Power Issues in Wireless Sensor Nets

Transcription

1 Power Issues in Wireless Sensor Nets David Culler CS252 Spring /31/05 CS252 S05 1

2 Outline Basic model of operation Node Design a for low power consumption Operating System Issues Design of the power-supply subsystem MAC-level network design for power Higher-level network design for power Application level Important areas of development Discussion 3/31/05 CS252 S05 2

3 Model of operation Active Active Sleep WakeUP Work Sleep WakeUP Work Sleep Active [Wakeup / Work] Peak Power Essentially sum of subsystem components MW in supercomputer, kw in server, Watts in PDA milliwatts in mote class device Sleep power Minimal running components + leakage Microwatts in mote-class Average power P ave = = (1-f active )*P sleep + f active *P active P ave = f sleep *P sleep + f wakeup *P wakeup + f work *P work Lifetime EnergyStore / (P ave -P gen ) Duty Cycle 3/31/05 CS252 S05 3

4 Passive Vigilance Sense only when there is something useful to detect Listen only when there is something useful to hear How do you know? By arrangement By cascade of lower power triggers 3/31/05 CS252 S05 4

5 Mote Power Parameters 1s Microwatts sleep 10s of milliwatts active (wakeup or work) Wakeup substantial Milliseconds (1000s of instructions) 1% Duty Cycle is common Wakeup matters 3/31/05 CS252 S05 5

6 Batteries Still the best energy store Issues Voltage Source current Leakage Voltage profile Recharge 3/31/05 CS252 S05 6

7 Design of a Low Power Node 3/31/05 CS252 S05 7

8 Key Design Elements Flash Storage proc timers Sensor Interface ADC analog sensors digital sensors data logs pgm images Data SRAM WD pgm EPROM Low-power Standby & Wakeup Wireless Net Interface Wired Net Interface RF transceiver serial link USB,EN, antenna Efficient wireless protocol primitives Flexible sensor interface Ultra-low power standby Very Fast wakeup Watchdog and Monitoring Data SRAM is critical limiting resource 3/31/05 CS252 S05 8

9 Mote Platform Evolution 3/31/05 CS252 S05 9

10 Platforms Focused on low power Standards Based Sleep - Majority of the time IEEE , USB Telos: 2.4µA IEEE MicaZ: 30µA CC2420 radio Wakeup 250kbps As quickly as possible to process and return to sleep Telos: 290ns typical, 6µs max MicaZ: 60µs max internal oscillator, 4ms external Process Get your work done and get back to sleep Telos: 4MHz 16-bit MicaZ: 8MHz 8-bit TI MSP430 Ultra low power» 1.6µA sleep» 460µA active» 1.8V operation 2.4GHz ISM band TinyOS support New suite of radio stacks Pushing hardware abstraction Must conform to std link Ease of development and Test Program over USB Std connector header Interoperability Telos / MicaZ / ChipCon dev UCB Telos Xbow MicaZ 3/31/05 CS252 S05 10

11 TinyOS-driven architecture 3K RAM = 1.5 mm 2 CPU Core = 1mm 2 multithreaded RF COMM stack =.5mm 2 HW assists for SW stack Page mapping SmartDust RADIO =.25 mm 2 SmartDust ADC 1/64 mm 2 I/O PADS Expected sleep: 1 uw 400+ years on AA 150 uw per MHz Radio:.5mm2, -90dBm receive sensitivity 1 mw power at 100Kbps ADC: 20 pj/sample 3/31/05 10 Ksamps/second =.2 uw. jhill mar 6, 2003 CS252 S05 11

12 Microcontrollers Memory starved Far from Amdahl-Case 3M rule Fairly uniform active inst per nj Faster MCUs generally a bit better Improving with feature size Min operating voltage 1.8 volts => most of battery energy 2.7 volts => lose half of battery energy Standby power Recently a substantial improvement Probably due to design focus Fundamentally SRAM leakage Wake-up time is key Trade sleep power for wake-up time Memory restore DMA Support permits ADC sampling while processor is sleeping 3/31/05 CS252 S05 12

13 Radio Trade-offs: wakeup vs resilience / performance Wakeup vs interface level Ability to optimize vs dedicated support 3/31/05 CS252 S05 13

14 Flash Technology One write per bit per erase cycle Flash characteristics: Not used in current motes NOR AT45DB NAND Erase Slow (seconds) Fast (ms) Fast (ms) Erase Unit Large (64K-128K) Small (256) Medium (8K-32K) Writes Slow (100 kb/s) Slow (60 kb/s) Fast (MB/s) Write Unit 1 bit 256 bytes 100 s bytes Bit-Errors Low Low High (requires ECC) Read Fast Slow + I/O Bus Fast + I/O Bus Erase Cycles 10^4 10^5 10^4 10^5 10^7 Intended Use Code storage Data storage Data storage 3/31/05 CS252 S05 14

