Dynamic Channel Allocation in MIMO Mobile Networks

1
Dynamic Channel Allocation in MIMO Mobile Networks

• Performance Depends on Traffic and Interference
  Models and MAC Protocol
• Current Research Provides Example for Specific
  Traffic and Interference Models for a TDD Network
• DCA algorithm performance in simple TDD networks
  with Internet-based traffic
• 3GPP/3GPP2 MIMO urban microcell model
• Modified DCA algorithms for multi-sector cellular
  system
• Future Direction
• Increase MAC throughput in Ad Hoc Networks
  exploiting MIMO techniques

2
Noise and Interference Suppression Techniques

• Adaptive Interference Cancellation
• Requires reference data for the interference
• Nonlinear Adaptive Filtering
• Creates reference from error signal
• Adaptive Beamforming
• Selective Interference Cancellation
• Wideband waveforms (OFDM, spread spectrum)
• Co-channel Interference Suppression
• SIC, MUD, MMSE, ZF
• Dynamic Channel Access
• Schedules transmissions to minimize CCI

3
Transition from TDD DCA to Mobile Ad Hoc Networks
Derive Theoretical Traffic Models Measure
Internet based wireless traffic Create binomial
dynamic traffic models
Previous Research
Implement TDD Dynamic Channel Allocation
Rules Pseudorandom, orthogonal, symmetrical
intelligent BS-coordinated allocation
Analyze Measure Throughput Performance Generate
equations results for various traffic models
and DCA algorithms
Repeat for Realistic Cellular Network Integrate
3GPP/3GPP2 Spatial MIMO Channel Across
Multi-Sector Urban Microcell System
Medium Access Control in Ad Hoc Networks Key
differences with the TDD DCA problem, proposed
approach to increase network throughput with MIMO
Ref1 W.Cooper, J.R.Zeidler, R.R.Bitmead,
Modeling Dynamic Channel-Allocation Algorithms in
Multi-BS TDD Wireless Networks With
Internet-Based Traffic,
IEEE Trans. On
Vehicular. Technology, Vol 53, No. 3, May 2004
Ref2 W.Cooper, J.R.Zeidler, R.R.Bitmead,
Dynamic Channel-Allocation Algorithms in TDD
Wireless Networks with Internet-Based Traffic and
Space-Time
Channels,
submitted to IEEE Trans. On Vehicular. Technology
4
UCSD Campuswide 802.11b Network

• Network Traffic Measurements
• November 2002
• 150 official 802.11b network access points
  connected through central router.
• 4000 registered wireless MAC addresses
• 800 active users
• Note Estimated 100 private unregulated 802.11b
  access points in labs, offices and halls of
  residence. (Not included in traffic
  measurements).

5
UCSD Wireless Internet Traffic Analysis

• Internet datagram message logs fields
• Source and Destination IP Address ? Uplink vs.
  downlink identifier
• Source and Destination Port ?
  Internet application type client-server message
  identifier
• Message Size (Octets) ?
  Average message size per application type
• Number of Packets in Datagram ? Average
  packet size per application
• Datagram Interval (1st-last packet) ? Routing
  delay (not used in traffic models)
• Datagram Transmission time ?
  Corresponding client-server message identifier

Note Approximately 76 downlink Rx traffic, 24
uplink Tx traffic , Beta 0.75
6
Dynamic Traffic Model Normalized Flow
7
Wireless TDD Traffic Distribution Models

• Traffic Distribution Models for Asymmetric TDD

8

• Simple 2-BS TDD Interference Model

9
TDD Dynamic Channel Allocation Algorithms

• Pseudo Random Allocation
• Allocation Rule Assign timeslots completely
  independently, without trying to align or order
  requests

• There are 49 allowable traffic request
  permutations. Note Permutations shown in purple
  indicate inter-BS interference.

