Energy Conservation Techniques

1
Energy Conservation Techniques

• Prof. Sunggu Lee
• EE Dept., POSTECH

2
Overview

• Most of the material in this talk is based on
  material used in previous MANET course and
  Basagni 2004
• Material in Chaps. 4 and 14 of text is lacking
  and does not provide a comprehensive overview of
  energy conservation techniques
• Chap. 4 Sensor and Sensor-Actuator Networks
• Sensor-actuator networks are heterogeneous
  networks that comprise networked sensor and
  actuator nodes that communicate among each other
  using wireless links
• actuator node ? sink node with actuator
  capability
• Also emphases the need to use localized
  algorithms in order to provide scalable,
  distributed energy-conserving solutions
• Use only information from 1-3 hops away
• Chap. 14 mostly discusses the parts of the IEEE
  802.11 protocol used for energy conservation

3
Measuring Energy Consumption

• It is impossible to make all-encompassing general
  statements about energy consumption in all
  possible portable devices
• Due to the variety of devices available,
  operating modes, energy management techniques
  used and usage scenarios
• However, measurements on example ZigBee
  radio-based WSN devices demonstrates several
  relationships about the energy normally required
  for various types of operations
• Refer to Table 3.1 of text
• Reception and transmission require the most
  energy
• Idle state (CPU idle, radio idle) saves a LOT of
  energy
• Conclusions
• Conserving radio transmission and reception is
  more effective than conserving CPU cycles
• Thus, send and receive packets sparingly (as
  little as possible)
• Make nodes sleep (both radio and CPU idle) as
  much as possible to save energy

4
Energy Usage in FireFly Node
Table 3.1 of text
5
Approaches

• Approaches for energy-efficient communication
• Power-save protocols
• Attempt to maximize the amount of time nodes
  spend in the sleep state
• Power-control protocols
• Nodes transmit using minimum transmit-power
  levels required for network connectivity while
  attempting to forward traffic with least energy
  usage
• Maximum-lifetime routing
• Select paths with aim of maximizing network
  lifetime
• Try to balance the energy consumption across the
  nodes of the network
• Each node wants to avoid packet forwarding as
  much as possible
• However, if all nodes avoid packet forwarding,
  then multi-hop paths will become ineffective

6
Power-Save Protocols

• Network-Layer Power-Save Protocols
• Network-layer traffic is buffered at MAC layer
  for sleeping neighbors or routed using only
  nonsleeping neighbors
• Three basic strategies
• Synchronized power-save mechanism
• Nodes wake up periodically in a synchronized
  manner
• Network topology-based approach
• A dominating set of nodes is found to route
  packets
• Fully asynchronous operation
• Neighboring nodes all operate asynchronously, but
  use a schedule that allows them to meet eventually

7
Synchronized Power-Saving

• Nodes periodically wake up and exchange info
  about pending traffic
• Nodes required for the pending traffic remain
  awake while others go back to sleep
• Representative example method used in the IEEE
  802.11 standard
• Using an IBSS (independent basic service set) or
  BSS, synchronized beacon intervals are used
• At beginning of beacon interval, nodes contend to
  send the beacon frame each node sends its
  beacon after a random backoff interval
• First beacon received is used for the
  synchronization
• Each node that needs to send a packet then sends
  an ATIM (ad hoc traffic indication message)
  refer to Figs. 14.1 and 14.3-14.5
• Determines nodes that need to remain awake
• Simulation results
• Short beacon intervals result in higher energy
  savings, but at the cost of substantially reduced
  throughput
• The choice of the ATIM interval also effects the
  throughput
• Throughput maximized when ATIM window is 25 of
  beacon interval

8
Overlapping Beacon Intervals
Fig. 14.1 of text
Figs. 14.4-14.5 of text
9
Topology-Based Power-Saving

• Want to select a subset of nodes that cover all
  of the nodes in the network
• Dominating set
• A dominating set of a network is a subset of
  nodes, such that each node is either in the
  dominating set or is a neighbor of a node in the
  dominating set. Basagni 2004
• Also refer to general graph theory defn
  wikipedia.org
• Synonymous with maximal independent set
  NP-complete problem
• A connected dominating set can be used as routing
  backbone
• Nodes in the connected dominating set remain
  awake, while other nodes can go to sleep for long
  periods of time
• Finding a minimal connected dominating set
  requires a lot of computation NP-hard problem
• In general, we use a heuristic algorithm that
  finds a good connected dominating set
• What is a possible algorithm? Describe it.

