Neighbor-Aware Control in Ad Hoc Networks

1
Neighbor-Aware Control in Ad Hoc Networks

• Lichun (Luke) Bao
• Dissertation Defense
• University of California, Santa Cruz

2
Dissertation Committee

• Prof. J.J. Garcia-Luna-Aceves (Chair and Advisor)
• Prof. Katia Obraczka
• Prof. Patrick Mantey

3
Presentation Agenda

• Motivations
• Neighbor-aware contention resolution
• MACs using omnidirectional antennas
• MACs for unidirectional networks
• MACs using directional antennas
• Topology management
• Contributions and future work.

4
Motivation

• Contention resolution mechanisms
• On-demand (contention-based)
• MAC protocols (ALOHA, CSMA, CSMA/CA RTS/CTS
  schemes)
• Topology control (random election)
• Problem run-time control overhead
• Scheduled
• MAC protocol (UxDMA global topology)
• Schedule exchanges for setup.
• Problem background control overhead
• NCR with minimum topology

5
Presentation Progress

• Motivation
• Neighbor-aware contention resolution
• MACs using omnidirectional antennas
• MACs for unidirectional networks
• MACs using directional antennas
• Topology management
• Contributions and future work.

6
NCR (Neighbor-aware contention resolution)

• Assumptions
• Topology information contenders (two-hop
  neighbors in MANETs)
• Time synchronized between contenders
• Problem formulation
• In each time slot, how can an entity elect
  itself without conflicts from its contenders?

7
NCR Specification

• 1. Assign a priority to each entity using the
  message digest of its identifier and the current
  time slot number.
• Random, unique to each entity (fairness)
• 2. An entity is entitled the winner if it has
  the highest priority among its contenders.
• Conflict-free (deadlock free)

8
Presentation Progress

• Motivations
• Neighbor-aware contention resolution
• MACs using omnidirectional antennas
• MACs for unidirectional networks
• MACs using directional antennas
• Topology management
• Contributions and future work.

9
Channel Access in Ad Hoc Networks

• Network modeling
• Independent identical communicating and computing
  nodes.
• Communication happens over multi-hop.
• Time synchronized, and channel time-slotted.
• Contention modeling
• One hop neighbors directly shared the channel.
• Two hop neighbors hidden interfering source.
• Interference outside transmission range ignored.

10
Networks with Omni-Directional Antennas

• Antenna modeling
• Antennas have fixed transmission range
• Signal propagation in all directions
• Circular coverage of one-hop neighborhood
• Contenders of a node
• One-hop and two-hop neighbors
• Channel multiplexing technology
• Code-division using direct sequence spread
  spectrum (DSSS)

11
Channel Access Protocols1 NAMA Node activation
multiple access

• Require broadcast to all one-hop neighbors
• Nodes are the competing entities
• Contenders are one- and two-hop neighbors

12
Channel Access Protocols2 LAMA Link activation
multiple access

• Require unicast to a one-hop neighbor
• Nodes are competing entities
• Signals are scrambled with codes assigned to the
  receivers
• Contenders are one-hop neighbors of a node and
  its receiver

13
Channel Access Protocols3 PAMA Pair-wise
activation multiple access

• Require unicast to a one-hop neighbor
• Directional links are competing entities
• Signals are scrambled with codes assigned to the
  transmitters
• Contenders are incident links of the end-points
  of a link

14
Channel Access Protocols4 HAMA Hybrid
activation multiple access

• Allow broadcast to all one-hop neighbors and
  unicast to a one-hop neighbor
• Nodes are competing entities
• Signals are scrambled with codes assigned to the
  transmitters
• Contenders are one- and two-hop neighbors

15
Channel Access Protocolsgt Performance analysis

• Network modeling
• Uniformly distributed over infinite plain
• Node density ? , transmission range r .
• The number of nodes k over a given area S
  follows Poisson Distribution

16
Channel Access Protocolsgt Activation probability
of a node
17
Channel Access Protocolsgt Comparing the
activation probabilities
18
Channel Access Protocolsgt Comparing with CSMA
and CSMA/CA
19
Simulations Results and Comparisongt Throughput
in fully-connected networks
20
Simulations Results and Comparison gt Delay in
fully-connected networks
21
Simulations Results and Comparisongt Throughput
in multi-hop networks
22
Simulations Results and Comparison gt Delay in
multi-hop networks
23
Presentation Progress

• Motivations
• Neighbor-aware contention resolution
• MACs using omnidirectional antennas
• MACs for unidirectional networks
• MACs using directional antennas
• Topology management
• Contributions and future work.

