CS Colloquium Research Projects in Wireless Communication Networks Xin Liu Computer Sciences Department University of California, Davis.

CS Colloquium

Research Projects in Wireless Communication Networks

Xin Liu

Computer Sciences Department

University of California, Davis

2

Wireless Networks

Cellular systems 1G: analog 2G: digital 3G: data

Wireless LAN IEEE 802.11

Ad-hoc wireless networks Military, emergency, etc.

Wireless Sensor networks

3

Research Topics

Digital signal processing Smart antenna Scheduling Power management Topology management Mobility management Routing (for ad hoc networks) ……

4

Unique Features

Motivated by some unique features in wireless communication systems:

Scarce radio resource Limited power Timing-varying channel conditions Shared media

5

Scarce Radio Resource

Wireline networks High bandwidth and reliable channel Core router: Gbps-Tbps

Wireless systems Limited nature resource (radio frequency) Capacity is limited by available frequency 3G data rate: up to 2Mbps IEEE 802.11b: up to 11Mbps Requirement: spectrum efficiency

6

Power

Battery power is still the bottleneck Important for hand-held equipment Critical for wireless sensor networks

What can we do? Power management --- use the available

power efficiently

7

Channel Conditions

Decides transmission performance

Determined by Strength of desired signal Noise level

Interference from other transmissions Background noise

Time-varying and location-dependent.

8

Interference and Noise

Noise

Interference

Interference

Interference

DesiredSignal

9

Propagation Environment

Shadowing

Multi-path Fading

Strong

Weak

Path Loss

10

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)

11

Illustration of Channel Conditions

Based on Lee’s path loss model, log-normal shadowing, and Raleigh fading

12

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

13

Shared Media

Shared media: everyone can hear each other

Can hurt Can help Multi-user diversity

14

Interference

Noise

Interference

Interference

Interference

DesiredSignal

15

Helper

Relay:

Coherent Relay:

16

Multi-user Diversity

Radio tower

Laptop

Radio tower

Different users see different channels at different time

17

Opportunistic scheduling

Motivation: Spectrum efficiency Time-varying channel

conditions Multi-user diversity

Question: how to handle channel variability?

18

Opportunism

Traditional design: point to point Channel variability: source of unreliability

Opportunism: embrace channel variability Multiple users share resource Exploits favorable channel conditions.

19

Myopic Opportunism

Greedy algorithm: best user to transmit Good throughput Unfairness

Starvation

20

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 greedy.

Each scheduling decision depends on channel conditions fairness or QoS requirements.

21

System Model

Time-slotted systems

Each user has a certain requirement.

TDMA or time-slotted CDMA systems (e.g., IS-856, known as Qualcomm HDR)

Both uplink and downlink.

Time

22

Overview

EstimateUtility

Values

ApplyScheduling

Polity

UpdateParameters

MeasureChannel

Conditions

iU

iV

23

Performance Measure

Based on utility value Reflects channel condition. Uik : utility value of user i at time k .

If time slot k is assigned to user i, user i will receive a utility value of Uik.

Measures the worth of the time slot to user i. Examples of utility:

Throughput Throughput – cost of power consumption.

Utility values are comparable and additive.

24

Utility Values

{Uik, k=1,2,3…} is a stochastic process.

25

A Framework for Opportunistic Scheduling

Objective: Maximize the sum of all users’ utility values while satisfying the QoS requirements of users.

Scheduling decision depends on: Utility values (reflecting channel conditions) QoS/fairness requirements.

26

A Case Study: Temporal Fairness Scheduling

Uk1

Uk

N

Uk2

?)( ),,...(Given

},...2,1{)(

policy Scheduling :

users ofNumber :

1 0, :

1

1

U

U

QUU

NQ

Q

N

rrr

kN

k

N

i iii

time. theofportion getsuser each :fairness Temporal ir

27

Objective

Maximize average system utility subject to the

fairness constraints ri. System utility:

28

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 correlated.