• Symmetrical Allocation

• Interleaved Orthogonal Allocation

• No conflicting timeslot allocations allowed
• Throughput dependent on uplink-downlink ratio at
  BS.
• Very inefficient assignment although minimizes
  interference, since all users are orthogonal.

• No conflicting timeslot allocations allowed
• Traffic must be symmetric between BSs to achieve
  optimal capacity.
• Increased inter-BS interference vs. interleaved
  orthogonal

10
Overall BS Thruput with Dynamic Traffic
Intelligent
Pseudo-Random
Symmetric
Orthogonal
11
3GPP/3GPP2 MIMO Urban Microcell Channel Model
LOS and NLOS Urban microcell propagation model
(Walfish-Ikegami)

• Mobile Station (per Antenna)
• n 6 multipaths clusters
• Path azimuth N(0,s2)
• s 104.12(1-exp(-0.265log10(Pn)))
• Pn 10(tn zn /10), tn U(0,1.2µS), Zn
  N(0,3dB)
• M 20 subpaths per path cluster
• Fixed azimuth spread 35?

• Base Station (per Antenna)
• n 6 multipaths clusters
• Path azimuth U(-40o,40o)
• m 20 subpaths per path cluster
• Fixed azimuth spread 5o
• PLNLOS 34.53 38log10(d)
• SFNLOS (0, s2), sSFNLOS 10dB

MS Azimuth Spread
BS Azimuth Spread
n 6 paths
m 20 subpaths
12
Intelligent DCA Algorithm w/ ST Channel


13
Intelligent vs. Pseudo-Random DCA Total MS to
Target MS Interference Level
Total MS to Target MS Interference with 0dB
Lognormal Fading, Binomial Traffic, 48 MS per
Sector
Intelligent assignment has much lower total MS to
MS interference level due to fewer colliding Tx
Rx timeslots.
Mean MS to Target MS interference around
-125dBm for Pseudo Random allocation.
MS to Target MS Rx interference level lt -200 dBm
for timeslots when there are no MS co-channel
interferers.
Mean MS to Target MS interference around -180dBm
for Intelligent allocation, due to interfering MS
Tx packets on cell boundary being assigned first
to timeslots at end of timeframe, far away from
MS Target Rx packets located on cell boundary.
MS to Target MS interference around -60dBm when
interfering and Target MS close to each other on
adjacent cells.
14
TDD DCA vs. Ad Hoc Networks MAC

• Comparison
• Similarity goal is to increase throughput by
  controlling interference
• Differences
• random access in Ad Hoc Networks rather than
  dynamic channel allocation
• no central controller to provide
  co-ordination/time synchronization
• Standard Ad Hoc MAC (e.g WiFi)
• Nodes have single omni-directional antennas
• Medium access is contention-based (CSMA-CA)
• RTS/CTS handshake reserves the channel for a
  single pair of users before data
• transmission (collision/interference
  avoidance)
• Issue limited throughput (due to single data
  transmission at a time)
• Improvement through link optimization with MIMO
  techniques

X
C
D
A
B
CTS
RTS
15
RACH Packet Throughput with Antenna Selection
Diversity
Ref W.Cooper, J.R.Zeidler, S.McLaughlin,
Performance Analysis of Slotted Random Access
Channels for W-CDMA Systems in Nakagami Fading
Channels,
IEEE Trans. On Vehicular.
Technology, Vol 51, No. 3, May 2002
16
Incorporating Spatial Information in Ad Hoc
Networks

• Ad Hoc Networks with Directional Antennas
• It has been recently proposed to allow multiple
  simultaneous transmissions, limiting interference
  by use of directionality
• MAC protocols have been modified to accommodate
  directional transmissions / receptions (e.g.
  D-MAC protocol)
• Increased throughput
• Drawbacks
• Applicability in poor scattering environments (is
  there a direction in rich fading?)
• Transmitters must know (from upper layers) or
  acquire (control overhead) the direction of
  intended receiver
• Performance dependent on topology (e.g. worse if
  nodes in straight line)
• Hidden terminal and deafness problems
• Limits redundancy of network for
  reconfigurability

