Download Wireless Sensor Networks - Lecture Slides | ECE 284 and more Study notes Electrical and Electronics Engineering in PDF only on Docsity!
Wireless Sensor Networks Wireless Sensor Networks
Curt Schurgers
2 ECEECE 284284
Wireless Sensor Networks Wireless Sensor Networks
Networks of tiny autonomous data-gathering devices
Large-scale ● Dense local sampling, instead of distant more global view ● Small sensors have a limited range
Autonomous operation ● Network management and setup are intractable for large-scale networks ● Self-configuration and adaptation ● Examples Time synchronization Self-localization Mobile agents: instantiate functionality where needed
Data centric ● Interest in the data, not the sensor identity ● Tailor protocols to be data-centric rather than node-centric
Robust operation as a network ● Tiny cheap devices in often harsh environments ● Limited resource
UCB mote
The system is the
sensor network
3 ECEECE 284284
Energy Efficiency Energy Efficiency
Limited resources ● Energy and power: small form factor, thus small batteries and limited energy scavenging possibilities ● Communication energy is important, since we are interested in the behavior of the entire network ● Turning the radio off is one the main ways to be more energy efficient Strategies: adaptation and application specific design
[WaveLAN] Feeney, L.M.; Nilsson, M., “Investigating the energy consumption of a wireless network interface in an ad hoc networking environment”, INFOCOM 2001, pp. 1548-1557, 2001.
[CC1000] http://etd.adm.unipi.it/theses/available/etd-05252004-154652/unrestricted/Chap4.pdf
[TR1000] A. Savvides, C.-C. Han, M. Srivastava, “Dynamic fine-grained localization in ad-hoc networks of sensors,” MobiCom 2001 , Rome, Italy, pp. 166 – 179, July 2001.
0
4
8
12
16
12.48 (^) 12.
14.
0. Tx Rx idle sleep
d ~ 20 meters, 2.4 kbps
(mW)
RFM TR
0
500
1000
1500
1 2 3 4
967 844
1327
66 Tx Rx idle sleep
d ~ 250 meters, 2 Mbps (mW)
802.11 WaveLAN
0
20
40
60
(^48 )
54
0. Tx Rx idle sleep
d ~ 50 meters, 38.4 kbps
Chipcon CC
(mW)
4 ECEECE 284284
Localized Algorithms Localized Algorithms
Centralized algorithms ● Global information ● Decisions are made in a central location and this information is propagated ● E.g. centralized routing scheme Distributed algorithms ● Global information ● Decisions are made by the individual devices ● E.g. AODV Localized algorithms ● Local information only ● Decisions are made by the individual devices ● E.g. geo-routing
Simulation of localized algorithms ● An infinite sensor field can be mimicked by ignoring statistics from ‘edge’ nodes ● The definition of ‘edge’ depends on how localized the algorithm is: e.g. only information from 1-hop neighbors
edge
Scaling: ability for an algorithm to handle increasingly large networks ● Localized > distributed > centralized
7 ECEECE 284284
Leveraging Network Density Leveraging Network Density
The desired node density depends on the network deployment: random deployment requires typically a larger network density ● Statistical connectivity guarantees ● Statistical sensor coverage guarantees ● Fault tolerance and robustness
Result: the local network density is often larger than strictly needed, which can be exploited in topology management
Leveraging communication density ● Less nodes are needed to provide a sufficiently connected network ● Nodes can go to sleep to save communication energy
Leveraging sensing density ● Less nodes are needed to provide sufficient coverage ● Save sensing energy ● Save communication energy (less data to report)
R
A
8 ECEECE 284284
GAF [Xu01] GAF [Xu01]
Geographic Adaptive Fidelity (GAF) ● Leverage node density to put nodes to sleep ● Conserve the data forwarding capacity of the network ● Utilize geographic information ● Energy savings for very dense networks
Approach ● Divide the network in virtual grids ● Each node in a grid is equivalent in terms of traffic forwarding ● Each node in a grid has to be able to communicate with each node in a neighboring grid
R
G
G 5
2 G^2 = R
9 ECEECE 284284
Energy savings ● Only one node is active in a grid at each time: grid behaves as ‘virtual node’ ● Rotate functionality amongst nodes in the grid
Analysis
GAF [Xu01]GAF [Xu01]
Average energy savings factor
3.0 2.82 44.
2.5 2.22 35.
2.0 1.59 25.
1.5 0.87 13.
