Energy efficiency in wireless sensor networks

1
Energy efficiency inwireless sensor networks

• Michele Zorzi
• Dipartimento di Ingegneria dellinformazione
• CNIT Universita di Padova

2
before we start...

• This is by no means an exhaustive tutorial on the
  topic
• Rather, we will see a number of examples that
  hopefully give an idea of the various issues
  involved
• In this presentation I freely use material
  provided by various colleagues (besides my own),
  including Mani Srivastava (UCLA), Chiara Petrioli
  (Univ. Roma la Sapienza) and Stefano Basagni
  (Northeastern University)
• Also material from the web (which I try to
  properly acknowledge)

3
Ad hoc networks defined

• Absence of a fixed infrastructure
• All nodes are functionally equal
• Autonomous
• Reconfigurability
• Multihop data delivery
• Sensor vs. Ad hoc networks
• Application domain
• Data generation and handling
• Constraints/requirements/objectives

4
Main technical problems

• Media access mechanism
• Routing
• Discovery and self-organization
• Topology management
• Reliable end-to-end data delivery
• Data collection and dissemination
• Handling of mobile nodes
• Security
• Energy efficiency

5
Energy efficiency

• One of the main issues here
• Battery powered nodes, not easily recharged
• Especially important for sensor networks
• Low power electronics a traditional field
• No Moores law for battery technology
• More recent view low-energy architecture
• Main issues in ad hoc sensor networks
• Short range and sporadic communications may lead
  to dominance of radio electronics
• Power consumption to rx/listen comparable to (and
  sometimes even larger than) that to transmit
• Sleep times are good but turn-around time/energy
  overheads
• These observations call for new design criteria
  of the protocols

6
Generic Sensor Node
In-node processing
Wireless communication with neighboring nodes
Event detection
Acoustic, seismic, image, magnetic, etc. interface
Electro-magnetic interface
sensors
radio
CPU
Limited battery supply
battery
Energy efficiency is the crucial h/w and s/w
design criterion
7
MICA2
8
Mote Platform Evolution
9
The eyesIFX Mote
10
The eyesIFX Mote
11
Where does the energy go?

• Processing
• excluding low-level processing for radio,
  sensors, actuators
• Radio
• Sensors
• Actuators
• Power supply

12
Computation Communication
Energy breakdown for MPEG
Energy breakdown for voice
Decode
Decode
Transmit
Encode
Encode
Receive
Receive
Transmit
Radio Lucent WaveLAN at 2 Mbps
Processor StrongARM SA-1100 at 150 MIPS

• Radios benefit less from technology improvements
  than processors
• The relative impact of the communication
  subsystem on the system energy consumption will
  grow

13
Energy Consumption of Hardware
14
Power States at Node Level
Active
Active
Telos Enabling Ultra-Low Power Wireless
Research, Polastre, Szewczyk, Culler, IPSN/SPOTS
2005
15
Identifying the Energy Consumers

• Need to shutdown the radio

TX
RX
CPU
IDLE
SLEEP
SENSORS
RADIO
16
Transmission and PHY issues

• Wireless communications RF vs. IR
• Traditional schemes vs. spread spectrum
• How many bits per symbol?
• Affects the on-time, but increase complexity of
  transceiver (energy tradeoff here)
• Gain inversely proportional to start-up time
• Modulation scaling
• Dynamic control of the constellation size
• Reduces overall energy cost of transmitting a bit
• Tradeoffs in error control

17
Error control tradeoffs for multihop

• We address issues related to error control
• Problem deliver reliably a packet to its final
  destination
• Possibly via multi hop operation
• Some considerations in recent literature
• Multihop not necessarily good
• Even coding not necessarily good
• Need to rethink common wisdom in communications
  theory?
• Approach
• We compare various multihop solutions
• We assume coding is used

18
Communication View Coding is Always good for
Energy/Bit
Shih et. al., Mobicom 2001
19
Considered scenarios

• Preliminary evaluation to understand tradeoffs
• Consider case in which decoding energy is
  significant (wrt transmission)
• Three scenarios
• Direct transmission from source to destination
• Multihop transmission with end-to-end FEC
• Multihop transmission with hop-by-hop FEC