29

An Optimal Scheduling Policy

Choose the ``relatively-best'' user to transmit.

vi* : “off-sets” used to achieve the fairness requirement.

30

Property

Improves performance for all.

Gain depends on channel variability.

A certain level of average utility guarantee for each user.

31

Scheduling Gain

Opportunistic scheduling gain increases with

channel independence (across users)

channel variability (over time)

number of users.

32

System Performance

1.01] 0.99 1.00 1.01 1.01 0.99 1.00 99.0[ :ratio Fairness

33

Joint Scheduling and Power Allocation

Joint scheduling and power allocation: intercell-interference management.

Interference limits the system capacity. Power allocation: interference management. Opportunistic scheduling: multi-user diversity. Two decision variables:

which user how much power.

34

Objectives

Objective 1: minimize total transmission power guarantee a minimum-utility for each user.

Objective 2: maximize net utility

tradeoff between throughput and transmission power (interference to other cells).

guarantee a minimum-utility for each user.

35

A To-do List

May induce variability if needed. Can be used in distributed manners.

Many to many Large sensor networks

Real-time traffic Multi-carrier systems A different design aspect Problems in information theory Future wireless systems: exploit opportunistic

methods (IS-856).

36

Wireless Sensor Network Potential

Micro-sensors, on-board processing, and wireless interfaces all feasible at very small scale can monitor

phenomena “up close”

Will enable spatially and temporally dense environmental monitoring

will reveal previously unobservable phenomena

Seismic Structure response

Contaminant Transport

Marine Microorganisms

Ecosystems, Biocomplexity

Ref: based on slides by D. Estrin

37

Enabling Technologies

Embedded Networked

Sensing

Control system w/Small form factorUntethered nodes

ExploitcollaborativeSensing, action

Tightly coupled to physical world

Exploit spatially and temporally dense, in situ, sensing and actuation

Ref: based on slides by D. Estrin

38

Challenges

By no means this is a complete list: Self-configured

Random deployment of sensor networks Long-lived sensor systems

Sensors have very limited battery power Reliability

Harsh environment Unreliable sensors

Cost Scalability Massive data

Compression and aggregation Time synchronization, data query, localization, storage, etc.

39

A Random Deployed Sensor Network

GATEWAY

MAIN SERVER

CONTROLCENTER

40

Topology control

Many-to-one communication Unbalanced load Uneven power consumption “Important” nodes in the route die quickly

Possible approaches More power at closer nodes Data compression and aggregation

41

The Problem

Objective: minimize # of sensors needed to build a sensor network that covers a given area for a certain amount of time.

Communication consumes a lot of power

R: rate, D: distance between transmitter and receiver

Put nodes with heavier load closer

52, RDP

1

84.0,1,2,1422

411

2121

DRDR

DDRR

42

Approach

Non-trivial: sensor placement, routing, power management

To consider: Linear and planar network Random and non-random topology Other power consumption

Approaches: Understand fundamental principles Build practical solutions

P1

P2

43

Coverage and Connectivity

44

Coverage and Connectivity

Traditional work: full coverage and connectivity, K-coverage, etc.

Our objective: Cover and connect a large portion of the area Quantify the size of uncovered area How many nodes needed What is the density needed

45

Cost and Reliability

Layered structure More expensive nodes with more functionality

Objective: minimize the total cost, including different types (cost) of nodes, while maintaining the desired performance

Reliability important, especially for large scale network nodes damages, out of power, etc.

46

Parking Lot Patrol Problem

Sensors on parking meters Build a wireless sensor network to report

illegal parking Patrolman to find the reported events

Applications: Border patrol Speeding monitoring

47

What Do We Stand?

History: a successful story, an industry of $$$$$$ Current: Policy re-examination underway

Increased unlicensed spectrum allocation Exploration of “underlays”, e.g., UWB Exploration of “overlays”, e.g., opportunistic use of

committed but unused bandwidth Future:

more spectrum better ratio equipment, DSP technologies, longer

battery life Better networks Cool applications

48