METHOD FOR ATTACHING A USER TERMINAL TO A BASE
STATION OF A NETWORK
FIELD OF THE INVENTION
The present invention relates to a method for attaching a user terminal to a
base station of a network. The invention also relates to a network
management for carrying out said method.
Such a method may be used in any network system comprising
heterogeneous types of base stations.
BACKG ROU ND OF THE INVENTION
A method for attaching a user terminal to a base station of a network,
also called user association, well known by the man skilled in the art,
comprises the step of attaching the user terminal to the closest base station.
One problem of the well-known prior art is that this may lead to
unbalanced load, especially among heterogeneous types of base stations
(some with low maximum transmit power, and some with high maximum
transmit power), when small cells and macro cells co-exist. Another problem
is that high-speed user terminals attached to small cells need to handoff
frequently which results in extra cost of resources used for handover (for
example such as operation overhead during which no data transmission is
performed but only handover).
Moreover, it results also in low spectrum utilization efficiency. Indeed, when a
user terminal handoffs, there is a time gap due to switching from one base
station to the other. In this time gap (duration), data transmission has to be
held. For example, if this time gap has to be 1 second, however, the
switching/handoff will happen regularly immediately after every 1 second
of data transmission, then the time efficiency is only 50% (given by 1s / 2s).
SUMMARY OF THE INVE NTION
It is an object of the invention to provide a method for attaching a user
terminal to a base station of a network, which permits to resolve the
problems above-mentioned.
To this end, there is provided a method for attaching a user terminal to
a base station of a network, said method comprising:
- defining a global cost function which is a weighted sum of the user
terminal transmission delays, over all the user terminals in the
network, using a user terminal's context weighting factor representing
a user terminal's characteristic ;
- defining a local cost function for each user terminal from said global
cost function, said local cost function taking into account said user
terminal's context weighting factor for each user terminal, and being a
function of the base station to which this user terminal is associated ;
- running a Gibbs sampler with said local cost function for producing
user-base station association probabilities;
- choosing the user-base station association probability which favors
low local cost ; and
- attaching said user terminal to a base station according to the userbase
station association probability chosen.
As we will see in further details, the method permits to enhance the
user association by taking into account the external user context
characteristics of a user terminal within a heterogeneous surrounding cells
environment and by performing only local operation to achieve global
optimality.
In a first non-limiting embodiment, the user transmission delay is the
inverse of user throughput.
In a second non-limiting embodiment, said throughput is defined from
the SINR according to the Shannon capacity formula which is equal to:
ru = K\oge(1+SINR ) , where K is a constant.
In a third non-limiting embodiment, the user-base station association
probability to associate a user to a base station b is equal to
exp (-C;(b )/r)
, where:
fe exp (- (b)/ )
b is the set of the neighboring base stations for said user ;
C b ) is the local cost function considered at said user terminal
when said user is associated with said base station b ;
T is a parameter which is either a constant or decreases in time.
In a fourth non-limiting embodiment, T is equal to T0/ln(1 +t), where t is
the time and T0 is a constant.
In a fifth non-limiting embodiment, the running of the Gibbs sampler
starts with an arbitrary initial state with said user terminal attached to any one
of the base stations of the network from which a signal may be received.
In a sixth non-limiting embodiment, the user terminal's characteristic is
grade of service and/or the user terminal's velocity.
In a seventh non-limiting embodiment, the grade of service is a data
rate, or the jitter.
In a eighth non-limiting embodiment, when the user terminal's context
characteristic is the user terminal's velocity, a weighting factor is defined,
said weighting factor taking into account the user terminal's velocity and the
type of surrounding cells covering the geographic area corresponding to the
network.
In a ninth non-limiting embodiment, the user terminal's context
weighting factor is defined as being function of:
- a handoff frequency of a user terminal, said handoff frequency
depending on the user terminal's velocity, a cell density over the
geographic area corresponding to the network and a cell radius ;
a time taken by each handoff of the user terminal ; and
the user's required grade of service.
In a tenth non-limiting embodiment, the local cost function is equal to:
Where:
- WU(X) is the weighting factor associated to said user terminal;
- W v(X) is the weighting factor associated to another user terminal;
- Nu is the thermal noise at the user terminal ;
- I(b,u) is the path loss of the transmission from a base station to the
user terminal;
- Pu is the transmission power for the user terminal;
- y(v,u).P v.l(bv,u) is the interference to the user terminal from the
transmission destined to the other user terminal;
- y(u,v) is the orthogonality factor between the user terminal and the
other user terminal;
- I(b,v) is the path loss of the transmission from the base station to the
other user terminal; and
- Pv.l(bv,v) is the power of the received signal at the other user terminal
from the base station which said other user terminal is attached to.
Therefore, the method applies in downlink.
In an eleventh non-limiting embodiment, the local cost function (C ) is
equal to:
Where:
- WU(X) is the weighting factor associated to said user terminal ;
- WV(X) is the weighting factor associated to another user terminal;
- Nu is the thermal noise at the base station of said user terminal;
- I(u, b) is the path loss of the transmission from the user terminal to the
base station;
- Pu is the transmission power of the user terminal;
- Y(v,u).Pyl(v,b) is the interference due to the transmission of the other
user terminal applying on the signal transmitted by said user terminal;
- y(u,v) is the orthogonality factor between the user terminal and the
other user terminal;
- I(u,bv) is the path loss of the transmission from the user terminal to the
base station which the other user terminal is attached to; and
- Pv.l(v,bv) is the power of the signal received at the base station which
is transmitted by the other user terminal.