15 Power States at Node Level Active Active Sleep WakeUP Work Sleep WakeUP Work 3/31/05 CS252 S05 15

16 Tiny OS Concepts Scheduler + Graph of Components constrained two-level scheduling model: threads + events Component: Commands, Event Handlers Frame (storage) Tasks (concurrency) Constrained Storage Model frame per component, shared stack, no heap Very lean multithreading Efficient Layering structured event-driven execution Never wait or spin init Commands init Power(mode) TX_packet(buf) power(mode) send_msg (addr, type, data) Messaging Component internal thread Events Internal State TX_pack et_done msg_rec(type, data) msg_sen d_done) (success RX_pack et_done (buffer) ) 3/31/05 CS252 S05 16

17 Cooperative Interfaces Power management extends std control 1000-fold range of power draw Components informed of intention to go to sleep Take internal actions Propagate control Scoreboard determined permissible depth of sleep state Scheduler drops to sleep on idle Key interrupts drive wake-up Rich communication interfaces Signal strength Post-MAC time-stamping Sub-carrier signaling Application: query processing detection, reporting Sensor Comm Drivers & Stack Timer Filters Sensor 3 Sensor 2 Radio Clock Sensor1 Power Management 3/31/05 CS252 S05 17

18 Power-supply Subsystem Energy Store Power Source Consumer Management & Control 3/31/05 CS252 S05 18

19 Importance or primary buffer Node is able to operate from capacitors Moderate period of time (~week) Source active load (mas!) Absorb energy input Perform frequent charge cycles (daily) shallow Source high voltage recharge of secondary Power MCU during secondary recharge 3/31/05 CS252 S05 19

20 It s all about leakage Bigger isn t better More doesn t help We use two 22 F in series operate in the flat under load (1%, 10 ma) 3/31/05 CS252 S05 20

21 Recharging High density & Low leakage Software on MCU manages recharge sequence Include temperature compensation Pulsing charge current (~1x battery capacity) to 80% High level charge management and load control 3/31/05 CS252 S05 21

22 High Level Management 3/31/05 CS252 S05 22

23 Outline Basic model of operation Node Design a for low-power consumption Operating System Issues Design of the power-supply subsystem Communication Basics MAC-level network design for power Higher-level network design for power Important areas of development Discussion 3/31/05 CS252 S05 23

24 Basics Power required to transmit a given distance grows like the r 3 in free space with omnidirectional antenna Can be as bad as r 7 close to the ground Slower growth rate with directional, but Power required to route data hop-by-hop a given distance grows only linearly Connectivity determined by a host of factors SNR Transmission power, receiver sensitivity, distance Interference obstructions 3/31/05 CS252 S05 24

25 The Basic Primitive Transmit a packet Received by a set of nodes Dynamically determined Depends on physical environment at the time What other communication is on-going Each selects whether to retransmit Potentially after modification And if so, when 3/31/05 CS252 S05 25

26 Routing Mechanism Upon each transmission, one of the recipients retransmit determined by source, by receiver, by on the edge of the cell 3/31/05 CS252 S05 26

27 Communication and Power listen RX off RX off TX TX TX Costs power whenever radio is on Transmitting, receiving, or just listening Transmit is easy, Rcv is what s tricky Want to turn it on just when there is something to hear Two approaches Schedule transmission intervals» Statically, dynamically, globally, locally Make listening cheap 3/31/05 CS252 S05 27

28 S-MAC Ye, Heidemann, and Estrin, INFOCOM 2002 Traditional monolithic protocol design Synchronized protocol with periodic listen periods Black Box design Designed for a general set of workloads User sets radio duty cycle SMAC takes care of the rest so you don t have to Integrates higher layer functionality into link protocol Hard to maintain set of schedules T-MAC [van Dam and Langendoen, Sensys 2003] Reduces power consumption by returning to sleep if no traffic is detected at the beginning of a listen period Node 1 Node 2 listen sleep listen sleep Schedule 1 listen sleep listen sleep Wei Ye, USC/ISI 3/31/05 CS252 S05 28 sync sync sync sync Schedule 2

29 TDMA variants Time Division Media Access Each node has a schedule of awake times Typically used in star around coordinator» Bluetooth, ZIGBEE» Coordinator hands out slots Far more difficult with multihop (mesh) networks Further complicated by network dynamics» Noise, overhearing, interference 3/31/05 CS252 S05 29

30 Complexity of Connectivity Direct Reception Non-isotropic Large variation in affinity Asymmetric links Long, stable high quality links Short bad ones Varies with traffic load Collisions Distant nodes raise noise floor Reduce SNR for nearer ones Many poor neighbors Good ones mostly near, some far 3/31/05 CS252 S05 30

31 Low Power Listening (LPL) Energy Cost = RX + TX + Listen Scheduling tries to reduce listening Alnternatively, reduce listen cost Example of a typical low level protocol mechanism Periodically wake up, sample channel, sleep Properties Wakeup time fixed Check Time between wakeups variable Preamble length matches wakeup interval Robust to variation Complementary to scheduling Overhear all data packets in cell Duty cycle depends on number of neighbors and cell traffic Node 1 Node 2 wakeup sleep wakeup wakeup sleep RX sleep sleep 3/31/05 CS252 S05 31 wakeup TX sleep wakeup wakeup sleep wakeup time time

