WO2014082686A1

WO2014082686A1 - Method for determining powers and locations of radio transmitters

Info

Publication number: WO2014082686A1
Application number: PCT/EP2012/074174
Authority: WO
Inventors: Jaap Van De Beek; Andreas Polydoros; Ioannis DAGRES; Liljana GAVRILOVSKA; Vladimir ATANASOVSKI; Daniel DENKOVSKI
Original assignee: Huawei Technologies Sweden Ab
Priority date: 2012-11-30
Filing date: 2012-11-30
Publication date: 2014-06-05
Also published as: CN104871615A

Abstract

The present invention relates to a method for determining powers and locations of j = 1,..., K numbers of radio transmitters based on a plurality of received signal strengths measured at i = 1,..., N _s number of sensors, where K and N _s are positive integers and K ≥ 2 and N _s ≥ 1, respectively; the method comprising the step of: using the expectation-maximization (EM) method for determining the powers P _j and locations X _j of the j = 1,..., K numbers of radio transmitters based on the plurality of received signal strengths measured at the i = 1,..., N _s number of sensors. Furthermore, the invention also relates to a computer program, a computer program product, and a central processing unit thereof.

Description

METHOD FOR DETERMINING POWERS AND LOCATIONS OF RADIO

TRANSMITTERS

Technical Field

The present invention relates to a method for determining powers and locations of radio transmitters. Furthermore, the invention also relates to a computer program, a computer program product, and a central processing unit thereof.

Background of the Invention

Passive transmitter localization via sensor networks is an important topic for many wireless applications since localization of unknown radio sources is likely to become a key component in future, next generation radio communication networks. Endowed with the capability to sense and record radio-activity in the surrounding environment, central operators or other relevant radio network management entities can then efficiently plan, decide upon and execute respective actions by their networks.

In general, the environmental awareness that comes with the knowledge of unknown transmitters' locations is likely to become instrumental in future self-organizing networks where, for instance, the power, frequency or other crucial radio parameters are (automatically) adapted based on the environmental knowledge gathered from massive-scale sensor networks and their affiliated measurements ("observations", "observed data"). Next-generation cellular networks for instance can benefit from such self-organizing capabilities through increased cost efficiency (the reduction of the operational expenses, OPEX) and through increased spectrum and energy efficiency.

Benefits can result for operators and regulators alike; the first to improve network operation, the second to detect and locate malicious and non-licensed spectrum users.

Most projections and forecasts about future cellular networks predict a huge increase of the number of radio transmitters. A key enabler to providing higher data rates is the densification of network infrastructure, i.e. more transmitters per unit area. Figure 1 illustrates the problem scenario that is addressed by the present invention. In an area of interest, two or more active radio transmitters operate in a certain frequency band. The location, power, type or any other relevant parameter of these radio transmissions is unknown. The number of active radio transmitters is also unknown. A possibly high number of distributed sensors perform measurements of the Received Signal Strength (RSS) caused by these unknown transmitters at the respective sensor locations and forward these respective measurements to a central network node. The forwarded measurements may contain relevant labels such as the time of the measurements, the frequency band of the measurements, the location/coordinates of the sensor etc.

A common problem arising in these multi-transmitter scenarios is the complexity in distinguishing components of the measured RSS due to different radio transmitters. This complicates multi-transmitter localization problem. The problem is then to accurately determine the unknown number, the unknown power(s) and the unknown location(s) of the radio transmitters in the area.

Transmitter(s) localization techniques based on RSS are the most popular ones because of their simplicity and direct applicability to commercial off-the-shelf radio hardware. However, the literature extensively addresses the problem of single-source localization with the multiple source case receiving far less attention due to its inherent difficulty owed to:

i. the lack of commercial scenarios where multiple non-orthogonal sources overlap in the same band, and

ii. the rather high cost for deploying many sensors, which is a necessary condition for accurate localization.

However, the potential for cognitive-radio applications (see coexistence issues in IEEE 802.19 and 802.22) plus the decreasing cost of sensors are changing this reality. A number of approaches to the localization problem exist in the art. ML estimation

Maximum Likelihood (ML) is a popular estimation approach due to its asymptotic optimality. However, the likelihood function in a log-normal propagation environment typically possesses multiple maxima, thus resulting in a non-convex optimization problem. This becomes the most significant drawback of the ML approach. Standard techniques for such non-convex optimization tend to be very complex. In addition, computation increases exponentially with the number of sources which is another major drawback with the ML estimation.