Therefore, the method applies also in uplink.
In addition, there is provided a network management element for a
network apparatus for attaching a user terminal to a base station of a
network, said network comprising a plurality of base stations, said network
element being adapted to :
- defining a global cost function which is a weighted sum of the user
transmission delays, over all the users in the network ;
- using a user terminal's context weighting factor representing the user
terminal's characteristic ;
- defining a local cost function for each user terminal from said global
cost function, said local cost function taking into account said user
terminal's context weighting factor for each user, and being a function
of the base station to which this user is associated ;
- running a Gibbs sampler with said local cost functions for producing
user-base station association probabilities ;
- choosing the user-base station association probabilities which favors
low local cost ; and
- attaching said user terminal to a base station according to the userbase
station association probability chosen.
In a first non-limiting embodiment, said network apparatus is a base
station.
In a second non-limiting embodiment, said network apparatus is user
terminal network apparatus.
In a third non-limiting embodiment, the steps are performed in a
distributed way at each user terminal or in a centralized way in the base
station.
In addition, there is provided a computer program product, comprising
a set of instructions, which when loaded into said computer, causes the
computer to carry out the method for attaching a user terminal to a base
station of a network, according to any one of the previous characteristics.
BRIEF DESCRIPTION OF THE FIGURES
Some embodiments of methods and/or apparatus in accordance with
embodiments of the present invention are now described, by way of example
only, and with reference to the accompanying drawings, in which:
- Fig. 1 illustrates a schematic multi-layer network system with small and
macro cells, where the method for attaching a user terminal to a base
station according to the invention is to be used;
- Fig.2 illustrates a schematic organization chart of the method for
attaching a user terminal to a base station according to a non-limiting
embodiment of the invention;
- Fig.3 illustrates a schematic organization chart of a definition step of
weighting factor of the method for attaching a user terminal to a base
station of Fig. 2;
- Fig.4 illustrates schematically a network management element which
is adapted to carry out the method for attaching a user terminal to a
base station of Fig. 2.
DESCRIPTION OF EMBODIMENTS OF THE INVENTION
In the following description, well-known functions or constructions by
the man skilled in the art are not described in detail since they would obscure
the invention in unnecessary detail.
The present invention relates to a method for attaching a user terminal
to a base station of a network. More particularly, said method permits to
perform user association.
It is to be reminded that user association stands for the operation of
associating a user at time and location to a base station, which will be in
charge of serving it and exchanging information/data with it.
In the following description, the terms user terminal or user will be
used indifferently.
As will be described hereinafter, the method permits to enhance user
association by combining the Gibbs sampler with the external user context
(i.e. user terminal's characteristics) including the velocity (e.g. user velocity =
low, moderate, or high speed) and service demand (e.g. users may have
different data rate requirement) of the user terminals. In addition, the method
takes into account the heterogeneous nature of the surrounding cells (e.g.
the type of base station is a macro cell or a small cell). Hence, through the
establishment of weighing factors, user demand, user terminal velocity and
the characteristics of surrounding cells are taken into account. Besides, by
defining the attachment method in accordance with the user terminal context
as well as the multitude and types of the surrounding base stations, the
method may be performed in a distributed way.
Here, "distributed" means that one can and will run the Gibbs sampler with
the defined local cost function for himself / herself. However, everybody does
so, i.e., the Gibbs sampling is run for every user terminal. When everybody
in the network does so following the defined local cost function, the global
cost function will be optimized and driven to its minimum provided that a
parameter called the temperature is decreased in an appropriate way. In
other words, user terminals are working together in a joint activity with local
updates (i.e., state transition). Note that this joint activity does not require a
centralized control or coordinator. One can find that the collaborative result
of the above local updates (in a distributed way) will result in an optimization
of the global cost function (the lower, the better). Hence, the joint local
updates of user association favor low global cost.
Thus the performance of the whole network is optimized. The
distributed method only requires local operation and limited information
exchange (among neighboring base stations) for the achievement of global
optimality. Each user makes individual decision in choosing its serving base
station. Besides, there is no requirement on the order of decision making, i.e.
the user association adjustment can be conducted in a distributed and
asynchronous manner without a centralized coordinator. This matches
today's requirement of self-optimized networks.
A network NTW, as illustrated in Fig. 1 is composed of:
- macro cells MC and small cells SC, one base station being
associated to each cell ; and
- user terminal MT. In a non-limiting embodiment, a user terminal is
mobile terminal.