20
Coding performance block codes

• Error probability vs. tx energy for h2 hops

21
Coding performance convol. codes
3 hops, Prec -110dBW
3 hops, Prec -100dBW
22
Reality Coding Not Always Good Due to
Computation Energy

• Encoding energy ltlt Decoding energy
• Computation energy dominates at higher target BER

With Viterbi Decoder on an ASIC(5X more
efficient computation)
With Viterbi Decoder on a StrongARM
Shih et. al., Mobicom 2001
23
Media access control

• Main sources of energy consumption
• Collisions, listening/overhearing, protocol
  overhead
• Main issues here
• Collision avoidance
• Aggressive use of sleep times
• Random topology
• TDMA scheduling, centralized, signaling
• Random monitoring, uncoordinated, RTS/CTS
• Sleep modes how to coordinate, RTS/CTS?
• Other considerations node density, latency, ...

24
Media access control

• IEEE 802.11 (CSMA/CA) typically used in ad hoc
  networks
• Other proposals (none implemented) FAMA
• MAC design not optimized, lack of efficient
  energy savings strategies
• Effect of MAC mechanisms (e.g., backoff)
• Proposed protocols for sensor networks
• SMAC scheduled sleep in 802.11
• Scheduled TDMA with rendez-vous
• Sleep modes and wake-up low-power radio?

25
Topology maintenance

• How to build an ad hoc/sensor network
• Node association, power control
• Flat vs. Clustering
• Self-organizing neighbor discovery and
  management
• How to guarantee that a recipient node is
  available in the presence of sleep modes
• STEM Wait for recipient to wake up
• SPAN identify minimum sets of nodes for
  connectivity and bandwidth and rotate
• GAF nodes in areas equivalent for routing
• May incur significant latency (STEM) or
  inefficient forwarding (GAF)
• Are there better solutions?
• Random forwarding in dense networks

26
Using a backbone to save energy

• In this example, all nodes are within coverage of
  the four black nodes
• Same is true for the four red nodes
• We can let the red and black sets of nodes take
  turns
• Lifetime is doubled
• Very simple idea
• More details are needed
• Main idea in SPAN as well as other schemes

27
Routing issues

• Probably the most studied issue for ad hoc and
  sensor networks
• Various solutions with different objectives
• A number of proposals for energy efficiency
• Proactive vs. reactive protocols
• Flat vs. Hierarchical
• Energy efficient routing
• Issues in sensor networks
• Different traffic, data collection and addressing
  model
• Energy efficiency more critical
• In-network processing

28
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

29
MAC and routing integration

• MAC, routing and topology maintenance are
  intimately related to each other
• In some cases, the MAC protocol negatively
  interacts with the higher layers, esp. with
  multihop routing
• It seems reasonable to look for more tightly
  integrated solutions
• Some examples
• Energy efficient routing via power control
• Geographic random forwarding

30
Source placement event-radius model
31
Data aggregation

• Efficient data mgm
• E.g., suppression of duplicates, aggregation
• Support for data requests

32
A more general look at Data Centric vs. Address
Centric approach (Krishnamachari et al.)

• Address Centric
• Distinct paths from each source to sink.
• Data Centric
• Support aggregation in the network where
  paths/trees overlap
• Essential difference from traditional IP
  networking
• Building efficient trees for Data centric model
• Aggregation tree On a general graph if k nodes
  are sources and one is a sink, the aggregation
  tree that minimizes the number of transmissions
  is the minimum Steiner tree. NP-complete.Approxim
  ations
• Center at Nearest Source (CNSDC) All sources
  send through source nearest to the sink.
• Shortest Path Tree (SPTDC) Merge paths.
• Greedy Incremental Tree (GITDC) Start with path
  from sink to nearest source. Successively add
  next nearest source to the existing tree.