The problem of local maxima for a single source has been addressed via Projection On Convex Sets (POCS). The algorithm is of low complexity, robust to local maxima and possesses a distributable computation. The distance estimates also employ RSS, but the algorithm can also be used for any positioning method where sensor-source distance estimates are somehow available. The method, named "circular" POCS, performs well when the (single) source node is located inside the sensor convex hull defined by the outer perimeter nodes in the sensor network. It also requires knowledge of parameters such as path loss and transmission power. The disadvantage with this solution is that it solves the single-source case only.

Another method for addressing the non-convexity of conventional ML is the Semi-Definite Programming (SDP) relaxation technique. The idea here is to convert the non-convex quadratic-distance constraints into linear constraints by introducing a relaxation that removes the quadratic term in the formulation. The SDP method has been only applied to the single- source case. In addition, the transmit power of the radio transmitters must be known.

Approaches with simplified channel models and known number of sources

For multiple sources, research has mostly focused on the AWGN model, which is valid for acoustic sources. The same AWGN model has been applied to radio applications, but this is an over-simplification for log-normal shadowing environments. Other prior art proposes a global optimization methodology. A clustering algorithm has been introduced based on k- means, for enhancing convergence and reducing overall complexity. For the same AWGN model, a quasi EM method has been proposed. All the above techniques assume knowledge of the number of sources. The inherent sparsity (small number of sources, large number of measurements) has been exploited in order to improve estimation by suitably modifying the Least- Absolute Shrinkage and Selection Operator (LASSO).

Grid-based approaches The first attempt at a full-blown ML was by approximating the sum of log-normal random variables. The challenge of not knowing the number of sources was addressed by placing a large number of hypothetical sources on a grid (i.e. each grid point corresponds to a potential source), thus leading to over-fitting issues. Mentioned over-fitting issues lead to bad performance.

Other approaches

Other prior art solves the problem by adopting a deterministic approach. The disadvantage here is that no shadow fading is taken into account and the robustness of the method in such environments suffers.

Thus, there is a need in the art for an improved method for determining powers and locations of radio transmitters in an area. Summary of the Invention

An object of the present invention is to provide a solution which mitigates or solves the drawbacks and problems of prior art solutions. Another object of the invention is to provide simplified solution to the above problem. According to a first aspect of the invention, the above mentioned objects are achieved by a method for determining powers and locations of j = 1, ... , K numbers of radio transmitters based on a plurality of received signal strengths measured at i = 1, ... , N_S number of sensors, where K and N_s are positive integers and K≥ 2 and N_s ≥ 1, respectively; the method comprising the step of:

using the expectation-maximization method for determining the powers P_j and locations

Xj of the j = 1, ... , K numbers of radio transmitters based on the plurality of received signal strengths measured at the i = 1, ... , N_s number of sensors.

Preferred embodiments of the invention are defined in the appended dependent claims. Any method according to the present invention can be executed as a computer program in processing means, and may be comprised in a computer program product. According to a second aspect of the invention, the above mentioned objects are achieved with a central processing unit comprising processing means for processing data and memory means for storing data, and further comprising receiving means and transmitting means for receiving and transmitting data, respectively, the central processing unit being arranged to determining powers and locations of j = 1, ... , K numbers of radio transmitters based on a plurality of received signal strengths measured at i = 1, ... , N_S number of sensors, where K and N_s are positive integers and K≥ 2 and N_s ≥ 1, respectively, by further being arranged to use the expectation-maximization method for determining the powers P_j and locations Xj of the j = 1, ... , K numbers of radio transmitters based on the plurality of received signal strengths measured at the i = 1, ... , N_s number of sensors.

The present invention provides a solution having an improved computational complexity (e.g. compared to the full-blown ML approach). Further, the present solution has better performance in terms of higher accuracy of the estimations compared to prior art solutions.

Further applications and advantages of the invention will be apparent from the following detailed description.

Brief Description of the Drawings

The appended drawings are intended to clarify and explain different embodiments of the present invention in which:

Fig. 1 shows an example problem scenario for 2 radio transmitters and 5 sensors in which the 2 transmitters operate at unknown locations in an area. The 5 distributed sensors measure received signal strength from the radio transmitters and forward measurements to a central processing unit; and

Fig. 2 shows a flow chart of an embodiment of the present invention.

