1 / 61

Deeply Embedded Networks for Habitat Monitoring

This article reviews different network architectures for habitat or environmental monitoring, including multi-levels, extent, duration, and energy sources. It discusses the role of mobility and adaptation and the involvement of distributed control. The insight from description vs empirical evaluation is also examined.

rpetersen
Download Presentation

Deeply Embedded Networks for Habitat Monitoring

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. CS294-1 Deeply Embedded NetworksPlatformsSept 2, 2003 David Culler Fall 2003 University of California, Berkeley

  2. Review of last time • 3 different network architectures for habitat or environmental monitoring • Multi-levels • Extent, duration, energy sources • Role of mobility and adaptation • Where was distributed control involved? • Insight from description vs empirical evaluation? CS294-1 F03

  3. Reading Materials • G. J. Pottie and W. J. Kaiser, Wireless Integrated Network Sensors , CACM, 43(5), May 2000. • Mica: A Wireless Platform for Deeply Embedded Networks, Jason Hill and David Culler, IEEE Micro., vol 22(6), Nov/Dec 2002. • L. Doherty, B.A. Warneke, B.E. Boser, and K.S.J. Pister, "Energy and Performance Considerations for Smart Dust." International Journal of Parallel Distributed Systems and Networks, Volume 4, Number 3, 2001. • The PicoRadio Test Bed, Burghardt, Mellors, Rabaey • PicoRadio Supports Ad Hoc Ultra-Low Power Wireless Networking, Rabey, Ammer, da Silva, Patel, and Roundy CS294-1 F03

  4. Outline • Touch of history • 3 Platform Perspectives • WINS – estimation theory • SmartDust – analog design, chip technology • Mica - systems • WINS vision • bi-part architecture for passive vigilance • PicoRadio • Incorporate programmable logic for protocol processing, as well as signal processing • Smart Dust • Really serious about energy limits • Simple, un-partitioned architecture • Berkeley TinyOS motes take off • Mica • accelerate primitives not solutions • Rich interfaces and flexible composition for cross-layer optimization CS294-1 F03

  5. Early History • 1966 Igloo White • http://home.att.net/~c.jeppeson/igloo_white.html • 1986 DARPA packet Radio program Name  Model  Variant   Type  Length  Weight ADSID I (N)- - - - - -  normal   seismic  31.00 ins    26.0 lbs ADSID I (S)MA-36  short   seismic   20.10 ins    13.7 lbs ACOUSID II TC-415     seismic-acoustic   53.14 ins    38.8 lbs ACOUSID III MA-31    seismic- acoustic   47.63 ins .  37.2 lbs ADSID III (N)MA-33  normal   seismic  37.66 ins .  37.2 lbs ADSID III (S)MA-37  short  acoustic  20.10 ins .  13.7 lbs MODS81 mm  mortar  seismic  33.00 ins .  9.6 lbs CS294-1 F03

  6. Classic Wireless Architecture • Partitioned system with narrow, std interfaces • Similar picture to 802.11 NIC • uP are poor at driving radios directly • Over sampling, start symbol, … • Partitioning driven by many factors • Local optimization • Standardization • IP protection • Works best in well-established areas • Encumbers deep innovation • Ex: ppp for data on cell CS294-1 F03

  7. Emergences of WINS • 1994 Pottie and Kaiser propose Low Power Wireless Integrated Microsensor • LWIM nodes built around 1996 • DARPA Sensit Program • Late 97-98 handhelds emerge • palm • ITSY, BWRC PicoRadio, Srivastava UCLA, Chandrakasan MIT, … • Matchbox PCs • Bluetooth promised • Berkeley SmartDust • 1999 WeC mote offshoot • SCADDS (USC/UCLA) pc104s & tags • 00 Mote / TinyOS platforms • WINS ng finally appears in Linux for Sensit • 02 Mica NEST OEP creates de facto platform • 03 Bluetooth revival CS294-1 F03