32 The Common Case: Data Gathering Collection of nodes take periodic samples Stream data towards a root node Root announces interest depth = 0 Nodes listen When hear neighbor with smaller depth start transmitting data to best lower neighbor set own depth to one greater (and include with data) Data transmission continuously reinforces & adjusts routes Aggregation within nodes or within the tree 3/31/05 CS252 S05 32

33 Radio Cells 3/31/05 CS252 S05 33

34 Continuous Network Discovery /31/05 CS252 S05 34

35 Local Operations => Global Behavior Nodes continually sense network environment uncertain, partial information Packets directed to a parent neighbor all other neighbors hear too carry additional organizational information Each nodes builds estimate of neighborhood adjusted with every packet and with time Interactively selects parent # trans := 1/ParentRate + #trans(parent->root) Routes traffic upward Collectively they build and maintain a stable spanning tree takes energy to maintain structure node # child? dept h parent? % link goodnes s 17 1 yes yes /31/05 CS252 S05 35

36 Power-aware Routing Cost-based Routing Minimize number of hops Minimize loss rate along the path Perform local retransmissions, minimize number along path Energy balance Utilize nodes with larger energy resources Utilize redundancy Nodes near the sink route more traffic, hence use more energy Give them bigger batteries or provide more of them and spread the load Randomize routes Utilize heterogeneity Route through nodes with abundant power sources 3/31/05 CS252 S05 36

37 Communication Scheduling TDMA-like scheduling of listening slots Node allocates listen slots for each child Transmission slots to parent Hailing slot to hear joins To join listen for full cycle Pick parent and announce self Get transmission slot CSMA to manage media Allows slot sharing Little contention Reduces loss & overhearing Connectivity changes cause mgmt traffic 3/31/05 CS252 S05 37

38 In-network Processing Best way to reduce communication cost is to not communicate Compute at the sensor Only communicate important events Compute over localized regions of the network Distributed detection Validation 3/31/05 CS252 S05 38

39 Exceptional Event Detection Challenge: Detecting Exceptional Events Rare most time spent monitoring noise Random nearly continuous sampling Ephemeral low bound of duty-cycling off time Approach: Passive Vigilance Multi-modal, low-power sensors Duty-cycled, where possible and arranged in Energy-Quality hierarchy with low (E, Q) sensors Triggering higher (E, Q) sensors, and so on How to Estimate Energy Consumption? Power = idle power + energy/event x events/time Estimate event rate probabilistically: p(tx) = Low False Alarm Rate Energy-Quality Hierarchy Energy Usage High from ROC curve and decision threshold for H 0 & H 1 How to Optimize Energy-Quality? Let x* = (x 1 *, x 2 *,..., x n *) be the n decision boundaries between H 0 & H 1. for n processes. Then, given a set of ROC curves, optimizing for energy-quality is a matter of minimizing the function f(x*) = E[power(x*)] subject to the power, probability of detection, and probability of false alarm constraints of the system. High Low Trigger network includes hardware wakeup, passive infrared, microphone, magnetic, fusion, and radio, arranged hierarchically Nodes: sensing, computing, and comm processes Edges: < Energy, P FA > < Energy, P FA > 3/31/05 CS252 S05 39 sub. SPOTS 05

40 Example: Detection & Tracking 3/31/05 CS252 S05 40

41 Key Areas to Improve Low leakage SRAM 1T sram Low leakage supercaps Communication accelerators Radio wake-up cascade Very low power detection of signal triggers receiver Robust communication protocols Sensor detection cascades 3/31/05 CS252 S05 41

42 Discussion 3/31/05 CS252 S05 42

43 Where to read for more Telos: Enabling Ultra-Low Power Wireless Research, Joseph Polastre, Robert Szewczyk, David Culler To appear in The Fourth International Conference on Information Processing in Sensor Networks: Special track on Platform Tools and Design Methods for Network Embedded Sensors (IPSN/SPOTS), April 25-27, 2005 Perpetual Environmentally Powered Sensor Networks, Xiaofan Jiang, Joseph Polastre, David Culler To appear in The Fourth International Conference on Information Processing in Sensor Networks: Special track on Platform Tools and Design Methods for Network Embedded Sensors (IPSN/SPOTS), April 25-27, 2005 Versatile Low Power Media Access for Wireless Sensor Networks, Joe Polastre and David Culler, The Second ACM Conference on Embedded Networked Sensor Systems, Nov "Design of a Wireless Sensor Network Platform for Detecting Rare, Random, and Ephemeral Events", Prabal Dutta, Mike Grimmer, Anish Arora, Steve Bibyk, and David Culler, In The Fourth International Conference on Information Processing in Sensor Networks (IPSN'05), /31/05 CS252 S05 43