Understanding Channel Access Control in Ad Hoc Networks with MIMO Nodes

1
Understanding ChannelAccess Control in Ad Hoc
Networks with MIMO Nodes

• J.J. Garcia-Luna-Aceves
• Hamid Sadjadpour
• Marcelo Carvalho
• Renato Moraes
• Xiaohui Yu
• University of California, Santa Cruz

Not funded by STC MURI
2
Outline

• Quick summary of research output
• Recent results on the capacity of ad hoc networks
  with MIMO nodes
• Recent results on the performance of MAC
  protocols with MIMO nodes
• Summary of research directions.

3
Summary of Research Output

• Papers accepted for publication
• 1-X. Yu, R. D. Moraes, H.R. Sadjadpour and J.J.
  Garcia-Luna-Aceves,, "Capacity of MIMO Wireless
  Ad Hoc Networks," IEEE WirelessCom 2005.
• R. D. Moraes, H.R. Sadjadpour and J.J.
  Garcia-Luna-Aceves, "A New Communication Scheme
  for MANETs," IEEE WirelessCom 2005.
• R.M. Moraes, H.R. Sadjadpour and JJ. Garcia-Luna,
  "Mobility-Capacity-Delay Trade-off in Wireless Ad
  Hoc Networks," Elsevier journal on ad hoc
  networks (to appear).
• Papers submitted for publication
• R. D. Moraes, H.R. Sadjadpour and J.J.
  Garcia-Luna-Aceves, "Opportunistic cooperation A
  new approach for scalable mobile ad hoc
  networks", submitted to INFOCOM2006.
• R. D. Moraes, H.R. Sadjadpour and J.J.
  Garcia-Luna-Aceves, "Ergodic capacity of MIMO
  MANETs with opportunistic cooperation," submitted
  to INFOCOM2006.
• R. D. Moraes, H.R. Sadjadpour and JJ.
  Garcia-Luna-Aceves, "Opportunistic cooperation A
  new communication scheme for MANETs," submitted
  to Asilomar 2005.
• R. D. Moraes, H.R. Sadjadpour and JJ.
  Garcia-Luna-Aceves, "Opportunistic Cooperation A
  new approach for scalable Mobile Ad hoc
  networks," submitted IEEE Transactions on
  Information Theory.
• R. D. Moraes, H.R. Sadjadpour and JJ.
  Garcia-Luna-Aceves, "Taking full advantage of
  Multiuser diversity in Mobile Ad hoc networks,"
  submitted to IEEE Transactions on Communications.
  
• M. Carvalho and J.J. Garcia-Luna-Aceves,
  Modeling Ad hoc Networks with Directional
  Antennas, submitted to INFOCOM 2006.

4
Summary of Research Output

• Ph.D. sudents
• Renato Moraes and Marcelo Carvalho will
  be graduating before the end of the Fall 05
  quarter.

5
Key Research Areas(from kickoff meeting)

• Help the understanding of the fundamental
  limitations to the scaling of ad hoc networks
  with cross-layer optimization (number of nodes,
  energy consumption, bandwidth utilization).
• Study the impact of the physical layer on
  communication protocol stack.
• Modular protocol stacks to bridge the gap between
  the applications of large ad hoc networks and the
  new hardware available with ST coding and other
  technologies. Emphasis on the MAC layer.
• Complementing MURI research with other ongoing
  research work at UCSC

6
Network Capacity
7
Motivation for MURI Research (from kickoff
meeting)

• How can we improve the throughput performance by
  sending packet via least resistance paths?
• We need cross layer optimization! Constant hop
  count and constant interference per hop

8
Multi-copy Forwarding (from kickoff meeting)
9
Conventional Communication Models

• Conventional communication models are based on
  competition-driven approaches.
• Two-phase forwarding model by GT clearly
  increase the capacity!

Competition leads to scaling problems (G-K
model) Uses storage as bandwidth, a form of
cooperation
10
New Vision Opportunistic Cooperation

• Why should nodes fight for the channel?
• Can we exploit MIMO processing to allow nodes to
  cooperate?
• How much do we gain?
• What type of cooperative protocols should we be
  designing?

Many-to-Many Communication Model
Start by exploiting MIMO in the two-phase
delivery model of G-T.
11
Opportunistic Cooperation with MIMO Systems

• Assume
• n mobile nodes
• Each square cell has same area size (a_cell)
• Each node picks an arbitrary destination
• Uniform mobility model.
• A node transmits at power P to another node.
• Each node has M antennas
• Sources do not have CSI
• Simple path propagation model.
• Node knows its location and the freq map

Consider a simplified FDMA/MIMO system as an
instantiation of the many-to-many communication
concept.
MISO 9A frequency bands
12
Ergodic Channel Capacity(submitted to Infocom 06)
13
Capacity Behavior
14
Performance Analysisof MAC Protocols
15
Summary of Work

• Used our interference matrix model as the
  baseline model
• M. Carvalho and J.J. Garcia-Luna-Aceves, A
  Scalable Model for Channel Access Protocols in
  Multihop Ad Hoc Networks,'' Proc. ACM Mobicom
  2004, Philadelphia, Pennsylvania, Sept. 26--Oct.
  1, 2004.
• Developed first analytical model of wireless ad
  hoc networks that considers realistic
  antenna-gain patterns and applied it to DVCS
  protocol.
• Developed first analytical model of wireless ad
  hoc networks that considers MIMO Space-Time
  Coding (STC) technology.
• Space-time coding based on space-time block
  coding (STBC)
• Alamouti scheme for MT 2 transmit antennas and
  MR receive antennas
• New Markov model for the IEEE 802.11 DCF MAC
• Impact of carrier-sensing mechanism.
• Impact of transmission errors in both control and
  data frames previous models assumed errors only
  in control frames (data frames were assumed to be
  transmitted error-free!).
• Applicable to other back-off schemes with carrier
  sensing.