  8. WINS case for distributed sensor nets • must distribute to detect reliably regardless of $ • All signals decay with distance (r^2) + absorption, scattering, dispersion, …even with line of sight • Often need to track multiple objects • Obstructions, clustering • Detection and estimation theory • observables {Xj} – sample outputs of sensors • target signal plus background noise & interference • features {fk} – reduced representation of observations • Fourier, LPC, wavelet coefficients • hypotheses {hi} – presence/absence based on estimates of feature set • Choose hi if P(hi | {fk} ) > P(hj | {fk} ) for j != I • Reliability: number of independent observations and SNR • Complexity: dimensions of feature space, # hypotheses => More observations, rather than more processing per observation => Short range means better SNR => Fewer targets (hypoth’s) in range of set of sensors => Nearly homogeneous over small regions CS294-1 F03

  9. Radio propagation • Energy required to transmit distance d • Et = bdn • n is about 2 in freespace, about 4 near ground • Indoor has lots of other complications • Small energy => short range + Allows spatial multiplexing • Multihop routing required to achieve distance • Energy per hop is more + routes around obstacles • Requires discovery, topology formation, maintenance • may dominate cost of communication • Requires media access control • Time, space, frequency, … • Energy to receive ~ Et at short range • Dominated by listening time (potential receive) • Radio must be OFF most of the time! • WINS asserts diversity through spreading & coding CS294-1 F03

  10. Passive Vigilance Parts of the system must be sampling environment all the time • Reliable detection costs too much energy • use low-cost, low resolution techniques to detect potential event • Bring in more powerful, more costly options (infrequently) when important Example: seismic sensor triggering camera Processing hierarchy => Introduces processing, storage, and communication issues CS294-1 F03

  11. WINS node arch • Two-part architecture • Combine network performance info, synch, BW reservations, into data messages • Morphed into Sensoria automotive telematics • www.sensoria.com • Observe variant on classical partitioning CS294-1 F03

  12. PicoRadio Concept Arch • Embedded proc. for appln and protocol stack • High complexity, low frequency • Reconfigurable logic between uP and simple radio front end • Powerful data processing (SW radio) on the side • Potentially rich interfaces through FPGA • Propose 3-stage development • 2000: COTS, 2002:chip, 2004:better chip CS294-1 F03

  13. PicoRadio I implementation (2002) • ARM 1100 proc, 4 MB • Bluetooth front-end (sans baseband controller) • 130 – 300 mw @ 60 MHz, 430 – 700 @ 200Mhz • day(s) operation on set of cell phone batteries CS294-1 F03

  14. Smart Dust • Focused even more sharply on energy CS294-1 F03

  15. Resolution & range sensitivity env ADC proc transducer amp Ex: acc 1 V to 16 bits => 15 uV resolution mV/G Energy Picture • Li ion battery 2 J / mm2, Ultracap: 1 mJ/ mm2 • Solar 1 mW/mm2 (~1 J/day/ mm2) x 30% • 1 uW/ mm2 indoors • Vibration - 10 uW/g (2 mg in a mm3 of silicon) • 1000x energy gap bottom-up analysis vs uP • 10 – 20 pJ/inst (16 bit) • Lots of circuit tricks (ripple carry, bus design, address reuse) • Leakage dominated by memory • ADC Energy increases fiercely with resolution CS294-1 F03

  16. Communication • Radio • P Rmin > kTB Nf SNRmin • Pr = Pt Gant / 16 p2(d/l)n • 1 uJ/bit for 100m at 1 kbps ground-to-ground • 0.1 pJ/bit in free space • Alternatives? • Directional • Optical • Powered by larger transmitters • Optical LOS Pr = Pt Gant Arx/ 4 pd2, Gant ~ 4p/ q2 CS294-1 F03

  17. 4Mhz, 8bit MCU (ATMEL) 512 bytes RAM, 8K ROM 900Mhz Radio (RF Monolithics) 10-100 ft. range Temperature Sensor Light Sensor LED outputs Serial Port COTS Mote (Seth Hollar) COTS Dust weC Mote All physical subsystems “exposed” to processor => fine-grain multithreading in SW to coordinate flows. CS294-1 F03

  18. System Characteristics of WSNs • Small physical size and low power consumption => Limited Physical Parallelism and Controller Hierarchy => primitive direct-to-device interface • Concurrency-intensive operation • flow-thru, not wait-command-respond => must handle multiple inputs and outputs simultaneously • Diverse in Design and Usage • application specific, not general purpose • huge device variation => efficient modularity => migration across HW/SW boundary • Largely Unattended & Numerous => robust operation => narrow SW interfaces (but highly application specific) CS294-1 F03