The method M for attaching a user terminal u to a base station b of a
network NTW, said network comprising a plurality of base stations, said
method comprising as illustrated in Fig. 2:
- defining a global cost function Cw which is a weighted sum of the user
terminal transmission delay, over all the user terminals in the network
NTW, using a user terminal's context weighting factor WU(X)
representing a user terminal's characteristic X (step DEF_Cw(r u,
WU(X)) illustrated in Fig. 2) ;
- defining a local cost function Cu for each user terminal u from said
global cost function Cw, said local cost function Cu taking into account
said user terminal's context weighting factor Wu for each user u, and
being a function of the base station to which this user is associated
(step DEF_Cu(Cw, c )) illustrated in Fig. 2) ;
- running a Gibbs sampler with said local cost functions Cu for
producing user-base station association probabilities (step GIBBS(Cu,
V, u{b) , SO) illustrated in Fig. 2);
- choosing the user-base station association probabilities which favors
low local cost (step SELEC(V, opt(Cw) illustrated in Fig. 2) ; and
- attaching said user terminal u to a base station b according to the
user-base station association probability chosen p ) (step ATTCH
(u, b, p ) illustrated in Fig. 2).
The attachment method is described in details below.
Reference to Fig. 2 and 3 will be made.
In a first step 1), as illustrated in Fig. 2, one defines a global cost
function Cw which is a weighted sum of the user terminal transmission
delays, over all the user terminals in the network NTW, using a user
terminal's context weighting factor WU(X) representing a user terminal's
characteristic X.
The user terminal's characteristic X represents the current external
context data of the user terminal, such as in a non-limiting example, grade of
service, user terminal's velocity.
Hence, one defines a global cost function that incorporates the
weighting factors which represent the user terminal's characteristics (e.g.,
grade of service and user terminal's velocity as described later in the
description) and reflect the preference in low handoff frequency and higher
radio spectrum (or bandwidth) utilization efficiency.
For a balance between throughput enhancement and also bandwidth
sharing fairness among users, one aims at minimizing the network's
aggregate transmission delay instead of the sum of throughputs, i.e. å uDu ,
where Du is the transmission delay experienced by user terminal u in the
network and Du= —.
u
It is to be reminded that the objective of the above minimization is to
minimize the overall potential delay of the data transfers in progress in the
network NTW. This (minimizing the overall potential delay) provides an
intermediate solution between max-min and proportional fairness, penalizing
long routes less severely than the latter. It is to be noted that in the case of
max-min fairness, the performances such as the throughput, SINR, etc. will
be the same for a user terminal at the center of a cell than for a user terminal
at the border of the cell, but in the case of proportional fairness it won't be
the case, neither in the case of the above minimization.
Therefore, the cost function chosen minimizes the sum of the inverse
of throughput, or equivalent^ the total delay to send an information unit to all
users, which penalizes very low throughputs.
Hence, the following global cost function, which is the network's
weighted aggregate transmission delay, is defined as:
Cw = [1]
u
Where u is the set of user terminals to be served by the set of base
stations in the network NTW.
Under the additive white Gaussian noise (AWGN) model, the
throughput (also called data rate) in bits/s/Hz at the user terminal u is defined
by r = log (l+SINR ) [2], where K is a constant. K depends on the width
of the frequency band allocated to the user terminal u.
Therefore, the global cost function C is defined as
C W= W(X) _ W(X)
uU uUSINRu
This global cost function emulates the potential delay fairness.
It is to be noted that Cw will be the global energy of the Gibbs sampler.
Therefore, in a non-limiting embodiment, the user transmission delay
is the inverse of user throughput ru and said throughput ru is defined from
the SINR (Signal to Interference plus Noise Ratio) according to the Shannon
capacity formula which is equal to:
r = K\og (1+ SINR ) , where K is a bandwidth constant. [4]
As for the optimization, the constant K does not have any impact, one
may ignore this constant. This is done in the following description.
Hence, with this formula, one can have an indication of the throughput
obtained at the user terminal u.
It is to be noted that for each user terminal, one assumes that there is
a pair of orthogonal channels for uplink (from the user terminal to a base
station) and downlink (from a base station to the user terminal)
communications respectively. Since there is no interference between the
uplink and the downlink, for simplicity, one considers initially only the
downlink.
1. Downlink
Therefore, for a user terminal u and a base station b, the SINR at u is
expressible as:
P - b u)
v l ,v¹u
where:
- u is the set of user terminals u which are served by a set of base
stations b of the cellular radio network NTW ;
- bu, is the serving base station of user terminal u ;
- Pu is the transmission power for the user terminal u ;
- I(bu,u) is the path loss of the transmission from the base station bu to
the user terminal u ;
- Nu is the thermal noise at the user terminal u (also called receiver
noise) ;
- Y(v,u).Pyl(b v,u) is the interference to the user terminal u from the
transmission destined to another user terminal v.
By substituting the above SINR expression [5] into [4], one obtains
[6]:
N + å v,u)P-l(bv,u)
º v
Where:
y(v,u) is the orthogonality factor on the transmission signal destined to
the other terminal v ;
Pv.l(bv,u) is the power of interference at the user terminal u due to the
signal transmitted to the user terminal v. In other words, it is the signal
transmitted to/for the user terminal v by its base station bv. it is to be noted
that "its station" means the base station to which said user terminal v is
attached to.
It is to be noted that the throughput and the SINR can be calculated
with other methods than Shannon.
In non-limiting embodiments, one or a plurality of characteristics X are
taken into account.
• Characteristic X is a grade of service.
In a first non-limiting embodiment, the characteristic X is a grade of
service Q. In variants of this non-limiting embodiment, the grade of service Q
is a data rate or the jitter.
As user terminals may have different service demands, one takes into
account this user terminal's characteristic.