Detailed Description of the Invention

To achieve the aforementioned and other objects, the present invention relates to a method for determining powers and locations of active radio transmitters in an area based on received signal strengths measured at a number of sensors distributed over the area. As mentioned above, this particular problem leads to very complex mathematical problems but the present method provides a simplified model which is possible to solve.

With reference to the problem statement, the received power at sensor z is, in linear scale (small letters signifying linear scale):

rss_t =∑f₌₁ rssi (1) where K is the number of radio transmitters (radio sources) in an area. A standard log- distance model for the received energy RSS_{t j} (capital letters signifying scale in dB) of the z^'-th sensor at location s_t from source j with transmit power P_j at distance d_{t j} is

RSSi_j = P_j + L_Q - PL (Xj, _Si) + Ui (2) where L₀ is a reference path loss and n_£ is the log-normal shadow-fading component with variance σ² and PL(x_j, s^) is a suitable deterministic path loss function. One possible choice of this function is

\\x-— s \\

PLixp St) = 10g log₁₀ ^J where d₀ is the reference distance, typically close to the transmitter chosen such that the path loss at d₀ (in free space) equals L₀, and is the path loss exponent.

Using eq. (2), the eq. (1) becomes a sum of log-normal variables and, with the standard Gaussian approximation for the sum of log-normal variables to approximate (1), the received power at sensor i is Gaussian distributed with mean _έ and variance a :

β55_έ~Ν(μ_έ (β), σ_έ ² (0)) (3) The means and variances are:

and

μ. (0) = c-¹

- - c a {e) + - ca² (5)

where

m_i (e_j) = P_j - PL(x_j, s_i) (6) is the mean energy received at the z^'-th sensor when only the j-th source is active, Θ = [x±, yi, 2 _> ?2,— ' ^χκΎκ> PR] is ^me J- T-length vector of the parameters to be estimated (x-coordinate, y-coordinate, and power), σ² is the common shadow fading variance and c = (In 10)/10 is a dB-scale normalization constant.

Using this signal model, the problem solved by the present invention is to then find the locations and powers collected in the vector Θ from a (large) number N_s of observed values rsS_j. Each of the N_s sensors gives one such observation.

One solution to the problem is found by the ML estimation, which involves the minimization of the negative log-likelihood function with respect to 0 (see [14]):

arg min L (RSS; 0) (7)

Θ

where RSS = RSS₂, ... ,

is the vector of observed measurements and

L (RSS; 0) = In (inaf (0)) + *⁵ ff > (8) where N_s is the number of sensors in the area. This minimization (search for 0) involves non- convex optimization of dimension 3K with multiple minima. In general this minimization is prohibitively complex.

The inventors have however imposed an approximation for each received power sample, which allows the mapping of the original problem to the Gaussian Mixture Model (GMM) fitting problem. The GMM is a probabilistic model representing the appearance of several Gaussian variables underlying the observed data. It is a probabilistic model for representing the presence of subpopulations within an overall population without requiring that an observed-dataset should identify the sub-population to which any individual observation belongs. Each observed data sample is modelled as having been generated by one of a group of Gaussian distributions (typically, with different means and variances); however, unknown by which at the time of observation. The process of associating the collected observations to these distributions and also finding their corresponding moments is usually referred to as "clustering". Typically, a GMM involves random variable y modelled as a weighted sum of Gaussian variables, i.e. : y~∑f₌₁ w^N^_j, a ^~) . The model not only represents the means and variances of the component Gaussian distributions, but also the prior probability (or prior mixture weight) for each of the components indicating the prior probability that an observed sample originates from that particular component. The present approximation consists of assigning one dominant source per sensor. If it is assumed that a single source contributes most of the measured power to each sensor, then:

rsSi « max_j rss; _j (9)

Empirical experience has shown this to be valid, especially for indoor scenarios where the path loss is a higher number. For example, assume that a sensor is at the same rough distance from two equal-power sources. Then the sum received power will just be 3dB above the single- source one. This 3 dB falls well within the usual indoor shadowing standard deviation. With this assumption the model becomes a GMM. Namely, using eq. (9) (in dB), the measurements are then approximated as N_s independent, identically distributed samples RSS , RSS₂ , . .. , RSS_Ns , derived from a GMM with K components of associated parameters 8_t =

Furthermore, each signal measurement is assumed to originate from exactly one of the unknown Gaussian distributions. The means of the K components are parameterized by i, because the mean value of each mode at the z^'-th sensor depends on its position.