  19. Rene: Second Generation ‘Mote’ • A tool for experimentation • 3 Board Sandwich. • Main CPU boardwith Radio Communication. • Sensor Board • Power Board • Rich interconnection • Allows for expansion andcustomization. • Many sensor boards:Acceleration, Magnetic Field,Temperature, Pressure,Humidity, Light, and RF Signal Strength. • Can control RF transmission strength & Sense Reception Strength. CS294-1 F03

  20. Rene @ Sensit • UAV drops 10 nodes along road, • hot-water pipe insulation for package • Nodes self-configure into linear network • Synchronize (to 1/32 s) • Calibrate magnetometers • Each detects passing vehicle • Share filtered sensor data with 5 neighbors • Each calculates estimated direction & velocity • Share results • As plane passes by, • joins network • upload as much of missing dataset as possible from each node when in range • 7.5 KB of code! • While servicing the radio in SW every 50 us! 29 palms, 3/01 CS294-1 F03

  21. Services Networking TinyOS www.tinyos.net Rene 11/00 Mica 1/02 Dot 9/01 Demonstrate scale - Intel • Designed for experimentation • sensor boards • power boards • DARPA SENSIT, Expeditions NEST open exp. platform 128 KB code, 4 KB data 50 KB radio 512 KB Flash comm accelerators - DARPA NEST Mote Platform and Community WeC 99 “Smart Rock” Small microcontroller - 8 kb code, 512 B data Simple, low-power radio - 10 kb EEPROM storage (32 KB) Simple sensors Crossbow CS294-1 F03

  22. Panasonic CR2354 560 mAh Rene Power Breakdown • But what does this mean? • Lithium Battery runs for 35 hours at peak load and years at minimum load! • That’s two orders of magnitude difference! • A one byte transmission uses the same power as approx 11000 instructions. CS294-1 F03

  23. Panasonic CR2354 560 mAh System tradeoffs: simple radio, fast on/off, application behavior CS294-1 F03

  24. Mote Expansion Connector (v1) • Documented hardware interface • Swap components on either side of the connector while preserving investment in sensors or main boards • Sensor interfaces • 4 lines dedicated to switching components on and off • 7 analog voltage sensing lines • 2 I2C busses • SPI • UART lines • Debugging aids • All radio-related signals: RX, TX, base band, control signals, signal strength • Programming interfaces • SPI and reset signals for the main processor and the coprocessor • Ground, Vcc for both analog and digital circuits • 12 lines reserved for future use CS294-1 F03

  25. Rene Mote Connector Schematics CS294-1 F03

  26. Power Lines • Need to control on/off state of individual sensors • Independently switched, used as outputs • Capability • Sink up to 20 mA, source a bit less • If more current is required by a sensor circuit, use MOSFETs • No higher level protocol attached • The place to implement functionality not provided by standard interfaces CS294-1 F03

  27. Analog Lines and AD Conversion • Many sensors provide an analog interface • Simple voltage dividers with photo resistors, thermistors, etc. • Whetstone bridges, condenser microphones, etc. • Need analog voltage sensing lines • 10 bit ADC,  2 LSB=> 8 usable bits •  3 mV error • Rail-to-rail range • Sampling rate • Max 15.4 ksps in continuous sampling mode • Max 4 ksps in a single sampling mode • 8 multiplexed channels • One dedicated to sampling BBOUT • Sampling rate high enough for most environmental phenomena • Interrupt driven, or polling version CS294-1 F03

  28. I2C Bus • 2-wire serial bus: clock and data • Clock is an “or” of all clock generators, the slowest clock generator dictates the speed • Bi-directional data line • Higher level protocol than UART or SPI • Defines a protocol for multiple device access, up to 128 devices per bus • Defines a multiple master arbitration • Allows for multi-byte transactions • Speed: up to 400 kbps • Software and HW implementations • Soft timing constraints mean that the use of timers is not mandatory, use tasks instead • Many slave devices available • EEPROM used for logging and other permanent storage • IO extenders, ADC converters, sensors, etc. • other significant chip-chip interconnect is 1-wire CS294-1 F03

