10.07.2015 Views

Realizing RFID Sensor Networks with the Intel WISP

Realizing RFID Sensor Networks with the Intel WISP

Realizing RFID Sensor Networks with the Intel WISP

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Realizing</strong> <strong>RFID</strong> <strong>Sensor</strong> <strong>Networks</strong><strong>with</strong> <strong>the</strong> <strong>Intel</strong> <strong>WISP</strong>Joshua Smith, Alanson Sample, Dan Yeager,Michael Buettner, Ben Greenstein, Polly Powledge,Richa Prasad, and David We<strong>the</strong>rall<strong>Intel</strong> Research & University of WashingtonCommercial (EPC Gen2) <strong>RFID</strong> Tag<strong>Intel</strong> (EPC Gen2) <strong>WISP</strong>


CreditsCore <strong>WISP</strong>:• Josh Smith, Alanson Sample, Dan Yeager, Polly Powledge<strong>RFID</strong> sensor networks:• Michael Buettner, Ben Greenstein, Richa Prasad, David We<strong>the</strong>rall2djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


<strong>Intel</strong> Research SeattleAn exploratory research lab in <strong>the</strong> <strong>Intel</strong>Research network of labs– Roughly 20 people + students/visitorsFocus on computing systems that arewoven into <strong>the</strong> fabric of everyday life– Interdisciplinary projects, prototypes ofdevices/systems, user-centered view– “Vertical” expertise from sensing tomobile systems to ML to HCIDeep ties to universities and <strong>Intel</strong> groups– Open Collaborative Research model3djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


4University of WashingtonComputer Science & EngineeringConsistently ranked among <strong>the</strong> top 10 programs in <strong>the</strong> nationExcellent breadth and depth in <strong>the</strong> field• Embedded systems; systems & networking; security & privacy; animation & games;computing & biology; intelligent internet systems; computer graphics & vision; AI &robotics; technology in education; programming systems; data management;software engineering; computer architecture; <strong>the</strong>ory …A terrific environment for students and faculty• UW is a premier research university• Seattle is a great place to live – and to docomputer science (UW, Microsoft Research,<strong>Intel</strong> Research, Adobe Research, Google,Amazon.com, Real<strong>Networks</strong>, …)• Top academic programs and labs are filled<strong>with</strong> UW CSE alumsCome join us!djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


<strong>RFID</strong> <strong>Sensor</strong> <strong>Networks</strong> (RSNs)Argue <strong>the</strong>y can combine <strong>the</strong> best of <strong>RFID</strong> and sensor networks :• Enable new applications of physically embedded sensing• Retain <strong>the</strong> simplicity of <strong>the</strong> <strong>RFID</strong> model and ride its deployment• Retain <strong>the</strong> sensing/computing potential of wireless sensor networksResearch challenges:• Feasibility• Applications• OS issues / Intermittent power• Network issues / Asymmetric protocols5djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Wireless <strong>Sensor</strong> Network (WSN) – Vision• <strong>Sensor</strong>s on items in everyday life to monitor and enable smartenvironmentsHabitatsHospitalHome•Bird migration•Temperature changes•Patient movement•Vitals, medicine, etc.•Find lost objects•HVAC control, etc.6djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Radio Frequency IDentification (<strong>RFID</strong>) – Vision• Many kinds of <strong>RFID</strong>, mostly driven by business needs aroundinventory and supply chain managementActive <strong>RFID</strong>:100 ft, $10sPassive HF:10 cm, < 1¢Passive UHF:30 ft, < 5¢7djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Wireless <strong>Sensor</strong> <strong>Networks</strong> – Key FeaturesTelosB mote(sensing + code)richdata• Collection of small, battery-powered sensing devices (motes)• Peer-to-peer communication (multi-hop network)• But form-factor limits for embedding sensing, and tend to bepeople-intensive and complex8djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


<strong>RFID</strong> + WSN = <strong>RFID</strong> <strong>Sensor</strong> <strong>Networks</strong>EPC Gen2readerapplicationcoderich data<strong>WISP</strong>s(sensing + code)• Combine <strong>the</strong> best of both:– Small / inexpensive passive tags that sense/compute (<strong>WISP</strong>s) andcommunicate directly <strong>with</strong> apps in infrastructure (via readers)– Go beyond efforts that are battery-assisted or sense “physically”• Key advantages:– Ubiquitous, long-lived instrumentation, business simplicity, flexibility10djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


ChallengesFeasibility– Can <strong>WISP</strong> devices be built?Applications– Can RSNs do anything useful?OS issues– How do we program <strong>with</strong> intermittent power?Network issues– How do we communicate <strong>with</strong> asymmetric protocols?11djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008Focusto dateNot yetexplored!