1.0 0 0
M’ M λ
= ⋅ = 5 ⋅ λ
2
A
G
M N
eM
M
M −
m
E
E
M
E
E
λ
1
0
− ⎥ ⎦
⎤ ⎢ ⎣
⎡ E
E
Average number of nodes in a grid (distribution is approx Poisson)
Average number of nodes in an occupied grid
Energy of node in a grid with m nodes
10 ECEECE 284284
SPAN [Che01] SPAN [Che01]
SPAN
● Leverage node density to put nodes to sleep ● Conserve the data forwarding capacity of the network ● On top of 802.11 PSM
Operation ● Coordinator nodes stay awake and forward data ● Non-coordinator nodes are in PSM (reachable as destinations) ● Goal: minimize number of coordinators while not significantly affecting the forwarding capacity of the network ● Selection rule: a node becomes a coordinator of two of its neighbors cannot reach each other directly, or via 1 or 2 coordinators ● Collisions and selection priority are handled by a random backoff of HELLO messages, which depends on the remaining energy and the ‘benefit’ of selecting the node (i.e. how many of its neighbors become connected) (^) Assume geographic forwarding
13 ECEECE 284284
Breach Path [Meg01a] Breach Path [Meg01a]
Sensing model: the signal energy detected by the sensor decays exponentially with the distanced between the signal sourcep and the sensors
Maximum breach path: ● Path through network that maximizes the minimum distance to any node ● This minimizes the maximum detection probability of any node at any point in time
Solution ● Maximum breach path must lie on the bounded Voronoi tessellation ● Apply a search algorithm on this graph to find the best path
Note: The Voronoi tessellation is such that all points in the tile of node A are closer to A then to any other node
14 ECEECE 284284
Exposure [Mer01b] Exposure [Mer01b]
Exposure takes into account the signal energy measured: ● Signal strength at each sensor ● Time duration of the sensing
Two possible definitions of the signal intensityI ● The sum of the signal strengthS at all sensors ● The maximum signal strengthS over all sensors
The exposure over a time interval [t 1 ,t 2 ] is defined as:
Goal: find the path with the minimum total exposure
Algorithm based on an overlay grid and Dijkstra’s shortest path algorithm
All sensor intensity model Closest sensor intensity model
15 ECEECE 284284
Wakeup Protocols Wakeup Protocols
Non-uniform node activity in time ● React to events ● Periodic data with small or large period
Inefficient to optimize the node behavior under the assumption of continuous traffic ● Go to low power sleep mode ● Wake up on-demand when there is activity: wakeup protocol ● Optimize the behavior when there is activity: MAC protocol
Activity profile
time
time
16 ECEECE 284284
STEM [Sch02a] STEM [Sch02a]
Sparse Topology and Energy Management (STEM)
Nodes are equipped with two radios (with possibly different power consumption) ● Data channel for regular communication ● Wakeup channel for neighbor wakeup
A sleeping node turns the data radio off (sleep) and uses a periodic listen-sleep cycle for the wakeup radio with periodT
Wakeup procedure ● A wakeup signal is sent, consisting of a succession of small wakeup beacons ● If the sleeping nodes receives a wakeup beacon, it responds with an acknowledgement ● At this point, data transmissions use the data radio (using any MAC protocol)
Send wakeup signal
on
off
A
B
Only the wakeup channel is shown here A decides towake up B wakeup signalB receives the
T
19 ECEECE 284284
GeRaFGeRaF [Zor03][Zor03]
Energy-delay tradeoff by varying the period of the listen-sleep cycle
GeRaF also leverages density statistically through geographic forwarding: several nodes are possible relays
STEM can also leverage density by combining it with GAF, SPAN, etc. [Sch02a]
E 0
E Data: 1 %
λ = N
20 ECEECE 284284
MAC Protocols MAC Protocols
Goal ● Minimize energy ● Allow possible hit on latency, fairness (node-level versus application-level), …
Tackle the sources of energy wastage ● Collisions ● Overhearing: receiving messages not meant for this node ● Control overhead ● Idle listening