24
Neighbor Protocol (Need)

• Purpose
• propagate neighbor updates, time synchronization
• Cannot be based on NCR or TSMA
• Requires a priori topology information.
• Only efficient way
• Random access.
• Broadcast.
• No acknowledgement why? Efficiency, broadcast.
• Use retransmission to improve reliability.

25
Neighbor Protocol (Method)

• Insert random access section after scheduled
  access
• Send short signal frames carrying neighbor
  updates (256 bytes).
• Problem formulation
• How to regulate interval t and number n of
  retransmissions to deliver a piece of information
  with given (high) probability p with the least
  delay.

26
Neighbor Protocol (Results)

• Reliability deliver-probability p 99.
• Retransmission interval t 1.44N only depends
  on N (the number of two hop neighbors).
• Number of retransmission n 6.77 only depends
  on p .
• Suppose 2Mbps bandwidth, 2 second delay, 20
  two-hop neighbors random access sections cost
  9.6 of the channel bandwidth.

27
Presentation Progress

• Motivations
• Neighbor-aware contention resolution
• MACs using omnidirectional antennas
• MACs for unidirectional networks
• MACs using directional antennas
• Topology management
• Contributions and future work.

28
Networks with Unidirectional Links

• Antennas are omnidirectional with different
  transmission ranges, capable of code-division
  channelization (DSSS)
• Unidirectional network properties
• Can not provide two-way handshakes
• Network may still partition inclusive cycle of
  unidirectional link is required for two-way
  communication ULPC

29
Link-State Routing with Unidirectional Links

• Unidirectional link
• Link (a,b) is unidirectional if link (b,a)
  non-exists.
• ULPC (Unidirectional Link-state Routing Protocol
  with Propagation Control)
• Need to maintain the inclusive cycle of a
  unidirectional link when using it in routing
• The neighbor protocol for ULPC maintains partial
  topology graph for the discovery
• Only utilize links with small inclusive cycles to
  reduce control overhead limited propagation

30
Channel Access Protocols1. NAMA-UN NAMA for
unidirectional networks

• Node a is the Upstream-only neighbor of node b if
  link (a,b) has no inclusive cycle.
• Node a inadvertently interferes at node b
• Collision avoidance
• Code-division channelization assign codes to
  transmitters by priority.
• Dont transmit to B on A s code when node A is
  possible to transmit.

31
Channel Access Protocols2. PAMA-UN PAMA for
unidirectional networks

• Links are the contending entities
• Avoid colliding with Upstream-only neighbor of a
  node

32
PANAMA NAMA-UNPAMA-UN

• Provides both broadcast and unicast
• 25 time slots for NAMA-UN, 95 time slots for
  PAMA-UN
• Compare with UxDMA that uses global topology
  information for scheduling
• Factors
• Transmission range variations
• Ratio of usable unidirectional links
• Traffic types and portion unicast and broadcast

33
Simulations Delays
34
Simulations Throughput
35
Presentation Progress

• Motivations
• Neighbor-aware contention resolution
• MACs using omnidirectional antennas
• MACs for unidirectional networks
• MACs using directional antennas
• Topology management
• Contributions and future work.

36
Networks withDirectional Antennas

• DSP advances enable space-time processing using
  multiple antenna elements directional
  transmission and direction-sensitive reception
• MIMO (multiple input multiple output) becomes
  practical MBAA (Multi-Beam Adaptive Array)
  motivates MAC research
• Benefits reduced CCI, multipath fading, higher
  throughput.

37
Communication with MBAA Antennas

• Issues when using directional antennas
• Neighbor tracking for directional transmissions.
• Who transmits, and who listens coupling.
• Node d has two transmissions, node b is ready
  for two receptions.

38
Network Assumptions

• Antenna system MBAA
• Beam width pencil (10) fan (120 )
• Tx or Rx, not both.
• K simultaneous Tx or Rx.
• Neighbor position profiling requirements
• Accurate for aiming antenna beam
• Yet holds for a while to avoid volatility

39
Neighbor Position Profiling

• Azimuth of a is cut into 360/(ß/2) 720/ß
  sections.
• Two adjacent sections form a group.
• Node c sits in overlapping two groups Ac2,3,
  b in Ab1,2, d in Ad3,4 w.r.t node a.
• Antenna beam pointing to c interferes at b and d.
• How? Easy to compute
• Ac ?Ab ?F, and Ac ?Ad ?F
• Cannot activate (a,c) and (a,b) simultaneously.

40
Channel Access Protocolgt ROMA Receiver-Oriented
Multiple Access

• Require unicast s to multiple one-hop neighbors
• Links are competing entities
• Contenders are incoming links at the receivers
• Steps
• Receiver
• Sort incoming links according to their
  priorities.
• Select top K of the sorted links for reception.
• Transmitter i
• Compute top K active incoming links of each
  one-hop receiver, from which derive all active
  outgoing links of itself.
• Select K of the active outgoing links for packet
  transmissions.