Challenge: Feasibility<strong>Intel</strong> prototype shows passive yet programmable <strong>WISP</strong>s can be built– [Sample et. al.], IEEE Trans. on Instrumentation & Measurement<strong>Intel</strong> <strong>WISP</strong> ~2008<strong>WISP</strong> <strong>with</strong> standard reader12djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Evolution of <strong>the</strong> <strong>Intel</strong> <strong>WISP</strong> PlatformCapabilities willimprove <strong>with</strong>Moore’s law:• smaller devices• lower energy• longer range• more function13djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


<strong>WISP</strong> features ~2008/9Features:• Ultra-low power microcontroller(MSP430)• Power harvesting circuitry <strong>with</strong> storagecapacitor (to run away from reader)• 32K program space + 8K storage• <strong>WISP</strong> to reader backscatter and readerto <strong>WISP</strong> data communication (ASK)• <strong>Sensor</strong>s: 3D-accelerometer, light,temperature• Real-time clock• Works <strong>with</strong> EPC Gen 2 readers• Range of up to 10ft14djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008Mature discreteimplementation


Flexible <strong>RFID</strong> reader for applications• Can control commercial Gen2 readers to an increasing extent (LLRP)which supports some but not all applications• Software-defined radio (USRP) reader allows for low-levelprogrammability and complements <strong>the</strong> <strong>WISP</strong>– Also valuable for <strong>RFID</strong> measurementUSRPEPC Gen 2protocolCommercial tagGNU Radio + LinuxNew <strong>RFID</strong>protocol<strong>Intel</strong> <strong>WISP</strong>15djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Commercial readers – “tags per second”HS mode16 tags8 tags4 tags2 tags1 tagTags read per secondTag Reads/Sec0 50 100 150 2002 4 6 8 10 12 14Distance (ft)Distance (ft)• Common metric of gross tags/second is uninformative– Larger tag set amortizes overhead– Some tags may be read multiple times or no times16djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Commercial readers – Tag SetsCost of lower data ratesTag sets Tag read Sets/Sec per second0 2 4 6 8 102 4 6 8 10 12 14 16 18 20Distance (ft)Distance (ft)16 Tags• Physical layer effects matter• Encoding and bit errors, frequency-selective fading• [Buettner, We<strong>the</strong>rall] in Mobicom 2008HSDR17djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Challenge: ApplicationsRSNs enable new applications that leverage <strong>WISP</strong> features:• Passive long-lived, can be installed in inaccessible locations for<strong>the</strong> duration, “infrastructure <strong>with</strong>out wires”• Small and inexpensive can be used ubiquitously/widelyExample applications that stretch WSNs follow …18djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


“Milk”Cold chain monitoring:• Track many small items, e.g., bags ofblood, items in fridge, <strong>with</strong> cheapand/or disposable tag• Use proximity of readers to sensetemperature and vital statistics whenrefrigerated most of <strong>the</strong> time• Use stored energy (super-cap) tosense for brief periods when items areaway from reader and exposed19djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008Milk carton <strong>with</strong> <strong>WISP</strong> andtemp and fullness sensors[Yeager, Prasad, Powledge,Smith, We<strong>the</strong>rall], <strong>RFID</strong> 2008


“Planes”Embedded sensing <strong>with</strong>out wiresor access, e.g., for planes,temperature and pressure intires, o<strong>the</strong>r infrastructure.<strong>WISP</strong> <strong>with</strong> strain gageembedded in compositematerial20djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


“Brains”For non-intrusive physiological sensing in implantable medical devices,e.g., neural pros<strong>the</strong>tics, pacemakersNeural <strong>WISP</strong> [Hollman,Yeager, Prasad, Smith,Otis], BioCAS 200821djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