10
ASCENT Protocol

• Adaptive self-configuring sensor network
  topologies (ASCENT) protocol is designed to help
  a node find an equilibrium that saves power while
  preserving connectivity
• Nodes using the ASCENT protocol are in one of
  four state (active, test, passive, sleep)
• Active nodes form a communication backbone
• When node is in sleep state, it periodically
  wakes up and enters passive state
• When node is in passive state, it listens to
  neighboring nodes and transitions to sleep or
  test state based on number of active neighbor
  nodes and loss rate
• A node in test state transitions to active state
  (after a timeout) if it determines that it can
  improve the network connectivity

11
Figure 14.13 of text
12
Figure 14.14 of text
13
Span Power-Save Protocol

• Coordinators are defined as those nodes in a
  connected dominating set
• Coordinators form a low-latency routing backbone
  for the network
• A node is eligible to be a coordinator if it
  finds two neighbors that cannot communicate with
  each other
• Nodes with greater effectiveness at connecting
  pairs of neighbors are given higher priority
• Nodes with higher energy reserves are given
  higher priority
• A higher priority node uses a smaller backoff
  interval before sending out a coordinator
  announcement
• After spending some time as a coordinator, a node
  withdraws itself from the coordinator role
• Coordinator role is rotated in order to conserve
  and balance energy usage
• Span is a synchronous power-save protocol
• Nodes must be awake at the same time to exchange
  traffic in order to participate in coordinator
  election
• Underlying buffering and traffic announcement
  based on IEEE 802.11

14
Geometric Adaptive Fidelity (GAF)

• Power-save protocol that selects its covering
  set based on geographic position information
• GAF partitions the network into a grid such
  that each node in a grid square can communicate
  with any node in an adjacent grid square
• Implies grid size of R/sqrt(5), where R is the
  node transmission range
• Selecting the active node in a grid square does
  not require exchange of connectivity information
• Nodes periodically wake up and transition to
  discovery state
• One node in a grid square is chosen as an active
  node based on its remaining energy in a random
  manner
• Higher remaining energy ? shorter random backoff
  interval

15
Asynchronous Power-Saving

• Nodes wake up periodically in an asynchronous
  manner, but with a period that eventually
  overlaps with adjacent nodes
• BECA/AFECA protocol
• Like GAF, intended for use in sensor networks
• In basic energy conservation algorithm (BECA),
  nodes independently transition between sleep and
  listening/active states
• Once a node transmits or receives traffic, it
  transitions to active state
• Then transition back to sleep state if no traffic
  for a while
• Due to independent behavior patterns, there is no
  guarantee that there is a live path to the
  destination
• In order to ensure that route request (RREQ)
  reaches its destination, sleep interval is set to
  a multiple k of the listening/retry interval
• Once route is established, only nodes that
  forward traffic need to remain active

16

• Adaptive fidelity energy conservation algorithm
  (AFECA)
• Extension of BECA in which nodes adapt their
  sleeping interval depending on the estimated
  network density
• Higher network density ? ?
• Simulation evaluation of BECA/AFECA using ns-2
  with AODV as the routing protocol
• With a minimum sleep interval of 10 seconds,
  overall energy savings was about 35-45
• There was a significant increase in route latency
• Under one second for unmodified AODV
• Between 6 and 10 seconds using AFECA
• Network half-life increases by as much as 50,
  but there is almost no increase in time to first
  node failure

17
Power Control Techniques

• Nodes modify their transmit power to increase
  network capacity and reduce energy usage
• Topology control
• Assign per-node transmit powers to minimize the
  maximum transmit power used in the network while
  maintaining network connectivity
• Refer to Fig. 14.12 of text
• Also has effect of increasing throughput since
  interference is reduced

