mrn -3 – mobility - politecnico di milanohome.deib.polimi.it/capone/wn/mrn-en-3-mobility.pdf ·...

Politecnico di MilanoFacoltà di Ingegneria dell’Informazione

MRN -3 – Mobility

Mobile Radio NetworksProf. Antonio Capone

Mobility management

o In mobile radio networks, users can freely move in the area of the system changing the cell to which they are associated

o This obviously poses routing problems (bot for voice calls and data sessions)

o All the procedures that the system adopts for allowing users mobility with seamless connectivity are denoted as “mobility management”

A. Capone: Mobile Radio Networks 2

Mobility management

o Users while roaming the in service area can:n Calln Be called n Have an active call/session

o It is necessary to use some mechanisms for tracking user position in the network and adapt routing accordingly


Mobility management

o In the case of voice service, mobility management procedures differs according to the user state, IDLE (no call in progress) or ACTIVE (call in progress)n ACTIVE: there is an active connection that must be

rerouted at each cell change to guaranty continuous connectivity (Handover)

n IDLE: user must be localized in order to be reached in case of an incoming call (Location Update, Cell Selection, Cell Reselection)


Mobility management: Cell selection

o User terminal autonomously selects the base station based on the level of the signal received by base station s

o Base stations periodically transmits broadcast message on a common control channel, which includes system information and its identity (cell ID)

o User terminal scans frequencies for receiving broadcast messages of base stations in the neighborhood

o Terminal select the base station from which it receives the strongest signal

o Terminal continues to scan frequencies and whenever a base station with a stronger signal is received the new cell is reselected

o In the cell selection and reselection process also other system information are usually considered (like the network identification)



Handover

Mobility management: Handover

o This is the procedure that allows a terminal with an active session to change its base station without loosing connectivity

o Handover is always started by the network, based on measurements (signal strength, quality, etc.) performed both on user and network sides

o In order to guaranty connectivity and avoid information losses, handover procedures need to be fast and efficient

o For the MRN technologies we’ll see how handover is handled with signaling and rerouting procedures




o The selection of handover activation thresholds is a critical element

o If h is too small we may loose the connection because time is not enough

o If h is too big we may have too many handover procedures even when not necessary

When to start a Handover?

Dt t

Handover TH

Receiver THh



o There are different possible approaches1 – strongest signal

o handover is initiated in A

Due to signal fluctuations connection can be switched back-and-forth several times (ping-pong effect)



2 – strongest signal with thresholdo If signal of previous BS is below a threshold (e.g. T2) and the

power of the new BS is stronger; handover is initiated in B



3 – stronger signal with hysteresiso If the power of the new BS is stronger with a difference ³ h;

handover is initiated in C

Mobility management: Handover types


o Hard Handover (GSM-2G)

o Soft Handover (UMTS-3G)

In each time only one connection is active

User is contemporary connected to more than one base station

Handover performance

o When a handover is necessary a channel is released in the previous cell and another one must be available in the new cell

o Let us define the handover failure probability (Phf) as the probability that a handover request is rejected since there are no channel available, and blocking probability (Pb) as the probability that a new call is rejected

o If the system does not make distinction between handover requests and call requestswe have Phf=Pb

o Actually, it is usually more tolerable to block new calls rather than dropping already active ones

o We can try to privilege handovers


Guard channels


o Guard Channelsn Some channels are reserved for handover requestsn Phf becomes lower but system capacity for new calls

is reduced (Pb is increased)n Dimensioning guard channels may be critical and

requires traffic and mobility estimations (how many guard channels do we need?)



Assumptions:• Traffic flows of new calls and

handovers are Poisson processes and are independent

• m available channels

New calls

Handover requests

Completed calls

hl

il

cµ

We can simply use Erlang-B with B(A,m)

c

hiAµll +

=

Without guard channels



Let us define:

c

hh

c

ii

µlr

µlr

=

=Assumptions:• Traffic flows of new calls and handovers are

Poisson processes and are independent• m = c + g available channels• g reserved for handovers

New calls

Handover requests

Completed calls

hl

il

cµ

With guard channels



0 1 2 c c+g

hlih ll +

cµ cµ2 ccµ ( ) cc µ1+ ( ) cgc µ+

c+1

hlih ll +ih ll +

… …

π1 =λh +λiµc

π 0 = ρh + ρi( )π 0

π 2 =12ρh + ρi( )2 π 0

...