When the grade of service is a data rate, in a non-limiting example a
low or high data rate, one defines two corresponding weighting factors Wu
for Q equal to low rate, and Wu for Q equal to high rate such that Wu >Wu
where 0 0.
The weighting factor WU(S) represents the effective time for
throughput, which is the utility time to transmit data when there is no handoff.
In a second step 2), one defines a local cost function Cu for each
user terminal u from said global cost function Cw, said local cost function Cu
taking into account said user terminal's context weighting factor Wu for each
user terminal u, and being a function of the base station to which this user
terminal is associated.
From [6], one defines the following local cost function C (b) for each
user terminal u such that:
where:
- WU(X) is the weighting factor associated to said user terminal u;
- Wv(X) is the weighting factor associated to another user terminal v ;
- Nu is the thermal noise at the user terminal u ;
- I(b,u) is the path loss of the transmission from a base station b to the
user terminal u;
- Pu is the transmission power for the user terminal u ;
- y(v,u).P v.l(bv,u) is the interference to the user terminal u from the
transmission destined to the other user terminal v ;
- y(u,v) is the orthogonality factor between the user terminal u and the
other user terminal v ;
- l(b,v) is the path loss of the transmission from the base station b to the
other user terminal v ; and
- Pv.l(bv,v) is the power of the received signal at the other user terminal
v from the base station bv which said other user terminal v is attached
to.
It is to be noted that the user terminal v could be any user terminal
including u.
[7] can be expressed as
A (b)
C u = " + u u which is a function of b. [8]
Where:
that is to say ~
SINK
and
It is to be noted that the term u(b)P ' can be seen as the "selfish" part
of the energy or cost which is small if the SINR of the user terminal u is large.
In other words, it tells how good the signal received by the user terminal u is
compared to the other user terminal v.
The term Bu{b)P can be seen as the "altruistic" part of the energy or
cost, which is small if the power of interference incurred by all the other
terminal users (i.e. v ¹ u) because of P is small compared to the power
received from their own base stations bv. In other words, it tells how much
damages the user terminal u can do to other user terminals v.
It is to be noted that the local cost function Cu is a function of b, said feature
being used as described further in the description.
• Characteristic X is a grade of service.
When the characteristic X is a grade of service Q, from [7], one defines a
local cost function such that:
which corresponds to the global cost function
er K - • Characteristic X is the user terminal's velocity S
When the characteristic X is the user terminal's velocity S, one defines a
local cost function such that:
which corresponds to the global cost function:
er K - Of course, in a non-limiting embodiment, the characteristics Q and S may be
combined together, such that:
which corresponds to the global cost function:
In a third step 3), one runs a Gibbs sampler with said local cost function Cu
for producing user-base station association probabilities.
It is to be noted that the Gibbs sampler (or Gibbs sampling) is a
sampling method to generate a sequence of samples from the probability
distribution of one or more random variables. It is a stochastic method (i.e. a
method that makes use of random numbers, here, the random variable is b,
and the state transition is not in a deterministic approach but in a
probabilistic approach governed by the probability distribution p ) ) .
As the local cost function Cu
w is a function of b, one samples a random
variable b on the set of base stations b (i.e. the set of neighboring base
stations) according to a probability distribution described hereinafter.
As described hereinafter, when running the Gibbs sampler for all the user
terminals, one obtains for each user terminal a base station b to which it can
be attached, said base station being chosen according to the probability
distribution obtained.
In a non-limiting embodiment, the user-base station association
probability p ) to associate a user terminal u to a base station b is equal
exp (- (b) / )
to: —— — . - , where:
fe exp (- (b) / )
- b is the set of the neighboring base stations for said user terminal u;
- C b) is the local cost function considered at said user terminal u when
said user terminal u is associated with said base station b ; and
T is a parameter called the temperature which is either a constant or
decreases in time.
It is to be noted that this probability distribution favors low costs.
In a first non-limiting embodiment, the temperature T is a constant.
In a second non-limiting embodiment, the temperature T decreases in time
and is equal to T0/ln(1 +t), where t is the time and T0 is a constant.
When T is a constant, the network will converge to a stationary distribution
which favors low energy states. In other words, the network will be driven to
a network configuration which has a low energy. This resulting energy (cost)
may not be a global minimum point.
It is to be noted that by conducting annealing, that is to say when T
decreases in time and is equal to T0/ln(1 +t) where t is the time, one
guarantees convergence to an optimal configuration of minimal global cost
Cw . The network will be driven to a state of minimal energy (i.e. global cost
function). In other words, the network will be driven to a global minimum
point.
• Gibbs sampler
As will be described below, Gibbs sampler means that the state
transition refers to user-base station association, where the probability
distribution follows the Gibbs's defined probability p ) .
In a non-limiting embodiment, the Gibbs sampler operates on the
graph G defined below:
-the set of nodes of the graph is the set of user terminals u;
-each node has a state which is its user-base station association;
-the set of neighbors of node u in this graph is the set of all users v¹ u such
that the power of the signal received from the base station bv at user terminal
u is above a specific threshold Q.
Through information exchange between neighboring base stations,
the local cost function C (b) to determine p ) are evaluated as follows.