The EM algorithm applied to estimation of the unknown parameters μ^ , af , and Wj of the GMM is

We apply it here to the localization problem and model derived in the previous section for the particular case where the variance for all Gaussian components are the same: af = σ² for all

]^■

The log-likelihood function of the G

L (RSS; Θ) = (e), a²)j

where the variances are no longer functions of the distances, since the shadowing is modeled to be the same for all transmitters. Hence, by carefully study the above stated problem and by making the above mentioned assumptions, the ML problem can now be solved with the EM algorithm. So, by using the EM algorithm on the measured signal strengths at the sensors the above stated problem can be solved as a simpler ML problem, i.e. GMM. It is noted that the EM algorithm has never before been used for solving this type of problem.

The method steps of an embodiment of the present invention (for a fixed number of radio transmitters) are given in the flow chart of Figure 2 and Table I below.

Table 1 : Algorithm for EM-GMM based localization of K radio transmitters sources from N_s sensor measurements for channel model in eq. (2) (where d₀ and L₀ are the reference distance and power loss, a is the path loss exponent and the variance of the log-normal shadow-fading component is σ²).

Input: number of sources K, measurements RSS s_£ for i = 1, ... , N_s

Output: locations $_j, powers P_j, for j = 1, ... , K, and likelihood L.

a) (Initialization step) Initialize the iteration-index m = 0, and, for all j = Ι, .,. , Κ, choose initial estimates: (powers), x ^ (locations), along with initial weights

Compute the initial means

and the initial log-likelihood

b) (Expectation-step) For all i = 1, ... , N_s, and j = 1, ... , K,

Compute

c) (Maximization step) Compute the new weight-estimates yJ _s (m)

(m+1) _

w J, γΚ yN_{s v}(m) ' (weights)

For each _/ = 1, ... , K, compute the location-estimate +¹⁾ = (locations)

where

Finally, for each j = 1, ... , K, compute P- = P_j ( ) ), (powers) d) (Convergence step) Compute the means

μ «+1) ₌ _ _Lq _ _PL (^⁺¹)_{j S} .) and the log-likelihood σ²))

if |L^^m+1^— Z ^m) I < 5 (a fixed predetermined threshold)

increase m: m = m + 1 (m is the iteration index) and return to step b, otherwise

output J,- = xj^m+1) and P_j = P_y ^(m+1) for ; = 1, AT, and L = Z ^m+1) .

In the first initialization-step, an initial estimate of the positions and powers in the vector Θ and the weights Wj are computed (for instance at random). These initialized values completely determine the RSS. The (scalar) likelihood of this initial solution Θ is also computed.

In the expectation step, for the assumed RSS, the probabilities that a measurement i is due to radio transmitter j is then computed. In the maximization step, estimate of the positions and powers in the vector Θ and the weights Wj are re-computed, now using the probabilities of the expectation- step.

In the convergence step, the scalar likelihood of the solution Θ is computed and if this likelihood exhibits an increase higher than a certain threshold value δ, (which can be an arbitrary small number heuristically chosen to control the convergence), and a new iteration is started by jumping back to the expectation step.

The specific equations involved in each of the steps are shown in Table I. The maximization step involves an optimization step, but, in contrast to the original ML-search, this step now involves a standard non-linear weighted Least-Squares (LS) problem, usually encountered in the problem of localizing a single transmitter source. Regardless of the searching algorithm employed to solve this optimization, the complexity of this search grows only linearly with the number of transmitters K.

According to a preferred embodiment of the invention, the number of radio transmitters K may also be determined which is another advantage with the present method. So far it has been assumed the number of radio sources K is known. Since the actual number is typically not known in advance, either the Akaike Information Criterion (AIC) or the Minimum Description Length (MDL) principle is applied.

The basic idea is to run the method in Table I for different number of assumed number of transmitters K and choose the number K for which the likelihood is highest. Since the likelihood increases with K, both the AIK and the MDL method in the literature provide the means to normalize the likelihood with respect to K and are used in a stopping criterion in a sequential search for K. In Table 2 the further method steps of the proposed localization algorithm without a prior assumption on the number of radio transmitters K is summarized.