33
Comparison of energy costs
Data centric has many fewer transmissions than
does Address Centric independent on the tree
building algorithm.
Address Centric Shortest path data centric Greedy
tree data centric Nearest source data
centric Lower Bound
34
Energy Efficiency in MAC

• Major sources of energy waste
• Idle listening
• Long idle time when no sensing event happens
• Collisions
• Control overhead
• Overhearing
• Try to reduce energy consumption from all above
  sources
• TDMA requires slot allocation and time
  synchronization
• Combine benefits of TDMA contention protocols

35
Sensor-MAC (S-MAC) Design(Wei et al. 2002)

• Tradeoffs
• Major components of S-MAC
• Periodic listen and sleep
• Collision avoidance
• Overhearing avoidance
• Message passing

36
Periodic Listen and Sleep

• Problem Idle listening consumes significant
  energy
• Nodes do not sleep in IEEE 802.11 ad hoc
  mode
• Solution Periodic listen and sleep
• Turn off radio when sleeping
• Reduce duty cycle to 10 (200 ms on/2s off)
• Increased latency for reduced energy

37
Periodic Listen and Sleep

• Schedules can differ

• Preferable if neighboring nodes have same
  schedule
• easy broadcast low control overhead

Border nodes two schedules broadcast twice
38
Periodic Listen and Sleep

• Schedule maintenance
• Remember neighbors schedules
• to know when to send to them
• Each node broadcasts its schedule every few
  periods
• Refresh on neighbors schedule when receiving an
  update
• Schedule packets also serve as beacons for new
  nodes to join a neighborhood

39
Collision Avoidance

• Problem Multiple senders want to talk
• Options Contention vs. TDMA
• Solution Similar to IEEE 802.11 ad hoc mode
  (DCF)
• Physical and virtual carrier sense
• Randomized backoff time
• RTS/CTS for hidden terminal problem
• RTS/CTS/DATA/ACK sequence

40
Overhearing Avoidance

• Problem Receive packets destined to others
• Solution Sleep when neighbors talk
• Basic idea from PAMAS (Singh 1998)
• But we only use in-channel signaling
• Who should sleep?
• All immediate neighbors of sender and receiver
• How long to sleep?
• The duration field in each packet informs other
  nodes about the sleep interval

41
Message Passing

• Problem In-network processing requires entire
  message
• Solution Dont interleave different messages
• Long message is fragmented sent in burst
• RTS/CTS reserve medium for entire message
• Fragment-level error recovery
• extend Tx time and re-transmit immediately
• Other nodes sleep for whole message time

42
Msg Passing vs. 802.11 fragmentation

• Time reservation by duration field

• If ACK is not received, give up Tx fairness
• No indication of entire time other nodes keep
  listening

43
S-MAC Experimental results(implemented on UCB
Mote over RFM radio)

• Topology and measured energy consumption on
  source nodes

• Each source node sends 10 messages
• Each message has 10 fragments x 40B
• Measure total energy
• Data control idle

44
Implementation and Experiments

• Platform Mica Motes
• Topology 10-hop linear network
• S-MAC saved a lot of energy compared with a MAC
  without sleep

45
Motivation for Topology Control

• High power
• High interference
• Low Throughput

46
Motivation for Topology Control

• Low power
• Low interference
• High Throughput
• Global Connectivity

47
Topology Control

• Two major ways to do TC
• Controlling the power at the node to create
  energy effective topologies
• Taking advantage of the network density for
  turning off the radio interface
• Node sleep saves a lot
• Only a (connected) fraction of the nodes stays up
  for performing network functions
• Another idea
• backbone formation can also be used for the sake
  of topology control (only backbone nodes are
  awake, or better, nodes have different duty
  cycles depending on whether they belong to the
  backbone or are ordinary nodes)

48
Using a backbone to save energy

• In this example, all nodes are within coverage of
  the four black nodes
• Same is true for the four red nodes
• We can let the red and black sets of nodes take
  turns
• Lifetime is doubled
• Very simple idea
• More details are needed
• Main idea in SPAN as well as other schemes