A state transition is based on the local cost function C b) . So the base
station bu needs to gather some information to determine the coefficients of
C b) . To do so, each user VGU reports the following data to its base station
bv.
- (a) Its SINRV;
- (b) The power of its received signal, i.e. Pv -l(b v ,v) ; and
- (c) The power of the signal received from the other base stations,
i.e. r(u,v)Pu l bu ,v) .
From the collected information, each base station bu is able to compute the
local cost function Cu {b) .
When running the Gibbs sampler, one starts with an arbitrary initial
state SO with a user terminal u randomly selected. Then, one runs the Gibbs
sampler to decide to which base station b the user terminal u will be
attached. The Gibbs sampler is performed for each user terminal randomly
chosen. This results in at least one user-association probability p ) , such
that p ) = 1, or this results in a plurality of user-association probabilities
p (b) , such that 0< p (b) < 1 , for each b.
Therefore, the result is a probability distribution (set of p ) )
describing the probabilities that the user-base station association should be
selected.
As a result, in each state transition, the Gibbs sampler samples a
random variable b having more likely a small local cost function.
It is to be noted that as the global cost function is finally expressible as
a sum of the inverse of SINR, the Gibbs sampler can be used with a
guarantee of optimal performance.
It is to be noted that in a non-limiting example, the Gibbs sampler is
run in the base station b since base station is computationally more powerful
and with more resources.
In another non-limiting embodiment, it is applicable if one may wish to
implement and run it the user terminal if there is no resource limitation
problem or possible constraints.
In a fourth step 4), one chooses the user-base station association
probability p ) which favors low local cost.
This selection is performed according to the probability distribution (set of
p ) ) obtained before.
In a non-limiting example, if a user terminal has two possible choices of base
stations, say b 1 and b2, which correspond to said user-base station
association probability 0.9 and 0.1 , respectively, as a result, said user-base
station association probability 0.9, and therefore said base station b 1 will be
chosen with a higher probability (i.e., 0.9).
Said user-association probability chosen p ) corresponds to a base station
b.
Therefore, in a fifth step 5), one attaches said user terminal u to a base
station b according to the user-base station association probability chosen
These steps 3, 4, and 5 are iterated with all the other user terminals
randomly selected.
It is to be noted that in practice, the network is dynamic, that means that the
attachment method will run all the time, i.e. continuously to tune and adapt
the network. Therefore, the optimization will not stop.
It is to be noted that when the user terminal u is attached to a base station b,
the value of the SINR changes. It means that the local cost function Cu
changes (especially the interference element of the local cost function) each
time the user terminal u is attached to a different base station b, this means
at each iteration when running the Gibbs sampler.
After running the Gibbs sampler, the result is: for each user terminal, a
selected base station b to which a terminal may be attached (based on a
local cost function); the whole set of user-base station association in the
network will result in an optimal global cost.
Hence, this attachment method permits to optimize automatically the
global cost function Cw by running Gibbs sampler with the local cost function
C (as the global cost function Cw is the sum of all the local cost function
C"), taken into account user terminal's context characteristic such as a
grade of service Q and the user terminal's velocity S within an
heterogeneous network NTW (which comprises macro and small cells).
It may be performed in a distributed way and optimizes the performance of
the whole network. Therefore, it is a fully distributed approach which optimize
the global cost function.
Hence, the user association procedure is defined in accordance with the
mobile user's characteristics (grade of service + mobile velocity) as well as
the multitude and types of the surrounding base stations.
The attachment method is carried out by a network apparatus NE for
attaching a user terminal u to a base station b of a network NTW, said
network NTW comprising a plurality of base stations b, as illustrated on
Fig.4.
Said network apparatus NE is adapted to:
- define a global cost function Cw which is a weighted sum of the user
terminal transmission delay, over all the user terminals in the network
NTW, using a user terminal's context weighting factor WU(X)
representing user terminal's characteristic X ;
- define a local cost function C for each user terminal u from said
global cost function Cw, said local cost function Cu taking into account
said user terminal's context weighting factor Wu for each user terminal
u, and being a function of the base station to which this user terminal
is associated ;
- run a Gibbs sampler with said local cost function Cu for producing
user-base station association probabilities p ) ;
- choose the user-base station association probability p ) which
favors low local cost ; and
- attaching said user terminal u to a base station b according to the
user-base station association probability p ) chosen.
In a first non-limiting embodiment, said network apparatus NE is a
base station b. In this case, the attachment method is performed in a
centralized way.
In a second non-limiting embodiment, said network apparatus NE is a
user terminal u. In this case, the attachment method is performed in a
distributed way.
It is to be understood that the present invention is non-limiting to the
aforementioned embodiments and variations and modifications may be
made without departing from the scope of the invention. In the respect, the
following remarks are made.
It is to be understood that the present invention is non-limiting to the
aforementioned application.
Hence, the present invention has been described for the downlink, but
it can be applied to the uplink as described hereinafter.
2. Uplink
What has been above-described in the downlink part description (the
fifth steps, the Gibbs sampler, the network apparatus . . .) applies to the
uplink in the same manner.
In the following, for concision purpose, only the differences between the
downlink and the uplink are described and for sake of clarity few paragraphs
described in the downlink part description are reminded.