Many MAC protocols have been proposed for sensor networks ● S-MAC [Ye02]: CSMA/CA MAC with periodic listen-sleep ● TRAMA [Raj03]: reservation-based TDMA ● SMACS [Soh00]: sparse FDMA – TDMA ● Pico Radio MAC [Guo01]: CDMA codes for each sender (local assignment algorithm) and a separate wakeup channel ● WiseMAC [Elh04]: non-persistent CSMA with preamble sampling (use long PHY preamble to wakeup the receiver, similar to the technique used in pagers) ● TDMA-W [Che04]: TDMA with send slots and wakeup slots ● T-MAC [Van03]: CSMA/CA MAC with periodic listen-sleep and burst transmission ● …
21 ECEECE 284284
S- S-MAC [Ye02]MAC [Ye02]
Sensor MAC (S-MAC) ● Contention-based MAC for wireless sensor networks ● Loose synchronization
Operation ● Periodic listen – sleep ● Listen interval has two subintervals: one for SYNC messages, one for RTS ● Exchange SYNC packets Synchronize neighbors or learn their schedule Use slotted carrier sense ● Exchange RTS packets Acquire the medium Use slotted carrier sense ● Data is exchanged using RTS/CTS/DATA/ACK handshake ● Virtual carrier sense: set NAV and nodes go to sleep if a neighbor is transmitting
listen sleep
For SYNC
For RTS
22 ECEECE 284284
S- S-MAC [Ye02]MAC [Ye02]
How is energy saved? ● Avoid overhearing: go to sleep based on virtual carrier sense ● Periodic listen – sleep: adapts to the traffic load Loose synchronization ● Nodes that adopt the same schedule form virtual clusters ● Node can adopt multiple schedules
0 2 4 6 8 10
200
400
600
800
1000
1200
1400
1600
1800
Average energy consumption in the source nodes
Message inter-arrival period (second)
Energy consumption (mJ)
802.11-like protocol Overhearing avoidance S-MAC
Synch region 1 (^) Synch region 2
25 ECEECE 284284
Localization [Sav04] Localization [Sav04]
Location information of the sensor nodes is needed for ● Geographic routing ● Locating an event ● Beamforming ● Target tracking and localization Additionally, there is the problem of locating a sensed target with respect to the sensors
[Source: http://www.eng.yale.edu/enalab/courses/eeng460a/]
Taxonomy ● Active localization: send signals into environment to measure range Non-cooperative: e.g. radar Cooperative target: targets emit known signal Cooperative infrastructure: beacons/anchors emit known signal
● Passive localization: passively monitoring existing signals Blind source localization: blind beamforming Passive target localization: some knowledge of the source Passive self-localization: e.g. use RF signal properties, such as RSSI
Beacon
Unkown Location
Randomly Deployed Sensor Network
26 ECEECE 284284
Localization [Sav04] Localization [Sav04]
Ranging ● RF connectivity information ● RF signal strength: need accurate hardware and is typically crude ● RF time of flight (ToF): need synchronization (easier for infrastructure), could use UWB ● Acoustics: easier synchronization (e.g. use RF and acoustic signal in parallel and look at the difference in arrival time) ● Angle of arrival
MK-2 with ultrasound ranging
[Source: http://www.eng.yale.edu/enalab/courses/eeng460a/]
Technology Range Accuracy
RF RSSI 10 m 2 – 3 m (motes) Laser ToF 200 m 2 cm (very directional) RFIDs and Infrared sensors A few meters Proximity metric Acoustic angle of arrival Tens of meters 5 degrees
RF ToF Tens of meters 15 cm (UWB claim)
Acoustic ToF Tens of meters 10 cm
Ultrasonic ToF (25 – 40KHz) A few meters 2-5 cm
27 ECEECE 284284
Multilateration [Sav05] Multilateration [Sav05]
Based on distance estimates ● Ultrasound ToF
Multilateration: find positions based on a set of known beacons ● Atomic multilateration: direct measurements, similar to GPS ● Collaborative multilateration: collaborate with other intermediate nodes Form collaborative subtree: group such that there are enough constraints to find all positions Compute initial estimates Refine position: Kalman filter
Distributed versus centralized operation
Known position
Unknown position
λ = 6 6 beacons R = 15 m
28 ECEECE 284284
MDS- MDS-MAP [Sha04]MAP [Sha04]
Based on multi-dimensional scaling (MDS) ● Gather connectivity information (pure connectivity or with distance measurements) ● Generate relative map using MDS ● Normalize using known anchor locations
5% distance measurement error