16
Modeling Approach

• PHY
• The probability of successful reception of a data
  packet and its acknowledgment, based on effect
  from all transmissions (which depend on
  scheduling by the MAC) and PHY parameters
• MAC
• Scheduling rates based on feedback from the PHY
  regarding the success of transmissions and the
  state of the channel (e.g., busy, idle)
• Topology
• Consider the effect of all nodes based on where
  they are and their transmission attempts
• Linearize the problem exploiting the fact that
  any MAC protocol will tend not to schedule
  transmissions when feedback from the PHY
  indicates unsuccessful transmissions or the
  channel is busy.

17
Modified IEEE 802.11 DCF MAC
New Markov Model
Errors in control and data frames are considered
previous models disregarded errors in data
frames Single retry counter it increments if
either a control or a data frame transmission is
unsuccessful Previous models did not consider
the IEEE 802.11 finite retry limit and carrier
sense activity jointly impact of physical layer
Linearized transmission probability
18
Preliminary Results (MIMO)
Tx 1, Rx 1 (SISO) Tx 2, Rx 1 (MISO) Tx
2, Rx 2 (MIMO) Tx 2, Rx 4 (MIMO)
Rician Fading
Random Topologies
Case Studies
100-Node Topology
Use of multiple antennas without the need of
fine-tuning MAC still increases throughput in
ad hoc networks MISO and MIMO systems are robust
to bad channel conditions (when K
decreases) Throughput gains compared to SISO at
K 5 MISO 65 (50 nodes) and 160 (100 nodes)
MIMO (2x2) 86 and 220, respectively MIMO
(2x4) 113 and 285, respectively Higher gains
for MIMO due to diversity and array gains, as
opposed to MISO, which gives only diversity
gain. MIMO with 2 transmit and 2 receive
antennas presents the best trade-off between
throughput and system complexity
19
Results for Directional Antennas(Directional
Virtual Carrier Sensing)
Simulation versus Analytical Model 10 random
100-node topologies
Realistic Antenna Gain Patterns
20
Realistic versus Simplified Antenna Gain Models
Throughput Comparison 10 random 100-node
topologies
Pie-slice Antenna Gain Model
21
Next Steps
22
Next Steps

• Network capacity
• Analyze fundamental trade-off limits of
  throughput, delay, and receiver complexity under
  opportunistic cooperation for MIMO systems
• We need a lower bound for capacity.
• Consider different instantiations of
  opportunistic cooperation for MIMO systems
• Consider different mobility models and static
  networks.
• Performance analysis
• Compare the performance of cooperative MIMO MAC
  protocols with the performance of
  non-cooperative MIMO MAC protocols.
• Compare the performance of contention-based and
  schedule-based MIMO MAC protocols.
• Study the effect of PHY parameters on MIMO MAC
  protocols.
• Study the effect of opportunistic cooperation
  parameters.

23
Next Steps

• Subsequent
• Develop MIMO MAC protocol(s) based on the
  opportunistic cooperation framework.
• Address the interaction between MAC and network
  layer.
• Develop cross-layer designs that exploit
  opportunistic cooperation in MIMO systems.
• Example Forwarding taking advantage of MIMO MAC

24
Thanks!
25
Transceiver
26
Ergodic Channel Capacity
27
Ergodic Channel Capacity(submitted to Infocom 06)
28
Interference Effect
29
Interference Effect
30
Interference Effect
31
Space-Time Block Coding (STBC)The Alamouti
Scheme

• Scheme supports maximum-likelihood detection
  based on linear processing at the receiver (as
  opposed to space-time trellis coding (STTC),
  which has exponential complexity)
• Part of W-CDMA and CDMA-2000 standards
• Channel information not necessary at the
  transmitter side
• Channel assumed to be known at the receiver side
  (easier to obtain)
• Energy evenly divided among transmit antennas
  no extra power required!
• Basic Alamouit scheme allows MIMO systems with 2
  transmit antennas and MR receive antennas.

32
MIMO Case 2 Tx and 2 Rx Antennas
Channel Matrix
Signals received at antenna array over
consecutive symbol periods
are zero-mean AWGN samples
Receiver forms the signal
is orthogonal irrespective of channel
realization, i.e.,
If
we get
where
The effective received symbols are
with received SNR
33
Bit Error Rate under Multipath Fading

• Because

the received SNR is equal to the sum of SNRs on
each path

• SNR shape suggests the method of the Moment
  Generating Function (MGF) for computation of the
  symbol error probability under multipath fading

• Example DBPSK modulation under Rician fading

K is the Rician factor ratio of the power of
LOS signal to the power of NLOS signals
is the average SNR extracted from each path
(dominated by path-loss propagation effects only)