As mentioned in the downlink part description, the user transmission
delay is the inverse of user throughput. Throughput is defined from the SINR
according to:
ru = K \og ( 1+ SINR ) , where K is a constant.
The user-base station association probability to associate a user to a
exp (-C (b )/ )
base station is equal to —— — , where:
fe exp (- (b ) / )
- b is the set of the neighboring base stations for said user ;
- C b ) is the local cost function considered at said user terminal
when said user is associated with said base station ;
- T is a parameter which is either a constant or decreases in time.
The global cost function, which is the network's weighted aggregate
transmission delay, is defined as:
C W= U X _ WU(X)
e K - 1
where, for a user terminal u and a base station b, the SINR at u is
expressible as, for uplink:
where:
- u is the set of user terminals u which are served by a set of base
stations b of the cellular radio network NTW ;
- bu, is the serving base station of user terminal u ;
- Pu is the transmission power of the user terminal u ;
- I(u, bu) is the path loss of the transmission from the user terminal u to
the base station bu ;
- Nu is the thermal noise at the receiver in the base station (bu) for user
terminal u (also called as receiver noise) ;
- Y(v,u).Pyl(v,b u) is the interference due to the transmission of user
terminal v applying on the signal transmitted by user terminal u (which
is destined to bu) .
By substituting the above SINR expression into Cw, one obtains, for
uplink:
Where:
Y(v,u) is the orthogonality factor on the transmission signal destined to
the other terminal v ;
Pv.l(v,bu) is the power of interference received at the base station
(which is the receiver) of user terminal u due to the transmission signal of
user terminal v.
The user terminal's characteristic is grade of service and/or the user
terminal's velocity.
For uplink, the local cost function is equal to:
Where:
- WU(X) is the weighting factor associated to said user terminal u ;
- Wv(X) is the weighting factor associated to another user terminal v;
- Nu is the thermal noise at the base station b of said user terminal u
(also called as receiver noise) ;
- I(u, b) is the path loss of the transmission from the user terminal u to
the base station b;
- Pu is the transmission power of the user terminal u;
- y(v,u).Pv-l(v,b) is the interference due to the transmission of the other
user terminal v applying on the signal transmitted by said user
terminal u;
- y(u,v) is the orthogonality factor between the user terminal u and the
other user terminal v ;
- I(u,bv) is the path loss of the transmission from the user terminal u to
the base station bv which the other user terminal v is attached to; and
- Pv.l(v,bv) is the power of the signal received at the base station b
which is transmitted by the other user terminal v.
It is to be noted that the user terminal v could be any user terminal
including u.
As mentioned in the downlink part description, in non-limiting
embodiments, one or a plurality of characteristics X are taken into account.
• Characteristic X is a grade of service.
In a first non-limiting embodiment, the characteristic X is a grade of
service Q. In variants of this non-limiting embodiment, the grade of service Q
is a data rate or the jitter.
What has been described in the downlink part description in the
corresponding paragraph, applies here.
• Characteristic X is the user terminal's velocity S
In a second non-limiting embodiment, the characteristic X is a user
terminal's velocity S.
What has been described in the downlink part description in the
corresponding paragraph, is applies here.
As mentioned in the downlink part description, one defines a local cost
function Cu for each user terminal u from said global cost function Cw, said
local cost function C taking into account said user terminal's context
weighting factor Wu for each user terminal u, and being a function of the
base station to which this user terminal is associated.
One defines the following local cost function C (b) for each user
terminal u such that, for uplink:
which is a function of b, where the first term can be seen as the
"selfish" part of the energy or cost which is small if the SINR of the signal
transmitted by user terminal u received at said base station b is large while
the second term can be seen as the "altruistic" part of the energy or cost,
which is small if the power of interference incurred by all the other
transmissions (i.e. v ¹ u) because of u is small compared to the power
received at their own base stations.
• Characteristic X is a grade of service.
When the characteristic X is a grade of service Q, one defines a local cost
function such that, for uplink:
which corresponds to the global cost function:
• Characteristic X is the user terminal's velocity S
When the characteristic X is the user terminal's velocity S, one defines a
local cost function Cu such that, for uplink:
which corresponds to the global cost function:
er - Of course, in a non-limiting embodiment, the characteristics Q and S may be
combined together, such that:
which corresponds to the global cost function:
One runs a Gibbs sampler with said local cost function
ucing user-base station association probabilities.
Through information exchange between neighboring base stations,
the local cost function C (b) to determine p ) are evaluated as follows. A
state transition is based on the local cost function Cu
w b) . So, in uplink, the
base stations need to gather some information to determine the coefficients
of Cu
w b) . To do so, each user VGU reports its PV value to its base station bv
which (the information) would be exchanged among neighboring base
stations. By which, the base stations can estimate the coefficients of Cu
w b) ,
i.e., the path loss values in order to determine p ) .
Hence, the present invention has been described for heterogeneous
base stations such as of the type of small cells and macro cells, but it can
applied to the case of homogeneous base stations (for example a network of
small cells).
It is to be understood that the present invention is non-limiting to the
aforementioned embodiments.
Hence, other network apparatus NE than the base station b or the
user terminal u may be used. For examples, in a centralized implementation
of the attachment method, the network apparatus NE may be :
- an administrative owner AO which is a third party independent
from the operators ;
- a Network Management System, NMS ; or
- an Operational Support System, OSS.
