfer,” IEEE/ACM Trans. Networking, vol. 20, no. 6, 2012, pp. 1748–1761.

2. C. Martínez, T. Richardson and P. Cam-poy, “Towards Autonomous Air-to-Air Refuelling for UAVs Using Visual Infor-mation,” Proc. IEEE Int’l Conf. Robot-ics and Automation (ICRA), 2013, pp. 5756–5762.

3. S. Ahn, N.P. Suh, and D.H. Cho, “Charg-ing up the Road,” IEEE Spectrum, vol. 50, no. 4, 2013, pp. 48–54.

4. M. Magno et al., “Extended Wireless Monitoring Through Intelligent Hybrid Energy Supply,” IEEE Trans. Industrial Electronics, vol. 61, no. 4, 2014, pp. 1871–1881.

5. J. Xiaofan, J. Polastre, and D. Culler, “Perpetual Environmentally Powered Sensor Networks,” Proc. Fourth Int’l Symp. Information Processing in Sensor Networks (IPSN), 2005, pp. 463–468.

6. T. Esram and P.L. Chapman, “Com-parison of Photovoltaic Array Maximum Power Point Tracking Techniques,” IEEE Trans. Energy Conversion, vol. 22, no. 2, 2007, p. 439.

7. O.H. Stielau and G.A. Covic, “Design of Loosely Coupled Inductive Power Trans-fer Systems,” Proc. Int’l Conf. Power System Technology (PowerCon), vol. 1, 2000, pp. 85–90.

8. E. Ackerman, “Wireless Power Takes Charge,” IEEE Spectrum, vol. 53, no. 1, 2016, pp. 13–14.

9. A. Denisov and E. Yeatman, “Ultrasonic vs. Inductive Power Delivery for Minia-ture Biomedical Implants,” Proc. 2010 Int’l Conf. Body Sensor Networks (BSN), 2010, pp. 84–89.

10. M.E. Kiziroglou et al., “Acoustic Energy Transmission in Cast Iron Pipelines,” J. Phys-ics: Conference Series, vol. 660, 2015, article no. 012095; http://iopscience.iop.org/arti-cle/10.1088/1742-6596/660/1/012095/pdf.

11. M. Pinuela, P.D. Mitcheson, and S. Lucyszyn, “Ambient RF Energy Harvest-ing in Urban and Semi-Urban Environ-ments,” IEEE Trans. Microwave Theory and Techniques, vol. 61, no. 7, 2013, pp. 2715–2726.

12. M.E. Kiziroglou et al., “Design and Fab-rication of Heat Storage Thermoelectric Harvesting Devices,” IEEE Trans. Indus-trial Electronics, vol. 61, no. 1, 2014, pp. 302–309.

13. A. Kurs et al., “Wireless Power Transfer via Strongly Coupled Magnetic Reso-nances,” Science, vol. 317, no. 5834, 2007, pp. 83–86.

14. A.P. Sample, D.A. Meyer, and J.R. Smith, “Analysis, Experimental Results, and Range Adaptation of Magnetically Coupled Resonators for Wireless Power Transfer,” IEEE Trans. Industrial Elec-tronics, vol. 58, no. 2, 2011, pp. 544–554.

15. A. Cammarano, C. Petrioli, and D. Spenza, “Pro-Energy: A Novel Energy Prediction Model for Solar and Wind Energy-Harvesting Wireless Sensor Net-works,” IEEE 9th Int’l Conf. Mobile Adhoc and Sensor Systems (MASS), 2012, pp. 75–83.

16. R. Dall’Ora et al., “SensEH: From Simu-lation to Deployment of Energy Harvest-ing Wireless Sensor Networks,” IEEE 39th Conf. Local Computer Networks Workshops (LCN Workshops), 2014, pp. 566–573.

17. J. Hester, T. Scott, and J. Sorber, “Ekho: Realistic and Repeatable Experimenta-tion for Tiny Energy-Harvesting Sen-sors,” Proc. 12th ACM Conf. Embedded

Network Sensor Systems (SenSys), 2014, pp. 1–15.

18. J.T. Boys, G.A. Covic, and A.W. Green, “Stability and Control of Inductively Cou-pled Power Transfer Systems,” IEE Proc. Electric Power Applications, vol. 147, no. 1, IET, 2000, pp. 37–43.

19. C.H. Kwan et al., “Position-Insen-sitive Long Range Inductive Power Transfer,” J. Physics: Conference Series, vol. 557, no. 1, 2015, article no. 012053; http://iopscience.iop.org/arti-cle/10.1088/1742-6596/557/1/012053/pdf.

20. B. Kim et al., “Layer-by-layer Fully Printed Zn-Mno2 Batteries with Improved Inter-nal Resistance and Cycle Life,” J. Physics: Conf. Series, vol. 660, no. 1, 2015, article no. 012009.

Selected cS articles and columns are also available for free at http://computingNow.computer.org.

the AuThoRs

David E. Boyle is a research fellow in sensor networks with the department of electrical and electronic engineering at Imperial college London. His research interests include the design of pervasive computing applications, spanning var-ious topics in sensor networks, cyberphysical systems, and Internet of things technologies with industrial application. boyle received a Phd in electronic and computer engineering from the University of Limerick, Ireland. He is a member of IEEE. contact him at [email protected].

Michail E. Kiziroglou is a research associate with the optical and Semicon-ductor devices Group of Imperial college London, a lecturer in the department of automation engineering at the Alexander technological Educational Institute of thessaloniki, Greece, and a 2016 associate project scientist in the department of mechanical engineering at the University of california, berkeley. His research interests include energy harvesting devices, microengineering, and energy au-tonomous wireless sensors. Kiziroglou has a Phd in microelectronics and spin-tronics from the University of Southampton. He is a senior member of IEEE and a member of the Institute of Physics. contact him at [email protected].

Paul D. Mitcheson is a professor of electrical energy conversion with the control and Power research Group, Imperial college London. His research in-terests include using energy harvesting, power electronics, and wireless power transfer to provide power to applications when batteries and cables aren’t suit-able. mitcheson received a Phd in electrical and electronic engineering from Imperial college London. He is a senior member of IEEE, a Fellow of the Higher Education Academy, and is on the executive committee of the UK Power Elec-tronics centre. contact him at [email protected].

Eric M. Yeatman is a professor of micro-engineering and head of the electri-cal and electronic engineering department at Imperial college London. His research interests include motion and thermal energy harvesting for wireless devices, pervasive sensing, and sensor networks. Yeatman received a Phd in electrical engineering from Imperial college London. He is a Fellow and Silver medalist of the royal Academy of Engineering, and is a cofounder and director of microsaic Systems, which develops and markets miniature mass spectrom-eters for portable chemical analysis. contact him at [email protected].

energy provision and storage for pervasive computing

Documents