“Elders”Instrument environment to sense details of object use, e.g., to help<strong>with</strong> <strong>the</strong> long-term care of elders• Tags plus acceleration is fine-grained and beats identifiers aloneTLC project: wearable(HF) <strong>RFID</strong> reader[Philipose and o<strong>the</strong>rs] TLC project: Tagged objects and home setting22djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Challenge: OS issues / Intermittent PowerHow do we run a range of programson <strong>WISP</strong>s <strong>with</strong> intermittent power?<strong>WISP</strong> power model:• Ga<strong>the</strong>rs energy from readers,when in range at unpredictablerate and times• Has limited on board storage(capacitor


Strategy for intermittent power (1)• Duty cycle <strong>WISP</strong> task:– Wait until sufficient power is collected (preset threshold)– Run <strong>WISP</strong> task to completion; repeat• Implications– Variable frequency of operation; <strong>WISP</strong> responds occasionallyvoltagecharge runtime24djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


O<strong>the</strong>r strategies• Divide large computation into multiple small stages– Store state in non-volatile memory or on readers• <strong>WISP</strong> API to expose power state to application code– “Enough power yet to complete this step?”• Coordinate to match reader power to <strong>WISP</strong> needs– Continue to provide for a sufficient interval• Looking to refine model of “power tolerant tasking”25djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Challenge: Network issues / AsymmetryEPC Gen2readerReader incontrol<strong>WISP</strong>s onlyhear readerrich data <strong>WISP</strong>sHow do <strong>WISP</strong>s efficiently communicate data in an “<strong>RFID</strong>” setting?<strong>WISP</strong> communication model differs from <strong>RFID</strong> and WSN:• <strong>RFID</strong> protocol optimized for reading identifiers (slotted Aloha)• Readers control each step and <strong>WISP</strong>s only hear reader26djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Strategy for collecting sensor data (1)• Embed sensor data in loworder bits of tag identifiers• OK for small and sparseamounts of datareader tagQueryRN16• What about richer data?ACKID+dataQueryAcktime27djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Strategies for collecting sensor data (2)• Overload semantics of READcommand to return (changing)sensor data at known indices• Allows larger blocks to be read,but requires singulationreader tagQueryRN16ACK• Still not a good fit to <strong>the</strong> sameset of tags returning updatesIDReq_RNRN16timeReadx NDataQueryAck28djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


O<strong>the</strong>r strategies• Extend to “RSN protocols” for reading sensor data– Use short handles to <strong>WISP</strong>s ra<strong>the</strong>r than restart identification• Select sensors by value– Let <strong>WISP</strong>s respond only if <strong>the</strong>y have important data– Bias responses to favor <strong>WISP</strong>s <strong>with</strong> more important data• Yield reader control– Let <strong>WISP</strong>s drive interaction depending on <strong>the</strong>ir needs• May be able to approximate <strong>with</strong>in EPC Gen2– Coexist well <strong>with</strong> vanilla <strong>RFID</strong> tags and readers29djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Conclusion / Call to action<strong>RFID</strong> sensor networks:• Aim to combine <strong>the</strong> best of <strong>RFID</strong> and WSN• Ubiquitous, long-lived instrumentation,simplicity (reader/tag split), and flexibility(applications)<strong>WISP</strong> Research Challenge:• Give research community access to <strong>WISP</strong>s• Develop RSN systems & applications• Plus: <strong>RFID</strong> security, <strong>RFID</strong> measurement,educational projects30djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008


Thank you. Questions?Some <strong>WISP</strong> papers:• “Design of an <strong>RFID</strong>-Based Battery-Free Programmable Sensing Platform,” A.Sample, D. Yeager, P. Powledge, A. Mamishev, and J. Smith, IEEETransactions on Instrumentation and Measurement (accepted).• “Wirelessly-Charged UHF Tags for <strong>Sensor</strong> Data Collection,” D. Yeager, R.Prasad, D. We<strong>the</strong>rall, P. Powledge, and J. Smith, IEEE <strong>RFID</strong> 2008.• “Maximalist cryptography and computation on <strong>the</strong> <strong>WISP</strong> UHF <strong>RFID</strong> tag,” H.Chae, D. Yeager, J. Smith, and K. Fu, <strong>RFID</strong> Security, July 2007.• "Neural <strong>WISP</strong>: An Energy Harvesting Wireless Brain Interface <strong>with</strong> 1mRange", J. Holleman, D. Yeager, R. Prasad, J. Smith, and B. Otis, IEEEBiological Circuits and Systems 2008 (accepted).31djw \\ <strong>RFID</strong> sensor networks \\ 1 Nov 2008

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!