Transmission range R
Exact node locations (^) Estimated relative map
MDS-MAP(C,R)
λ = 14 5% distance measurement error
Variants ● C: on the entire network ● P: first build local maps and merge them ● R: added refinement step
Complexity: O(n^3 )
31 ECEECE 284284
TPSN [Gan03] TPSN [Gan03]
Link synchronization ● TPSN uses sender-receiver synchronization ● Alternative: receiver-receiver, e.g. RBS [Els02]) ● Consider relative drift RD between the two clocks
Resynchronization is needed periodically ● Drift of a node can be extrapolated based on previous measurements and partly compensated for
Results on MICA motes using TPSN: average error = 17 μs, worst case error = 44 μs
A B
S – R
A
A A
B
B B R – R
Beacon
( )^1
A B t t
UC UC UC SR
S P R RD
Error
− > −>
− =^ + + +
A B t t
UC UC
Error R R P R RD
−>
32 ECEECE 284284
Other Challenges Other Challenges
Network planning ● Flat architecture: direct transmission or multi-hop communication ● Hierarchical: multi-tier architecture Ad-hoc clustering: nodes are homogeneous and some are chosen to fulfill the role of cluster heads (rotating functionality) Heterogeneous networks with specialized nodes ● Static or mobile networks
Data collection ● Different types of data: periodic, reactive, … ● Data can be generated by one node or a collection of nodes ● Data can be compressed, merged into a decision, analyzed, used in classification, stored in a database, etc. ● Data can be merged and compressed, used in distributed signal processing ● Event tracking to warn nodes about likely upcoming events
Functionality ● Routing ● Security ● Data base management and querying ● Hardware and software (e.g. TinyOS RTOS)
33 ECEECE 284284
References References
[Est99] Estrin, D., Govindan, R., “Next Century Challenges: Scalable Coordination in Sensor Networks,” MobiCom’99, Seattle, WA, pp.263-270, Aug. 1999.
[Sch02a] Schurgers, C., Tsiatsis, V., Ganeriwal, S., Srivastava, M., "Optimizing Sensor Networks in the Energy-Latency-Density Design Space," IEEE Transactions on Mobile Computing, Vol.1, No.1, pp. 70-80, 2002.
[Zor03] Zorzi, M., Rao, R., "Geographic Random Forwarding (GeRaF) for Ad Hoc and Sensor Networks: Multihop Performance," Transactions on Mobile Computing, Vol. 2, No. 4, 2003.
[Che01] Chen, B., Jamieson, K., Balakrishnan, H., Morris, R., “Span: An Energy-Efficient Coordination Algorithm for Topology Maintenance in Ad Hoc Wireless Networks,” Mobicom’01, Rome, Italy, pp. 85-96, 2001.
[Xu01] Xu, Y., Heidemann, J., Estrin, D., “Geography-Informed Energy Conservation for Ad Hoc Routing,” Mobicom’01, Rome, Italy, pp.70-84, 2001.
[Ye02] Ye, W., Heidemann, J., Estrin, D., “An Energy-Efficient MAC Protocol for Wireless Sensor Networks,” Infocom '02, New York, NY, pp.1567-1576, 2002.
[Gup00] Gupta, P., Kumar, P., “The Capacity of Wireless Networks,” IEEE Trans. on Information Theory, Vol.46, pp.388-404, 2000.
34 ECEECE 284284
References References
[Raj03] Rajendran, V., Obraczka, K., Garcia-Luna-Aceves, J.J., “Energy-efficient, collision- free medium access control for wireless sensor networks,” SenSys’03, Los Angeles, CA, pp.182-192, 2003.
[Guo01] Guo, C., Zhong, L., Rabaey, J., “Low-power distributed MAC for ad hoc sensor radio networks,” Globecom’01, San Antonio, TX, pp. 2944-2948, 2001.
[Elh04] El-Hoiydi, A., Decotignie, J.-D., “WiseMAC: An Ultra Low Power MAC Protocol for Multi-hop Wireless Sensor Networks,” ALGOSENSORS’04, Lecture Notes in Computer Science, LNCS 3121, Springer-Verlag, pp.18-31, 2004.
[Ch04] Chen, Z., Khokhar, A., “Self organization and energy efficient TDMA MAC protocol by wake up for wireless sensor networks,” SECON'04, Santa Clara, CA, 2004.
[Van03] Van Dam, T., Langendoen, K., “An adaptive energy-efficient MAC protocol for wireless sensor networks,” SenSys’03, Los Angeles, CA, 2003.
[Soh00] Sohrabi, K., Gao, J., Ailawadhi, V., Pottie, G., “Protocols for Self-Organization of a Wireless Sensor Network,” IEEE Personal Communications Mag., Vol.7, No.5, pp.16-27, 2000.
[Els01] Elson, J., Estrin, D., “Time Synchronization for Wireless Sensor Networks,” IPDPS’01, pp.1965-1970, San Francisco, CA, 2001.
