Resource Allocation in Wireless Communication Networks

1
Resource Allocation in Wireless Communication
Networks

• Xin Liu
• Computer Science Dept.
• University of California, Davis

2
Wireless Communication Networks

• Cellular networks
• WiFi, WiMAX
• Ad hoc networks
• Mesh/community networks
• Wireless sensor networks

3
Resource Management

• Scarce radio resource
• Timing-varying and location-dependent channel
  conditions
• Limited battery power
• Shared medium
• Mobility

4
Research Topics

• Opportunistic scheduling
• Spectrum-agile communication
• Wireless sensor networks

5
Opportunistic Scheduling

• Objective
• Efficient spectrum utilization
• QoS provisioning
• Motivation
• Scarce radio resource
• Timing-varying channel conditions
• Multi-user diversity

6
Channel Conditions

• Decides transmission performance
• Determined by
• Strength of desired signal
• Noise level
• Interference from other transmissions
• Background noise
• Time-varying and location-dependent.

7
Interference and Noise
8
Propagation Environment
9
Time-varying Channel Conditions

• Due to users mobility and variability in the
  propagation environment, both desired signal and
  interference are time-varying and
  location-dependent
• A measure of channel quality
• SINR (Signal to Interference plus Noise Ratio)

10
Illustration of Channel Conditions
11
Performance vs. Channel Condition

• Voice users better voice quality at high SINR
  for a fixed transmission rate
• Data users higher transmission rate at high SINR
  for a given bit error rate
• Adaptation techniques are specified in 3G
  standards.
• TDMA adaptive coding and modulation
• CDMA variable spreading and coding

12
Multi-user Diversity
Scheduling question given this channel
condition, which user should transmit at a given
time?
13
A Greedy Scheduling Scheme

• Always choose the user with the best channel
  condition to transmit
• Improve the spectrum efficiency
• Unfairness among users

Starvation
14
Opportunistic Scheduling

• Basic idea schedule users in a way that exploits
  variability in channel conditions
• Opportunistic choose a user to transmit when its
  channel condition is good.
• Fairness/QoS requirements opportunism cannot be
  too myopic.
• Each scheduling decision depends on
• channel conditions
• fairness or QoS requirements
• Select the relatively-best user

15
System Model

• Time-slotted systems
• Each user has a certain requirement
• TDMA or time-slotted CDMA systems (e.g., IS-856)

16
Notion of Utility

• Uik data rate of user i at time k
• If time slot k is assigned to user i, user i will
  receive a throughput of Uik.
• Measures the worth of the time slot to user i.
• Generalize to the notion of utility
• throughput
• throughput cost of power consumption
• Uik, k1,2,3 is a stochastic process.
• Utility values are comparable and additive.

17
A Framework for Scheduling

• Objective Maximize the sum of all users
  throughput while satisfying the QoS requirements
  of users.
• Scheduling decision depends on
• Channel conditions
• QoS/fairness requirements

18
A Case Study Temporal Fairness Scheduling
19
Objective

• Maximize average system throughput subject to
  the fairness constraints ri.
• System utility
• is the indicator function

20
Scheduling Problem Formulation

• Optimal scheduling problem
• where ? is the set of all policies.
• No channel model assumed
• No assumption on utility functions
• General distributions of
• Users utility values can be arbitrarily
  correlated across time and among users.

21
An Optimal Scheduling Policy

• Choose the relatively-best'' user to transmit
• vi off-sets used to achieve the fairness
  requirement.

22
Parameter Estimation

• We estimate vi based on measurements of the
  channel using stochastic approximation.
• Consider the root-finding algorithm for each
  threshold vi
• vik ? vi with appropriately chosen
• However,

23
Parameter Estimation (Cont'd)

• vik ? vi w.p.1 under appropriate conditions
  (e.g., ak1/k).
• Simulation results show the estimation works
  well.

24
Scheduling Algorithm
25
Case 1 Simulation of a Wireless System

• Fair sharing ri1/N, N is number of active users
• Non-opportunistic scheme round-robin
• Concentrate on the downlink. Reuse factor is 3.
• Consider co-channel interference from first-ring
  neighbor cells
• Consider path loss (Lee's model) and log-normal
  shadowing
• Each user moves in the cell with a certain speed
  and its direction, which can change periodically
  
• 25 users/cell with exponentially distributed
  on-off periods.

26
Utility Values

• Step function - user 1-2
• Linear function - user 3-4
• S-shape function -user 5-8

27
System Performance
28
Conclusions on Opportunistic Scheduling

• Traditional setting performance of system
  depends on average channel conditions.
• Opportunistic setting performance of system
  depends on peak channel conditions.
• Opportunistic gain increases with
• channel variability (over time)
• number of users
• channel independence (across users).
• Current and Future wireless systems
• exploit opportunistic methods (IS-856).

29
Where do We Stand?

• History a successful story, a industry
• Current
• Rapid proliferation
• Policy evolution
• Future
• More spectrum
• Advanced DSP and radio technologies
• Cool applications

An Exciting Area, a Long Way to Go!
30
Recruitment

• I am looking for students
• Self-motivation
• Welcome background in algorithms, optimization,
  probability, etc.

• Thank You!