  29. SPI • 3-wire, full duplex serial bus: clock, MISO, MOSI • Connector bus defines 2 SPI busses • Speed: up to 1 mbps • Hardware support on the main processor, software implementation on the coprocessor • Low-voltage programming interface to ATMEL microcontrollers • Interaction with coprocessor • Cyclic 16 bit distributed register • Buffering within the SPI critical • TI MSP provides double buffering on TX => 2X radio performance CS294-1 F03

  30. UART • A standard way for exchanging information with a PC • Voltage conversion and connector required, provided by the programming board • Available speed: 2400 bps to 115 kbps • Most reliable for communication with PC: 19.2 kbps • Large clock rate errors at high sending rates • 1 byte FIFOs • Hardware doesn’t buffer multiple bytes of either input or output • Operate in either polling or interrupt driven mode • TinyOS uses UARTs in interrupt driven mode CS294-1 F03

  31. What is the sensor-interface of the future? CS294-1 F03

  32. AVR Processor Core • 8-bit “RISC” at 4, 8 or 16 MHz • Memory • 8 Kbytes of program memory (flash) • 512 bytes of data RAM • 512 bytes of EEPROM on chip (write: 4 ms/byte) • 32 8 bit registers • IO capabilities • 32 general purpose IO lines • Some lines also serve more specialized purposes, e.g. UART • 10-bit 8-channel ADC • Connector interface means that the IO lines serve a more specific purposes • Interrupts • No external interrupts available • No interrupt queuing • More recent ATMEGAs • Faster, more memory, better peripheral support • ARM thumb and other ARMs have little peripheral support CS294-1 F03

  33. Clocks and Timers a key resource • Clock sources • Internal high-speed resonator, low precision • External low-speed, high-precision oscillator • Only thing running when sleeping • External high-speed, high-preciosn oscillator • Counters • 3 – 4 hardware counters • 8 or 16 bits • Scale & Interval • Compare, interrupt, periodic • Uses • General free running clock • Wake-up (t0 is external 32 KHz) • Radio timing (mica t1 16-bit free running for bit capture, t2 8-bit for general op) • Critrical with RFM, not with ChipCon • Sensor sampling • Application scheduling (t1 free running) • Inherent tension between generality and precision CS294-1 F03

  34. Programming the Processor • Program the runtime memory image into a flash • No loader, no dynamic linker • Programming steps • Extract the code and data segments from an executable into an SREC file • Erase the current contents of the flash • Reset the processor • Download the program • Serial download protocol • Each download command contains the address, and the contents of to be stored • 10 ms delay per byte of code • Verify the program: read the flash, match it against the source image • After reset, start executing C runtime • rewrite the flash on the fly??? • Present in some uP (not Rene or Mica, yes Mica2 and dot) • Remapping? Protection? • Network Programming? CS294-1 F03

  35. Coprocessor Circuit • Coprocessor: ATMEL AVR AT90S2343 • 2 Kbytes of code space • 1 MHz internal clock • Required connections • SPI for programming the main microcontroller • RESET for the main controller • I2C for accessing EEPROM storage • Insufficient pins – I2C and SPI clock lines are shared • Minimal capability • Software SPI • Software I2C CS294-1 F03

  36. Radio Circuit • RFM Monolithics TR1000 916 MHz radio • On/off keying at 10 kbps (max. 19.2 kbps) • Capable of 115 kbps using amplitude shift keying • Capable of turning on in 30 us • Low-power listening by switching on and off on a sub-bit granularity • Processor interface • RFM CNTRL 0 and 1 – switch between transmit, receive, and sleep • Raw, unbuffered access to transmit (RFM TX) and receive (RFM RX) • Requires DC balanced signal – an equal number of 1’s and 0’s in the signal • Sampling on reception and modulating on transmission done is software • Too much noise in received signal to use UART for sampling • too little flexibility offered by UART • Imposes real time constraints on the system • Power usage • Transmit current: 7 mA • Receive current: 4.5 mA • Sleep: 5 uA CS294-1 F03