Table 2: EM-GMM localization for unknown number of radio transmitter sources. Input: RSS s_t for i = 1, ... , N_s.

Output: K and $_{j 5} Pj for j = 1, ... , K.

1. Set K = 1. Then execute EM-GMM and store the resulting likelihood as L (l). 2. Increase K: K = K + 1. Then execute EM-GMM and store the resulting likelihood as L (K) .

Compute C(K) = -L (K) + K (AlC-method),

or compute C(K) =—L (K) + - \og N_s (MDL-method).

3. if C(K) < C(K - 1) return to step 2,

otherwise output K, and the estimates Xj , P_j for j = 1, ... , K obtained in the most recent execution of the EM-GMM method.

In either of the cases, the EM-algorithm is run for increasing K, starting with K=l, and stopped as soon as the criterion- value stops decreasing, that is when:

C(K) ≥ C (K - l) (1 1)

The value for the criterion typically decreases when K= \ , 2,... However, for a certain value of K the decrease stops and the algorithm is terminated.

According to another embodiment of the invention the AIC computes:

C(K) = -L(K) + K (12)

where L(K) is the likelihood value computed in step d) above.

According to yet another embodiment of the invention the MDL computes:

C0O = -L0O + ^ log N_s (13)

where L(K) is the likelihood value computed in step d) above.

The sensors are distributed over the area and may comprise only one receive antenna each. There is no restriction to the distribution of the sensors in the area so they can e.g. be random or deterministic. Regarding the radio transmitters it is assumed that they operate in overlapping frequency bands, and often in the same frequency band(s). For example, the radio transmitters can be base stations of a cellular operating system such as LTE.

Furthermore, as understood by the person skilled in the art, any method according to the present invention may also be implemented in a computer program, having code means, which when run by processing means causes the processing means to execute the steps of the method. The computer program is included in a computer readable medium of a computer program product. The computer readable medium may comprises of essentially any memory, such as a ROM (Read-Only Memory), a PROM (Programmable Read-Only Memory), an EPROM (Erasable PROM), a Flash memory, an EEPROM (Electrically Erasable PROM), or a hard disk drive.

Moreover, the invention also relates to a central processing unit arranged to execute the present method. It is therefore realised that the central processing unit may be modified, mutatis mutandis, according to different embodiments of the present method. The central processing unit comprises processing means for processing data and memory means for storing data, and further comprises suitable means such as receiving means and transmitting means for receiving and transmitting data, such as received signal strengths measured at the sensors. The present device is further arranged to use the above described EM method for determining the powers P_j and locations Xj of the plurality of radio transmitters based on the plurality of received signal strengths measured at sensors.

The central processing unit may according to an embodiment be integrated in a communication node of a radio access network (RAN), e.g. a base station or a radio network controller. By integrating the central processing unit in the communication node of a RAN the cost can be reduced.

However, the central processing unit may instead be implemented in, and being a part of an operation and management (O&M) entity controlled by mobile operators. Thereby, the information about the number, locations and powers of the radio transmitters is directly available at the network level which means that the information can be used for the beneficial operation and management of the radio network.

Finally, it should be understood that the present invention is not limited to the embodiments described above, but also relates to and incorporates all embodiments within the scope of the appended independent claims.

Claims

1. Method for determining powers and locations of j = 1, ... , K numbers of radio transmitters based on a plurality of received signal strengths measured at i = 1, ... , N_s number of sensors, where K and N_s are positive integers and K≥ 2 and N_s ≥ 1, respectively; the method comprising the step of:

using the expectation-maximization (EM) method for determining the powers P_j and locations Xj of the j = 1, ... , K numbers of radio transmitters based on the plurality of received signal strengths measured at the i = 1, ... , N_s number of sensors.

2. Method according to claim 1 , wherein the received power at sensor i in linear scale is rssi rsSi , and the method further comprises the step of:

assuming that rss^ « max_j rss; _j .