It is to be understood that the methods and the elements according to
the invention are non-limiting to any implementation.
Thus, in a non-limiting embodiment, one may use binary encoding (and then
use log2 in the Shannon formula) such that the global cost function Cw is
define
There are numerous ways of implementing functions of the attachment
method by means of items of hardware or software, or both, provided that a
single item of hardware or software can carry out several functions. It does
not exclude that an assembly of items of hardware or software or both carry
out a function. For example, the step of building a path may be combined
with the step of updating an associated set of words, thus forming a single
function without modifying the building method M in accordance with the
invention.
Said hardware or software items can be implemented in several manners,
such as by means of wired electronic circuits or by means of a computer
program product that is suitable programmed respectively. A computer
program product PG can be contained in a computer or in a network
apparatus NE, said NE comprising a unit control UC, said unit control being
hardware or software items as above stated.
The computer program product PG comprises a first set of instructions.
Thus, said set of instructions contained, for example, in a computer
programming memory or in a network apparatus NE, may cause the
computer or the network apparatus NE to carry out the different steps of the
attachment method.
The set of instructions may be loaded into the programming memory by
reading a data carrier such as, for example, a disk. A service provider can
also make the set of instructions available via a communication network such
as, for example, the Internet.
Hence, some embodiments of the invention may comprise one or a
plurality of the following advantages:
- The attachment method (also called user association procedure) optimizes
the association/attachment of the users among the available base stations
within a given geographical area in a distributed way while taking into
account the user context (such as the target service and the mobile velocity)
as well as the characteristics of the surrounding cells so as to minimize the
overall transmission delay and overhead due to handoffs.
The optimization is conducted with the user terminal's context characteristic
such as the grade of service (e.g., target/expected data rate).
The optimization is conducted with respect to the user terminal's velocity and
the type of surrounding cells covering the geographic area which may bring
handoffs. The resulting optimization is to minimize the overall transmission
delay and overhead due to the handoffs.
Therefore, it enables the mobile terminal to exchange data with the base
station in the context of more and more complex and heterogeneous
networks.
Therefore, it offers better quality of service for each user terminal, especially
in the context of heterogeneous networks and taking into account the user
context.
Therefore, it optimizes the user association procedure with a multitude of
base stations (e.g. macro and small cells). Therefore, the invention tackles
the problem of new (emerging) networks with heterogeneous type of base
stations, consisting of macro and small cell base station.
- it permits to determine the best base station to which a user terminal is
attached in distributed way and such that it optimizes the overall functioning
of the network (including multi-layer networks) ;
- it avoids (if wanted) having a centralized coordinator to compute the global
function cost. It is based on local measurements and limited information
exchange and can adaptively drive the system to a state of global optimal
configuration. Therefore it avoids huge exchange of data information in the
network and a centralized user-association method with high complexity ;
- it does not only provide system throughput enhancement but also support
fair bandwidth sharing among the users in the network. Therefore, it permits
to have a user-association procedure which is relevant for large scale
wireless networks ;
- it permits to find the optimal (minimal) point of the (network/global) cost
function which may have multiple local optimal points where other methods
well-known by the man skilled in the art (e.g. hill climbing approaches) may
find a less optimal point.
- it applies to downlink and to uplink.
Any reference sign in the following claims should not be construed as
limiting the claim. It will be obvious that the verb "to comprise" and its
conjugations do not exclude the presence of any other steps or elements
beside those defined in any claim. The word "a" or "an" preceding an
element or step does not exclude the presence of a plurality of such
elements or steps.
The description and drawings merely illustrate the principles of the
invention. It will thus be appreciated that those skilled in the art will be able to
devise various arrangements that, although not explicitly described or shown
herein, embody the principles of the invention and are included within its
spirit and scope. Furthermore, all examples recited herein are principally
intended expressly to be only for pedagogical purposes to aid the reader in
understanding the principles of the invention and the concepts contributed by
the inventor to furthering the art, and are to be construed as being without
limitation to such specifically recited examples and conditions. Moreover, all
statements herein reciting principles, aspects, and embodiments of the
invention, as well as specific examples thereof, are intended to encompass
equivalents thereof.

CLAIMS
Method (M) for attaching a user terminal (u) to a base station (b) of a
network (NTW), said network comprising a plurality of base stations (b),
said method comprising :
- defining a global cost function (Cw) which is a weighted sum of the
user terminal transmission delays, over all the user terminals in the
network (NTW), using a user terminal's context weighting factor
(WU(X)) representing a user terminal's characteristic (X) ;
- defining a local cost function (C ) for each user terminal (u) from said
global cost function (Cw) , said local cost function (C ) taking into
account said user terminal's context weighting factor (Wu) for each
user terminal (u), and being a function of the base station (b) to which
this user terminal is associated ;
- running a Gibbs sampler with said local cost function (C ) for
producing user-base station association probabilities { p ) )
- choosing the user-base station association probability { p ) ) which
favors low local cost ; and
- attaching said user terminal (u) to a base station (b) according to the
user-base station association probability { p ) ) chosen.
A method (M) according to the previous claim 1, wherein the user
transmission delay is the inverse of user throughput ( r ) .