41
Simulations (Assumptions)

• Static topology for algorithm scheduling
  performance only.
• Two topology scenarios
• Fully connected (5, 10 nodes)
• Randomly generated topology (100 nodes on
  1000X1000 square torus with Tx range 200, 400).
• MBAA beam width 30.
• Number of beams 1, 2, 4.
• Packet arrival Poisson.
• Buffer per neighbor 20 packets.

42
Simulations (Throughput)

• Polygons ROMA
• Others UxDMA
• Unified framework for graph coloring.
• Polynomial algo.
• Adapted to handle MBAA.
• ROMA has higher throughput Why?

43
Simulations (Delay)

• ROMA has lower delay in any scenario because of
  its higher throughput.

44
Simulations (Packet Drop-rate)

• Maximum drop rate is one.
• The drop rate rises up later in ROMA than in
  UxDMA.

45
Presentation Progress

• Motivations
• Neighbor-aware contention resolution
• MACs using omnidirectional antennas
• MACs for unidirectional networks
• MACs using directional antennas
• Topology management
• Contributions and future work.

46
Topology Managementin Ad Hoc Networks Goals

• Virtual Overlay Topology Maintenance
• Less topology information presented to routing.
• Less topology updates due to mobility.
• Energy-Awareness
• Less nodes awake for communication.
• Load-balancing the higher the energy left, the
  more responsibilities for data forwarding.
• Basic Approach Clustering and interconnecting.
• Why not power control? Interference.
• Election via dynamic nodal priority assignment.

47
(No Transcript)
48
Topology Management by Priority Ordering
Assumptions

• Time synchronized
• Time counted by time slot and epoch
• Each time slot 100 millisecond.
• Each epoch 600 time slots 1 minute.
• Each node knows
• Topology within two hops clusterhead, doorway
  and gateway elections.
• Nodal speed stability.
• Nodal energy level energy-awareness.

49
Topology Management by Priority Ordering
Priority

• Willingness to join virtual topology
• Low energy, high mobility less willingness.
• Nodal priority for a node
• Is the message digest of the node identifier and
  the current time epoch, multiplied by its
  willingness value.
• Changes every epoch at unique starting point.
• Election Algorithms
• Nodes with higher priorities than their
  contenders compose virtual topology.

50
Topology Management by Priority Ordering
Election

• Clusterhead election a node that has the highest
  priority among
• The one-hop neighbors of itself
• The one-hop neighbors of one of its one-hop
  neighbors
• Gateway a node connecting clusterheads
• The maximum distance between clusterheads are
  three. Gateways are insufficient for
  connectivity.
• Doorway election a node extending the reach of a
  clusterhead

51
(No Transcript)
52
Simulation and Comparison

• Other clustering heuristics
• OPTIMUM least clusterheads.
• Lowest ID use ID instead of priority.
• Max Degree select nodes with high degree.
• MOBIC least neighbor signal strength variation.
• Load balance based on Lowest ID
• Compare
• Simulation duration.
• Combined metric the product of energy
  utilization (awareness), the number and the
  change rate of clusterheads (stability).

53
(No Transcript)
54
Presentation Progress

• Motivations
• Neighbor-aware contention resolution
• MACs using omnidirectional antennas
• MACs for unidirectional networks
• MACs using directional antennas
• Topology management
• Contributions and future work.

55
Contributions

• NCR algorithm using local topology information,
  and derived
• Four MACs for networks with omnidirectional
  antennas
• One routing protocol and two MACs for networks
  with unidirectional antennas
• One MAC for networks with directional antennas
• Topology management mechanism
• Neighbor protocol

56
Publications

• Two MOBICOM papers
• MACs using omnidirectional antennas (2001)
• MAC using directional antennas (2002)
• One ICNP paper
• Hybrid MAC using omnidirectional antennas (2002)
• Two journal papers
• JPDC 2002, MONET 2002
• Six other conference/workshop papers
• IC3N99,MoMuC00,MILCOM00/01,DialM01,NET02

57
Future Work

• Apply the neighbor protocol in wireless sensor
  networks
• Compare with TSMA, CSMA, 802.11
• Explore TMPO derivatives
• Unicast routing
• Multicast routing
• Power saving radio and MACs
• Flow oriented MAC

58
Acknowledgement

• My appreciations for the work of the dissertation
  committee
• Fellow CCRG members (Marc, Chane, Yu, Soumya,
  Long )
• The support from my wife and parents
• The funding from various agencies through J.J.