18
Power Control for Less Communication Interference
Fig. 14.12 of text
19

• Minimum energy routing
• Aims to minimize the total energy consumed in
  forwarding a packet from source to destination
• Sequence of low power transmissions may use less
  energy than a single direct transmission
• Another factor at a given transmit-power level,
  effective transmission ranges may differ in
  different directions due to terrain and obstacles
• Example methods
• Methods based on relay regions
• Given transmitter i and relay node r, relay
  region Ri(r) contains the set of receivers for
  which transmitting via relay node r reduces the
  total energy consumption. Basagni 2004, p. 317
• Power-aware routing optimization protocol (PARO)
• Nodes can nominate themselves as relay nodes
  based on observed traffic behavior

20
Power-Controlled MAC (1/2)

• Antennas can be designed to transmit/receive at
  different power levels
• Spatial reuse using power control
• Each node can be assigned a different maximum
  power
• Each packet (from each node) can be sent with
  different power levels
• More powerful method
• Basic method
• Send RTS and CTS at maximum power
• Based on received signal strengths of RTS/CTS,
  reduce the power of DATA and ACK packets
• Does this help spatial reuse a lot? Why or why
  not?

21
Power-Controlled MAC (2/2)

• RTS/CTS exchange does not prevent hidden-terminal
  problems when heterogeneous power levels are used
• Proposed Solutions
• Use busy tones for virtual carrier sensing
• Each active receiver advertises its noise
  tolerance using pulses in the busy tone channel
  (a separate channel)
• Then use Request-Power-to-Send (RTPS) and
  Acceptable-Power-to-Send (APTS) handshake in the
  normal data channel
• Implies use of a second transceiver
• Dont address the collisions simply send anyway
• Use backoffs and retransmissions to solve the
  problem

22
Minimum-Energy Broadcasting and Multicasting

• Power consumption for communication between two
  nodes at distance r is proportional to p(r) r a
  c, where a gt 2 and c is an overhead constant
• Local shortest path tree (LPST) algorithm Ref
  7 of text, Ch. 4 for broadcasting
• Each node u applies Dijkstras algorithm to find
  the shortest weighted path (weight p(r)) to
  each of its neighboring nodes using only 1-hop
  info
• Each node adjusts its transmission radius
  according to the links in a pruned logical tree
• Each node transmits with a radius equal to the
  longest adjacent link
• Refer to Sec. 4.3 of text for details

23
Maximum Lifetime Routing

• Maximum-lifetime routing, given a set of constant
  rate flows and a fixed network topology, can be
  formulated as a linear programming problem
• Analogous to maximizing minimum lifetime subject
  to multicommodity flow conservation
• The general maximum-lifetime routing problem is a
  much more difficult problem
• Flow-augmentation algorithm
• Hierarchical zone-based method
• Using game theory to determine if a node should
  forward another nodes packet

24
Alternative Approaches

• Battery-energy-efficient (BEE) routing
• Electrochemical behaviors of batteries
• Recovery effect
• Bursty discharge pattern is more efficient than a
  constant-current discharge
• Rate-capacity effect
• Drawing even a small percentage of current
  impulses that exceed the rated current capacity
  of the battery significantly degrades battery
  performance. text, p. 323
• BEE routing uses cost function with conventional
  max-min factor and penalty factor based on the
  above
• Reliable energy-aware routes
• Bit error rates vary depending on the transmit
  power used
• Cost function should reflect the expected cost of
  retransmissions at each link

25
Discussion

• Energy-efficient communication is a multi-faceted
  problem
• Trying to optimize a single facet may lead to
  suboptimal performance with respect to other
  facets, thereby leading to worse node and
  network lifetimes
• Energy consumption and wireless propagation
  models play a central role in the design and
  evaluation of energy-efficient systems
• Direct measurement studies of wireless systems
  are useful in designing and evaluating energy
  management techniques
• Better and more realistic simulation techniques
  and testbed environments are necessary in order
  to effectively evaluate and compare different
  protocols