A method (M) according to any one of the previous claims, wherein said
throughput ( r ) is defined from the SINR according to the Shannon
capacity formula which is equal to:
r = K \og e (1+SINR ) , where K is a constant.
A method (M) according to any one of the previous claims, wherein the
user-base station association probability { p ) ) to associate a user
exp(-C (b) / )
terminal (u) to a base station (b) is equal to —— — , — r , where:
å e P - )
- b is the set of the neighboring base stations for said user
terminal (u);
- C b) is the local cost function considered at said user terminal
(u) when said user terminal (u) is associated with said base
station (b) ; and
- T is a parameter which is either a constant or decreases in time.
5- A method (M) according to the previous claim 4, wherein the parameter T
decreases in time and is equal to T0/ln(1 +t), where t is the time and T0 is
a constant.
6- A method (M) according to any one of the previous claims, wherein the
running of the Gibbs sampler starts with an arbitrary initial state with said
user terminal (u) attached to any one of the base stations (b) of the
network (NTW) from which a signal may be received.
7- A method (M) according to any one of the previous claims, wherein the
user terminal's characteristic (X) is grade of service (Q) and/or the user
terminal's velocity (S).
8- A method (M) according to the preceding claim, wherein the grade of
service (Q) is a data rate, or the jitter.
9- A method (M) according to any one of the previous claims, wherein when
the user terminal's characteristic (X) is the user terminal's velocity (S), a
weighting factor (WU(X)) is defined, said weighting factor (WU(X)) taking
into account the user terminal's velocity (S) and the type of surrounding
cells (CI) covering the geographic area (A) corresponding to the network
(NTW).
10-A method (M) according to the preceding claim, wherein the user
terminal's context weighting factor (WU(X)) is defined as being function of
- a handoff frequency ( o ) of a user terminal (u), said handoff
frequency depending on the user terminal's velocity (S), a cell density
(cld) over the geographic area corresponding to the network (NTW)
and a cell radius (clr) ;
a time taken (T Ho) by each handoff of the user terminal (u).
11-A method (M) according to any one of the previous claims, wherein the
local cost function (C ) is equal to :
where:
- WU(X) is the weighting factor associated to said user terminal (u);
- Wv(X) is the weighting factor associated to another user terminal (v);
- Nu is the thermal noise at the user terminal (u) ;
- I(b,u) is the path loss of the transmission from a base station (b) to the
user terminal (u);
- Pu is the transmission power for the user terminal (u) ;
- Y(v,u).Pyl(b v,u) is the interference to the user terminal (u) from the
transmission destined to the other user terminal (v) ;
- y(u,v) is the orthogonality factor between the user terminal (u) and the
other user terminal (v) ;
- I(b,v) is the path loss of the transmission from the base station b to the
other user terminal (v) ; and
- Pv.l(bv,v) is the power of the received signal at the other user terminal
(v) from the base station (bv) which said other user terminal (v) is
attached to.
12-A method (M) according to any one of the previous claims 1 to 10,
wherein the local cost function (C ) is equal to:
Where:
- WU(X) is the weighting factor associated to said user terminal (u) ;
- Wv(X) is the weighting factor associated to another user terminal (v);
- Nu is the thermal noise at the base station (b) of said user terminal (u);
- I(u, b) is the path loss of the transmission from the user terminal (u) to
the base station (b);
- Pu is the transmission power of the user terminal (u);
- Y(v,u).Pyl(v,b) is the interference due to the transmission of the other
user terminal (v) applying on the signal transmitted by said user
terminal (u);
- y(u,v) is the orthogonality factor between the user terminal (u) and the
other user terminal (v);
- l(u,bv) is the path loss of the transmission from the user terminal (u) to
the base station (bv) which the other user terminal is attached to; and
- Pv.l(v,bv) is the power of the signal received at the base station (bv)
which is transmitted by the other user terminal (v).
13-A method (M) according to any one of the previous claims, wherein the
steps are performed in a distributed way at each user terminal (u) or in a
centralized way in the base station (b).
14-A network apparatus (NE) for attaching a user terminal (u) to a base
station (b) of a network (NTW), said network (NTW) comprising a plurality
of base stations (b), said network element (NME) being adapted to :
- defining a global cost function (Cw) which is a weighted sum of the
user transmission delays, over all the users in the network (NTW),
using a user terminal's context weighting factor (Wu) representing a
user terminal's characteristic (X) ;
- defining a local cost function (C ) for each user terminal (u) from said
global cost function (Cw) , said local cost function (C ) taking into
account said user terminal's context weighting factor (Wu) for each
user (u), and being a function of the base station to which this user is
associated ;
- running a Gibbs sampler with said local cost functions (C ) for
producing user-base station association probabilities ;
- choosing the user-base station association probabilities which favors
low local cost ; and
- attaching said user terminal (u) to a base station (b) according to the
user-base station association probability { p ) ) chosen.
15-A network apparatus (NE) according to the preceding claim, wherein said
network apparatus is a base station (b).
16-A network apparatus (NE) according to the preceding claim, wherein said
network apparatus is user terminal (u).
17-A computer program product (PG) for a computer, comprising a set of
instructions, which when loaded into said computer, causes the computer
to carry out the method for attaching a user terminal (u) to a base station
(b) of a network (NTW), according to any one of claims 1 to