CDMA/IP-based System for Interoperable Public Safety

Radio Communications  

 

Xin Wang

Director: Wireless Networking and Systems Lab (WINS)Department of Electrical and Computer Engineering

Stony Brook Universitywww.ece.sunysb.edu/~xwang

Problems in Public Safety Systems

Two main factors limiting the reliability and availability of public safety systems: – Different agencies use incompatible systems (different

frequencies, different modulation or encoding, etc). – Spectrum is limited and fragmented.

Problems of limited spectrum and incompatibility:– Can not interoperate– Cannot support wideband data and video communications

Real-time access to mug-shots, finger-prints, crime-scene Fire-fighting, crowd- and prison control

– Cannot share data among agencies

Short-term Solutions

Use dispatching or switch center to manually relay signals betweens systems– Requirements

Interfaces to all potential systemsCoordination and involvement of all public safety

agencies

– ChallengesScalability when allocating new frequency bandProprietary nature of public safety system

Long-term Solutions

Develop modular and scalable systems– Individual agencies can acquire and expand

their own wireless systems without compromising compatibility

– Cost offset: sharing the radio infrastructure from various agencies in a region

Use of more efficient radio technologies, especially for new frequency bands

CDMA/IP-based Wireless Systems

CDMA – Easy of deployment, higher capacity, improved quality,

greater coverage, increased privacy and talk time

IP interface between different systems– Allowing the interoperability of different bands– Sharing the networks independent of access

techniques– Easy of supporting new radio bands and new IP-based

technologies while supporting existing systems– Deployment of off-the-shelf and third-party products

Multimedia, location tracking, encryption, VPN

Future Network Architecture

-Wireless Local Area Networks(WLANs)

-

Bluetooth

Wireless Personal Area Networks(WPANs)

Cellular

Micro Base Station Wireless

Gateways

Base Station

Multimedia & MessagingServer

Content

Locationservice

IP RadioAccessNetwork(IP RAN)

Performance &Services

Radio Hub

Internet

A Sharing and Connection Structure

BS i

BSm

RNC

(PSTN)

BS j

Internet

IP RadioAccessNetwork(IP RAN)

PAG

Area 1 Area 2

IP RadioAccessNetwork(IP RAN)

RNC

PDSN/SGSNPDSN/SGSN

RNC

Public Switched Telephone Network

(Radio Network Controller)

(Packet Access Gateway)

BS k

(Packet Data Serving Node)

Benefit of IP RAN

More scalable, reliable, and cost effectiv– Instead of linking individual agency to

switching center through private or leased lines

Enable packet-based transportation– New applications

Statistical aggregation– High bandwidth utilization, reduced cost– Support both wire-line and wireless

Requirements of Public Safety System Round clock availability, secure and private

communications

Quality of services (QoS) guarantee– Voice (low delay and jitter)– Data (high throughput)– Video (QoS and throughput)

Maximize resource usage under scarce spectrum

Efficient resource management while guaranteeing – Availability, emergency, QoS

Challenges: Air Interface

Support transmission quality – Control power and rate to achieve target Eb/Io

Power and rate allocation for circuit-based transmission (e.g., multimedia)

Adapt rate of elastic data through scheduling Admission control for guaranteeing quality of on-going

transmissions

More efficient use of spectrum – Integrated support of various traffic

real-time circuit-based and elastic packet-based

Challenges: IP-based Backhaul

Traffic in RAN is different from general Internet– Significant amount of traffic is delay sensitive

Voice, radio frames involved in soft handoff

Majority of handoffs involve RAN– Interruptions during hard handoffs– Delayed handoffs and resource wastage during soft

handoffs– Reservation needs to be quick

Radio frame may contain both data and control– Loss and delay of control impact transmission, and

reduce air interface capacity

Proposed Work

Resource management for air interface Scalable backhaul management

Many interactions:

Resource allocation across multiple network layersEffect of air interface management and user mobility on RNAEffect of resource management in RAN on the air interface

Multicasting support: group communications Simulator design

Resource Management for Air Interface

Goal– Serve both circuit-based delay sensitive applications

and packet-based high speed data application– Support both user-to-user unicast and one-to-many

multicast for group communications

Approaches: Cross-layer– Physical layer: power control, rate control– Link layer: scheduling– Network layer: admission control

Rate Control

Basic rate control methods:– Fixed channel continuous transmission

Vary processing gain Assigning multiple codes

– Time-slotted scheduling Allocate different number of time slots Allocate different number of codes