17
Research Directions

• Ad Hoc Networks with MIMO nodes approach
• Allow limited number of mutually interfering MIMO
  transmissions
• Exploit spatial separation to discriminate
  between desired and undesired transmissions,
  using signal processing at the receiver
• (CCI suppression or multi-user detection ?
  multi-packet reception)
• What kind of MIMO Tx Technique?
• In such interference-limited scenarios, transmit
  diversity may be preferable to spatial
  multiplexing
• Spatial multiplexing incurs even more
  interference due to multiple transmitted streams
  per user
• Example
• With 3 antennas at the receiver and MMSE
  suppression, 2 interfering (undesired)
  transmissions are tolerated and 2nd order
  diversity for the desired transmission is
  achieved, by employing Alamoutis STC at the
  transmitter
• Issues
• Network synchronization for fast channel
  estimation
• Power control when no synchronization

18
Project Summary

• Management Plan
• Preselection 11 of the 13 PIs have already
  co-authored papers, 7 have co-advised PhD
  dissertations
• Collaborative simulation and testing
• Frequent interaction with DoD and other
  universities
• Future Reviews
• Educational Outreach
• 16 graduate students
• Significant Interest from Industry
• Workshops/ Short Courses?
• Deliverables

19
Statement of Work STC and Beamsteering (1)

• Determine SNR required for CSI estimation for
  different Doppler spreads
• Develop/simulate new STC structures quantify
  diversity/ rate tradeoffs
• Develop/test adaptive STC structures that allow
  different rates from different antennas
• Develop STC/BF hybrid structures to optimize
  performance/complexity tradeoffs in various CSI
  tactical scenarios
• Develop/implement/test closed-loop partial CSI
  estimation algorithms
• Quantify the capacity of feedback MIMO systems
  relative to the number of feedback bits and
  define performance/complexity tradeoffs
• Quantify the effects of STC on the connectivity
  of adhoc networks

20
Statement of Work STC and Beamsteering (2)

• Develop MIMO channel prediction algorithms/test
  with measured data
• Evaluate performance of antenna diversity with
  noisy channel estimates, correlated fading and
  intentional jamming
• Upgrade existing MIMO channel probing system
  and conduct outdoor measurements at various
  Doppler levels
• Evaluate impedance matching, mutual coupling
  and other performance issues for MIMO antenna
  topologies in realistic tactical ad-hoc network
• Prototype candidate antenna architecture in a
  real-time system
• Implement STC/BF hybrid structures in an ad-hoc
  network

21
Statement of Work Link Scheduling, MAC and
Routing (1)

• Develop dynamic scheduling algorithm to exploit
  STC and beamsteering for unicast, multicast and
  broadcast traffic, based on FAST (Flow Aware
  Scheduling of Transmission)
• Demonstrate increased capacity of FAST using MIMO
  channel estimates and MAC layer link reliability
  statistics
• Develop energy efficient routing protocols
  modeling energy consumption at all levels (RF,
  signal processing and link budget)
• Develop routing algorithms that support neighbor
  discovery and maintain reliable end-to-end data
  delivery in an ad-hoc network
• Develop and test routing protocols that
  utilize directional antennas and maximal path
  disjointness

22
Statement of Work Link Scheduling, MAC and
Routing (2)

• Develop low probability of detection (LPD) /
  anti-jam (AJ) waveforms that support link
  scheduling and routing in tactical environment
• Develop feedback interfaces from transport layer
  to lower layers to select the routing strategy
  that satisfies application layer requirements
• Develop space-time modulation and coding for
  parallel relay nodes and evaluate effects of
  co-channel interference
• Evaluate packet loss rate in a parallel relay
  assisted network and develop optimal routing
  algorithms
• Develop a simulation environment which allows
  evaluation of network capacity using physical
  layer information