π k =1k!

ρh + ρi( )k π 0 1≤ k ≤ c

π k =1k!

ρh + ρi( )c ρhk−cπ 0 c < k ≤ c+ g

#

$%%

&%%



( ) ( )

( )

( ) ( )

( )

( ) ( )å å

å å

åå

å å

=

+

+=

-+

=

+

+=

-

+

=

-

+

=

=

+

+=

-

+++

++

==

+++

+==

+++

=

c

k

gc

ck

ckhc

ih

kih

ghc

ih

gchf

c

k

gc

ck

ckhc

ih

kih

gc

ck

ckhc

ihgc

ckkb

c

k

gc

ck

ckhc

ih

kih

kk

gcP

kk

kP

kk

0 1

0 1

0 1

0

!!

)!(

!!

!

!!

1

rrrrr

rrrp

rrrrr

rrrp

rrrrrp


o The model takes as input the parameters: lh, li, µc

o Unfortunately, these parameters are dependent each other and basically determined by traffic intensity and user mobility

o Parameter tc=1/µc represents the average of the channel holding time Xc

o Let us define the parameter th=1/h as the average time spent by a user in a cell Xh



o Let us define the parameter t=1/µ as the average value of the call duration X

o We have:

o Assuming X and Xh are exponential random values we can observe that :


Xc =min X,Xh{ }

hµµt

+==

11

cc


o The probability that a call requires a handover is given by :

o Let us define n as the average number of handovers per call:


Ph = Pr X > Xh[ ] = dx ηe−ηxµe−µy dyx

+∞

∫0

+∞

∫ =

= −η e−(µ+η )y dx0

+∞

∫ =η

µ +η

ν =ττ h

=ηµ



o We can now compute lh as a function of li:

o Solving by lh:

( ) ( )[ ]hhfibhh PPP lll -+-= 11

( )( )[ ] i

hfh

bhh PP

PP ll---

=111

1/6 of traffic moving out of the cell


o Unfortunately the equation depends on the probabilities Pb and Phf which are the values we want to compute with the Markov chain

o We can adopt an iterative approach to solve the problem:1. We assign initial values to Pb and Phf and we compute lh

2. We solve the chain and compare the results with the two values3. If the difference us greater that a threshold we set the values of

Pb and Phf according to step 2 and than go back to step 1; otherwise STOP.


( )( )[ ] i

hfh

bhh PP

PP ll---

=111


o In most cases however, it is reasonable to assume that Pb and Phf are small

o And therefore:

o Obviously the assumption must be verified solving the chain


( )( )[ ] ( ) iii

h

hi

hfh

bhh P

PPPPP nll

µhlll ==

-@

---

=111

1


o The handover failure probability Phf is not however a parameter directly related to service quality

o We need to calculate the probability that a call is dropped due to handover failure:


( )( ) ( )

hfhfh

h

i hfh

hfhihf

ihhfh

hfhhfhhfd

PPPP

PPPP

PPPP

PPPPPP

n=-

@

=--

=-=

=+-+=

å¥

=

1

111

...1

0

2

1@



o We still need to calculate parameter h as a function of the terminal speed and cell size

o Let us consider first a simple case with square cells:

a

r

l

o # of handovers g = # of horizontal lines + # of vertical lines crossed



o We get:

( )

lV

rV

lr

lr

dl

rl

r

pgh

paa

p

aaap

g

p

p

4

4cossin2

sincos2

2/

0

2/

0

==

=úûù

êëé -

=+= ò

where V is the speed



o Proof: Let us consider an infinitesimal segment of the perimeter dl and the probability that a terminal crosses it in the time internal dt.

o Let fV(v) be the probability density of the speed and a the incidence angle on segment dl

o Terminal will cross the segment only if it is in the infinitesimal area:

o In general we have: S

LVp

h = Where L is the perimeter and S the area

dl

a acosvdt



o The probability is then given by the area divided by the total area:

o Averaging on v and on a, we get:

o Integrating along the perimeter:o And therefore the cell crossing frequency:

Sdldtvp acos

=

12πS

dv v fV v( )cosα dl dt−π /2

π /2

∫0

∞

∫ dα =

=12πS

2 v fV v( ) dl dt dv0

∞

∫ =VπS

dldt

dtSVLpp

=

SLVp

h =



LETTURA CONSIGLIATA:Trends in handover design Pollini, G.P.IEEE Communications Magazine , Volume: 34 Issue: 3 , March 1996 Page(s): 82 -90

Location update


Mobility Management: Location Update

o Location Area: topological entity hierarchically on top of the cell (group of cells)

o An IDLE terminal is localized based on the Location Area (and not based on cell)

o The last Location Area visited is stored into network data bases

A. Capone: Mobile Radio Networks 32LA 1 LA 2

Data Base

Mobility Management: Location Update


o If a IDLE terminal moves from a LA to another, it initiates a Location Update procedure

o The information on the LA currently visited by a terminal is used by the network to route incoming calls

LA 2

Data Base

LA 1

Pagingo When a new call is coming, the data base is searched to know the

LA currently isitedo Then a paging procedure is startedo Every base stations in the LA start transmitting a paging message

with the ID of the terminal on a broadcast control channelo When the terminal replies on the access channel the exact cell is

identified and the call is routed


Data Base

pagingpaging

reply

Paging vs. Location Updateo QUESION:

n What is the most convenient size (number of cells) of the Location Areas?o smallo big

n What does suggest to make them big and what small?


LA 2

Data Base

LA 1

Paging vs. Location Update

o The dimensioning of the location areas is a tradeoff between two opposite needs:n Increasing the location area size (number of

cells) we increase the signaling traffic due to paging

n Decreasing the size we increase the signaling traffic due to location updates

o Thus there is an optimal size that depends on a number of parameters including the terminal mobility and the arrival frequency of calls


Location Area dimensioningo A simple method for the dimensioning is based on the

calculation of the signaling cost per usero Let us consider squared cells of edge l, and squared location

areas with k ´ k cellso Let us denote with

n Cp the signaling cost per user per cell due to an incoming call [bytes/paging/cell]

n Cu the signaling cost due to a location update [bytes/update]

n l incoming call arrival frequency per user [calls/hour/user]

n uk location update frequency per user [update/hour/user]


Location Area dimensioning

o The signaling cost per user can be expressed as:

o Let us denote withn m the average number of users getting out of the

location area in the time unitn r the density of users per unit arean v the speed of the user (V=E[v])


( ) ukpk CuCkukC += ll 2,,



o Similarly to what we did for the calculation of the parameter h for the handover we can calculate:

o And thereforep

r Vklm ××=

4

ku

klV

klmuk 1

2

4)(

===pr


o Normalizing Cu to 1 and defining c(k,l)= C(k,l)/Cu we can give the normalized cost as:

o where


kukkc /),( 12 += lgl

up CC /=g

plVu 4

1 =

Location Area dimensioningo Let us define:

o The optimal value of k, kopt, can be computed looking for the maximum value of k for which D is negative:


)1()1()12(

),,1(),,(),,(

1

111

----

=

=--=D

kkukkk

ukcukcukgl

lll

kopt (λ,u1) =1 se Δ(2,λ,u1)> 0max k |Δ(k,λ,u1) ≤ 0[ ] otherwise

#$%

&%



o Example:

l = 300 mV = 8 km/hγ =Cp /Cu = 0.1λ = 0.6 [calls/h/user]

user][update/h/ 95.3341 ==

plVu

29.0),,8(03.0),,7(46.0),,6(16.1),,5(41.2),,4(36.5),,3(80.16),,2(

1

1

1

1

1

1

1

+=D-=D-=D-=D-=D-=D-=D

uuuuuuu

lllllll

o We get:

7=optk

mrn -3 – mobility - politecnico di milanohome.deib.polimi.it/capone/wn/mrn-en-3-mobility.pdf ·...

Documents