Supporting connectivity and availability– Reduce video resolution, reduce rate of elastic data

Different tradeoffs– Combating the reduction of Eb/Io: throttling the source-coding rate

or increasing the transmission power– Allowing for increasing bit error for less critical data– Apply more efficient error-resilient coding algorithms

Power Control

Optimal power allocations: different types of traffic, different transmission formats

Power sharing among real-time and non real-time traffic– Fixed rate transmission: iterative power control to

find the minimum power to guarantee the received quality

– Increased power for real-time traffic (increased load, or bad channel)

Reduce power for elastic data traffic Allocate more time slots to delay sensitive packet

scheduled data

Packet Scheduling

Support different QoS – Literature work only considers maximize

total throughput, cannot meet public safety requirement

Study tradeoffs between time-slotted scheduling and fixed-channel continuous transmissions. Feature of scheduling:– Pros: More efficient resource usage and overall

higher throughput, throughput gains from multi-user diversity

– Cons: complex in guaranteeing quality

Adaptive scheduling– Increase data rate when system load is low

Admission Control

Adaptive admission control for integrated traffic– Consider both circuit and packet transmissions– Cannot guarantee quality by purely scheduling

Different power for different usersVarying power for the same user due to varying

channel conditions and traffic rate

Prioritize handoffs– Consider both soft handoff and hard handoff

Study connection level performance

Backhaul Resource Management

Effective and scalable traffic engineering

Efficient handoffs

Scalable Traffic Engineering

Aggregate resource reservation and traffic multiplexing– Reservation at cell level instead of at mobile

level Minimize traffic dynamicsReduce management overhead

– Sink-tree based aggregation at upper link– Multicasting at downlink

Ensure fairness: different cells, different agencies, different users

Efficient Handoff Management

Handoff prediction and guard channel reservation– Dual time scale guard capacity control– More efficient than direct reservation– Prediction aggregation, fairness

Increase scalability– Blocked-based reservation

Packet rerouting and sequencing– Queuing at RNC or at base stations?

Load control and resource management at downlink– More effective diversity control to reduce error rate– Multicasting to speed up rerouting

Multicasting Support

Public safety agencies require: talk or share information within a group of users

Exploit the broadcast feature of downlink channels

Multicasting for circuit-based transmission

Multicasting for time-slotted packet-based transmission

Simulator Design

Build channel modelSimulate functions at air interfaceSimulator functions in the backhaulSimulated all the proposed functions,

performance evaluations

Work Completed

Work Completed So far

Data Traffic Analysis Preliminary simulator design

Traffic Analysis in CDMA Network

Internet data traffic exhibits long range dependency compared to voice traffic– Typical data users: heavy tailed ON/OFF users,

average file size 20KB (or 2.5seconds burst time with 64Kbps) –Long Range Dependent (LRD)

– Typical voice users: exponential ON/OFF users, average burst time 70ms.

CDMA network performance needs to be evaluated and protocols need to be enhanced to accommodate data traffic.

LRD Impact in CDMA Networks

LRD Impact on – Multi-Access Interference (MAI)– Signal to Interference and Noise Ratio (SINR)– Outage Probability

Can be used for traffic prediction– Call Admission Control (CAC)– Rate Control

Multi-access Interference

MAI:

– Xj is user’s activity indicator: when user j is transmitting (ON), Xj=1; when user is silent (OFF), Xj=0.

– Pj is power per sampling time.

– with perfect power control,

– Ki(u) is the equivalent number of active users transmitting with rate Ri

j

N

ijjji PuXuI

,1

)()(

)()()(,1

uKPR

RuXPuI ii

N

ijj i

jjii

jjii RPRP //

Statistics of MAI Distribution of MAI

– Instantaneous MAI I(u) is the sum of multiple independent random variables and approximates Gaussian distribution with variance

– Time-scaled MAI IT(t) is defined as

is the number of samples in T which remains as Gaussian

Long range dependency of MAI– Voice users: ON/OFF periods are exponentially

distributed, then I(u) is SRD.– Data users: ON/OFF periods are heavy tailed, then I(t) is

LRD.– MAI has a Weibull bounded tail distribution:

T

T

vS

SvuT

Ti uI

SvI

1)1()(

1)(

ST

Instantaneous SINR

Instantaneous SINR

Distribution– SINR has the distribution with impact combining N0

and Ki

Long range dependency– Voice users

N0 and Ki are both SRD, N0 +Ki -> SRD and SINR -> SRD.