49
Geographical Adaptive Fidelity(first solution
exploiting this idea)

• GAF Typical example of topology control
• Nodes are placed in grids of side r (needs
  location awareness)
• The side r is a function of Tx radius R If r
  R/v5 then every node has a neighbor in an
  adjacent grid
• IdeaRouting-wise all nodes in a grid are
  equivalent (not really true)
• A grid leader is elected and stays on
• Based on the node residual energy
• All other nodes in the grid go to sleep
• Leaders form a (connected) backbone
• Load balancing is realized by
• periodic re-election of the
• leader

50
GAF example
active node
sleeping node
Radio range

• It is sufficient to have one active node per area
• Advancement per hop is less than half the radio
  range
• Signaling needed (though not as heavy as in SPAN)

51
Energy Conservation Strategy

• Observe Most of the time, the network is only
  monitoring, waiting for an event to happen
• New strategy put nodes to sleep and only wake
  them up when they need to participate in data
  forwarding

Nodes have their radio in sleep mode to conserve
energy
Nodes turn on their radio, when they need to
communicate
52
STEM Data driven wakeup
Wake up the nodes along the path HOW ???
Sensor-triggered node wakeup
user
event
Zzz
Zzz
Zzz
Zzz
sensor network
Path nodes need to be woken up
53
How can a Sleeping Node be reached?
zzzzzzz

• Solutions
• Each node wakes up occasionally to listen for
  wake up messages
• Low duty cycle paging save energy.

54
Implementation Issues

• Problem Wake up messages can collide with
  ongoing data Tx
• Solution Use separate channel for wake up
  messages.

55
STEM Sparse Topology and Energy Management

• Need to separate Wakeup and Data Forwarding
  Planes
• Chosen two separate radios for the two planes
• Use separate radio for the paging channel to
  avoid interference with regular data forwarding
• Trades off energy savings for path setup latency

56
High Level Operation of STEM
57
Detailed Operation of STEM
Initiator node
f1
B1
B2
1. beacon received
Train of beacon packets
TRx
2. beacon acknowledge
f1
Target node
58
STEM Energy Analysis
fw wakeup frequency ? fw . tburst ? T /
TRx ? Pwakeup / Pdata
59
Exploiting Latency AND Density

• Other approaches (ASCENT, GAF, SPAN, etc) exploit
  node density
• Combine STEM and one of these schemes (chosen GAF)

60
Combining STEM and GAF

• As in GAF, 1 node is active in each grid
• the grid can be considered a virtual node
• Observe In GAF, the leader has to keep its radio
  on all the time
• Absence of traffic in the monitoring state not
  exploited
• This virtual node runs the STEM protocol
• Requires changes in leader election scheme

61
STEM GAF Energy Analysis
GAF leader election overhead increases with ? ?
affects choice of STEM-B or STEM-T
STEM alone
Energy Savings (compared to no topology
management) as a function of density. STEM GAF
provides energy savings, with a factor of 100 for
practical settings
62
Ultra-low power radio

• Main problem in topology control sleeping nodes
  cannot be contacted
• What if they could?
• Suppose we have an ultra-low power radio
• Very simple, must receive a single bit (or short
  address)
• Always on, when activity detected it activates
  main radio
• If this second radio consumes what can be
  gathered from the environment, infinite life!!
• Energy scavenging

63
Comparison of Energy Sources
With aggressive energy management, ENS might live
off the environment.
Source UC Berkeley
64
Prometheus Perpetually Powered Telos

• Solar energy scavenging system for Telos
• Super capacitors buffer energy
• Lithium rechargeable battery as a backup
• Uses MCU to manage charge cycles to extend system
  lifetime
• Manage limited recharges

Perpetual Environmentally Powered Sensor
Networks, Jiang, Polastre, Culler, IPSN/SPOTS,
2005
65
Saving Communication Power Data Mules
The observer (data collector) is mobile. Sensors
are static. The observer moves on the same
path (which is fixed and may not be chosen at
will) repeatedly.