  37. Radio Signal Strength • Measuring the quality of reception from a node is useful • Aid in multihop routing decisions • Help with location estimation • Signal strength measurement • Based on base band sampling • Demodulated and amplified analog signal, fed into ADC channel 0 • Raw samples further processed in TinyOS • Empirical evaluation* *Klemmer, Waterson, Whitehouse, 2000 CS294-1 F03

  38. Radio Transmission Strength Control • Useful in networking experiments • Control the cell size • Allow for power savings • Principles of operation • Radiated power is proportional to the square of the current on RFM TX pin • Signal strength control • Controlled through a digital potentiometer (DS 1804), 0-50 kOhms • Currently no observable effects on power usage • Evaluation • Optimal transmission at the pot setting of 10 kOhms • Changing the range from 45 to 90 feet in an indoor environment • Changing the transmit strength has no effect on power usage • Effects secondary to shape/length of the antenna CS294-1 F03

  39. LEDs • Basic mote UI and a great debugging tool • Power consumption: 6mA • LED consumes as much power as the main microcontroller • If power is a concern, turn them off! • LED 1 and LED 2 multiplexed with the signal strength potentiometer control • Use them for controlling any other DS 1804 pots, one at a time • Many groups have built add-on segmented displays CS294-1 F03

  40. narrow standardized intrerface rich physical interface New Architectures? Embedded Network Arch. Typical Wireless Arch. (cellphone) • Traditional approach is to partition design into specialized subsystems with rigid interfaces. • TinyOS allows low and high-level processing to be interleaved. • rich physical information can be exposed • specialized hardware to accelerate primitives • Enables cross-layer optimizations audio kbd / display Sensor / Actuators Codec Application Controller CoProc Multi-Purpose Controller DSP protocol accelerators Protocol Processor RF Transceiver RF Transceiver CS294-1 F03

  41. Mica Architecture CS294-1 F03

  42. Basic Packet Processing • Key accelerators • Start symbol detection • Channel Synchronization • Bit Spooling • Potentials? • CRC?, encrypt/decrypt • Carrier sense? Collision detect? CS294-1 F03

  43. Cross Layer Optimizations • Low-power listening • Shift cost from frequent receive to infrequent send so listening is cheap • No compromise on latency and bandwidth • HW alternatives? • RF wake up • Treat radio as a sensor, rather than communication device • Escalating event detection • Time Synchronization • Source/recv tightly synchronized for comm to work • Minimize system induced variance • Eliminate uncertainty in transmit path • Ranging for Localization • Signal strength, time-of-flight CS294-1 F03

  44. Example Application Environmental Monitoring • Sample period: 4 sec • Reporting period: 5 min • Data analysis: in-network aggregation Alarm Notification • Local processing for large variation detect • Report alarm within 4 secs CS294-1 F03

  45. Optimization evaluation • Sensing & Processing • 15 mw 17 mJ per day • Sleeping • 60 uw 5038 mJ per day • Communination • hardware accelerators for edge capture and serialization • 10 kbps => 50 kbps 2262 => 452 mJ/day5x • Rendezvous: 2x time-synchronization* • time-stamp packets: +- 100 ms • radio bit edge detection: +- 2 us • radio-level timesynch 669 => 33 mj/day20x • Wake-up • packet listen: 108 ms (21 ms) 54,000 => 25 mj/day2000x • sample radio channel for energy: 50 us • Combined: 2AA lifetime grows from 1 year to 9 years • dominated by sleep energy * receiver-based alternative (Elison) CS294-1 F03

  46. UC Berkeley Family of Motes CS294-1 F03

  47. Mica2 and Mica2Dot • ATmega128 CPU • Self-programming • Chipcon CC1000 • FSK • Manchester encoding • Tunable frequency • Byte spooling • Power usage scales with range 1 inch CS294-1 F03

  48. Basic Sensor Board • Light (Photo) • Temperature • Prototyping space for new hardware designs CS294-1 F03

  49. Light (Photo) Temperature Acceleration 2 axis Resolution: ±2mg Magnetometer Resolution: 134mG Microphone Tone Detector Sounder 4.5kHz Mica Sensor Board CS294-1 F03

  50. 5” x 2.5” x 3” size <$250 total 2-axis accelerometer COTS-BOTS (UCB)Commercial Off-The-Shelf roBOTS CS294-1 F03

More Related