– Data users N0 is SRD and Ki is LRD, N0

+Ki -> LRD and SINR -> LRD

)(/)()(

0 uKPWuN

GuSINR

ii

ii

Time-scaled SINR

Time-scaled SINR: average over a time window

Noise N0T has a Gaussian distribution with

variance Ki

T also follows a Gaussian distribution– Voice users: variance decreases fast with T– Data users: variance decreases slow with T as H>0.5

SINR has a “Gaussian like”distribution which is the reverse of WN0

T/Pi +KiT (Gaussian distribution)

)(/)()(

0 vKPWvN

GvSINR

Tii

TiT

i

Outage Probability

Outage probability – The probability that the average SINR

or time scaled SINR in a packet transmitting time is smaller than a threshold degraded quality

– Also decay slow.

Prediction in CDMA Networks

Active users K prediction– Predict K in the next window

Tm based on historical values

– Fixed Period (FP) vs. Variable Period (VP) prediction

Prediction is useful for – Rate control: in a relatively

small T– Call admission control: in a

relatively large TFP vs VP

Fixed Period Predictionvs. Variable Period Prediction

Fixed Period Prediction: (existing, simple)– Predict the next value based on the average

value in pervious m windows.Only count a finite number of historical valuesHistorical values are added to prediction with the

same weights.

Variable Period Prediction (more accurate)– Predict the next value based on all previously

measured values with proper weightsAll historical values are added to the predictionMulti time-scale predictionHistorical values are properly weighted in the

predictionRecursive algorithm, consumes less memory

Rate Control

Adjust user’s sending rate based on active user K prediction in a relatively smaller window T (2-10sec.)

Suppose the system can support at most Km (equivalent) active users (transmitting at maximum rate Rm), adjust user’s sending rate according to prediction:– If , increase each user’s rate with

– If , decrease each user’s rate with

um rKiK )1(ˆ

jum RiK

iKrK

)(

)1(ˆ

lm rKiK )1(ˆ

jlm R

iK

rKiK

)(

)1(ˆ

Call Admission Control

Admit new users based on prediction of network performance in a relatively large T (e.g., 5min).

CAC for voice users– Based on average performance– The users that the network can admit is at most

is the activity indicator CAC for data users

– Based on number of active users predicted in the next period

– If , then admit, otherwise reject.

)1(ˆ nK

0)1(ˆ KnK

)(][

11)1(

][

11 0

000 P

WN

SINR

G

XEK

XEM

X

User Throughput

Throughput:

Rate Control CAC

Conclusion for Traffic Analysis

Both MAI and SINR are LRD in a CDMA network with heavy tailed ON/OFF data users

Strong auto-correlation in MAI and SINR could be used for prediction in rate control and CAC

Variable period prediction scheme is proposed and proved to be better than the existing fixed period prediction in terms of– More accurate – Consumes less memory– Achieves better performance in rate control and CAC

Basic Simulator Design

Language: ANSI C++ The network topology

– Approximated as a square mesh.

Event Generator (Most important is handoff event)– Call arrival and departure are generated used

Poisson distribution– Handoff events are triggered on the basis of power

measurements.

Event queue and scheduling: tree-based– Need more efficient event scheduler

Simulator (cont’d)

Mobility model– Random Way Point

Power Measurement– Calculated based on mobile location

Channel Model– Fading, shadowing, path loss, interference

Network Model:– Mobile object, cell object– UMCast: major network functions with references

ALL mobile objects ALL Cell objects Stat class

Challenges:– How to run event generator and algorithm in parallel– Trade off scalability and event granularity

Basic Functions in Simulator

Call initiationCall arrivalCall departurePower measurementHandoff predictionGuard capacity managementAdmission controlPerformance statistics

On-going Work

Multicasting support for downlink circuit based transmissions (support of multimedia such as voice and video for group communications)– How to address heterogeneous requirements of users– How to transmit to different terminals?– How to guarantee quality for users with different channel

conditions?– How to guarantee multicast traffic quality?– How to guarantee un-interrupted communications for each

talk group?– How to tradeoff multicast and unicast transmissions?

Admission control for integrated circuit-based continuous media transmission and slotted-packet-based data– How to formulate resource consumption model?– How to interact with rate control and power control?

Future Work

The remaining of the proposal