3. Method according to claim 2, wherein the step of using involves the steps of:

a) determining K number of initial powers and locations for the K number of radio transmitters, and determining an initial likelihood value based on the initial powers and locations;

b) computing a probability gamma yfy for each sensor-radio transmitter pair i,j based on the plurality of received signal strengths measured at the i = 1, ... , N_s number of sensors; c) determining updated K number of powers p -^{m+ )} and locations Xj^m+1) based on the probability gamma for each sensor-radio transmitter pair

d) determining a likelihood value L^^m+1^ based on the updated K number of powers p(.^m+1 _an(j i_ocatj_ons ^^rn+1'⁾; and e) repeating steps b) - d) above m times until a convergence criterion based on the likelihood-value is satisfied, where m is the iteration index.

4. Method according to claim 3, wherein step a) involves:

computing the initial likelihood value as,

L = ∑£₁ ln (∑ ₌₁ w⁽⁰⁾N ((«55_i | ,⁽⁽:⁾, ²

i,J where N

and variance σ² evaluated for the value RSS_t, and denotes a weight value associated with the jth transmitter.

5. Method according to claim 3, wherein step b) involves:

com uting the probability gamma for each sensor-radio transmitter pair i,j as,

where and

variance σ² evaluated for the value RSS_T , and denotes a weight value associated with the jth transmitter.

6. Method according to claim 3, wherein step c) involves:

computing the updated K number of locations Xj^{m+ 1}^ and the associated K number of powers P_j \ Xj ) as, +¹⁾ = arg

(RSS, - P_j{x) + PL{x, sd) where Ρ,Ο) = -^^ Σ -Ι^ - PLfa s )²,

^^•i=i Yi ,j

and where PL(x, s ) is a deterministic path loss function and s_t is the location of the ith sensor.

Method according to claim 3, wherein step d) involves:

computin the likelihood value L^^m+1^ as,

_L(m₊i) ¹⁾N ((/?55_i| _[ ^{( +1)}, ² where denotes the Gaussian distribution function with mean

μΐ^⁺¹^ and variance σ² evaluated for the value RSS_t, and Wj ^m+1^ denotes a weight value associated with the jth transmitter.

8. Method according to claim 3, wherein the convergence criterion is compared to a threshold value δ such that

- < δ.

9. Method according to claim 3, wherein the steps a) - d) are repeated for an assumed number of radio transmitters until a stop criterion is satisfied so as to obtain the number of radio transmitters K.

10. Method according to claim 9, wherein the stop criterion is the Akaike information criterion (AIC) and based on the value of: C(K) =—L (K) + K, where L (K) is the likelihood value Z °) determined in step d).

1 1. Method according to claim 9, wherein the stop criterion is the Minimum Description Length (MDL) criterion and based on the value of: C(K) =—L (K) + - log N_s, where L(K) is the likelihood value determined in step d).

12. Method according to claim 1 , wherein the method further comprises the step of: receiving the plurality of received signal strengths measured at i = 1, ... , N_S number of sensors.

13. Method according to claim 1 , wherein the i = 1, ... , N_S number of sensors are distributed over an area.

14. Method according to claim 1 , wherein the i = 1, ... , N_S number of sensors only comprises one receiving antenna each.

15. Method according to claim 1 , wherein the j = 1, ... , K numbers of radio transmitters operates in overlapping frequency bands.

16. Computer program, characterised in code means, which when run by processing means causes said processing means to execute said method according to any of claims 1-15.

17. Computer program product comprising a computer readable medium and a computer program according to claim 16, wherein said computer program is included in the computer readable medium, and comprises of one or more from the group: ROM (Read-Only Memory), PROM (Programmable ROM), EPROM (Erasable PROM), Flash memory, EEPROM (Electrically EPROM) and hard disk drive.

18. Central processing unit comprising processing means for processing data and memory means for storing data, and further comprising receiving means and transmitting means for receiving and transmitting data, respectively, the central processing unit being arranged to determining powers and locations of j = 1, ... , K numbers of radio transmitters based on a plurality of received signal strengths measured at i = 1, ... , N_S number of sensors, where K and N_s are positive integers and K≥ 2 and N_s ≥ 1, respectively, by further being arranged to use the expectation-maximization (EM) method for determining the powers P_j and locations Xj of the j = 1, ... , K numbers of radio transmitters based on the plurality of received signal strengths measured at the i = 1, ... , N_s number of sensors.

19. Central processing unit according to claim 18, wherein said central processing unit is integrated in a communication node of a radio access network (RAN).

20. Central processing unit according to claim 18, wherein said central processing unit is part of an operation and management (O&M) entity.