COGNITIVE FEMTOCELL

Cognitive Radio The limited available spectrum and the inefficiency in spectrum

usage. Necessitate A new communication paradigm to exploit the existing wireless spectrum opportunistically.

The core technology behind frequency reuse is cognitive radio.

2\84

THE CONCEPT

wireless communication in which the transmission or reception parameters are changed to communicate efficiently without interfering with licensed users

parameter changes are based on the active monitoring of several factors in the radio environment (e.g. radio frequency spectrum, user behavior and network state).

3\84

IEEE 802.22 IEEE 802.22 standard is known as cognitive radio standard

because of the cognitive features it contains. Proposed to standardize a fixed wireless access system based on cognitive radio technology to enable spectrum access and sharing by the secondary system. provided that the secondary system will not interfere the primary systems.

One of the most distinctive features of the IEEE 802.22 standard is its spectrum sensing requirement.

IEEE 802.22 based wireless regional area network (WRAN) devices sense TV channels and identify transmission opportunities. The functional requirements of the standard require at least 90% probability of detection and at most 10% probability of false alarm for TV signals with -116 dBm power level or above.

4\84

BACKGROUND

Most of the radio frequency spectrum inefficiently utilized

spectrum utilization depends strongly on time and place

fixed spectrum allocation wastes resources

improved efficiency by allowing unlicensed users to exploit spectrum whenever it would not cause interference to licensed users

5\84

6\84

MULTI-DIMENSIONAL SPECTRUM AWARENESS

Spectrum Opportunity : a band of frequencies that are not being used by the primary user of that band at a particular time in a particular geographic area.

Other Dimensions :

Code Dimension

Angle Dimension

7\84

FREQUENCY AND TIME DIMENSION

This involves the availability of a specific part of the spectrum in time. In other words the band is not continuously used. There will be times where it will be available for opportunistic usage.

8\84

GEOGRAPHICAL SPACE ANGLE

The spectrum can be available in some parts of the geographical area while it is occupied in some other parts at a given time. This takes advantage of the propagation loss (path loss) in space.

These measurements can be avoided by simply looking at the interference level. No interference means no primary user transmission in a local area.

9\84

CODE DIMENSION

The spectrum over a wideband might be used at a given time through spread spectrum or frequency hopping. This does not mean that there is no availability over this band. Simultaneous transmission without interfering with `primary users would be possible in code domain with an orthogonal code with respect to codes that primary users are using. This requires the opportunity in code domain. 10\84

ANGLE DIMENSION

Along with the knowledge of the location/position or direction of primary users, spectrum opportunities in angle dimension can be created. For example, if a primary user is transmitting in a specific direction, the secondary user can transmit in other directions without creating interference on the primary user.

11\84

SPECTRUM SENSING IN CURRENT WIRELESS STANDARDS

802.11k is An IEEE standard for how a wireless local area network (WLAN) should perform channel selection and transmit power control (TPC) in order to optimize network performance.

802.11k is intended to improve the management of radio resources by sharing information about neighboring access points as well as more easily detecting transmission strength.

12\84

AP collects channel information from each mobile unit and makes its own measurements. This data is then used by the AP to regulate access to a given channel.

The sensing information is used to improve the traffic distribution within a network as well.

WLAN devices usually connect to the AP that has the strongest signal level. Sometimes, such an arrangement might not be optimum and can cause overloading on one AP and underutilization of others.

In 802.11k, when an AP with the strongest signal power is loaded to its full capacity, new subscriber units are assigned to one of the underutilized APs. Despite the fact that the received signal level is weaker, the overall system throughput is better due to more efficient utilization of network resources.

13\84

BLUETOOTH

A new feature, namely adaptive frequency hopping (AFH), is introduced to the Bluetooth standard to reduce interference between wireless technologies sharing the 2.4GHz unlicensed radio spectrum.

In this band, IEEE 802.11b/g devices, cordless telephones, and microwave ovens use the same wireless frequencies as Bluetooth.

AFH identifies the transmissions in the industrial, scientific and medical (ISM) band and avoids their frequencies.

14\84

Bluetooth transmission with and without adaptive frequency hopping. AFH prevents collusions between WLAN and Bluetooth transmissions.

AFH requires a sensing algorithm for determining whether there are other devices present in the ISM band and whether or not to avoid them.

15\84

TERMINOLOGY Capability of cognitive radio – Full cognitive radio Every possible parameter observable by a wireless

node or network is taken into account – Spectrum sensing cognitive radio Only radio frequency band is taken into account

Licenses of spectrum band – Licensed band cognitive radio • Primary network (user) • Secondary network (user) – Unlicensed band cognitive radio

16\84

COGNITIVE RADIO CHARACTERISTICS

Two main characteristics of the cognitive radio:

1 .Cognitive Capability.

2 .Reconfigurability:

– Operating Frequency

– Modulation

– Transmission Power

– Communication Technology

17\84

Spectrum sensingDetecting unused spectrum.

Spectrum managementCapturing the best available spectrum to meet user communication requirements.

Spectrum mobilityThe process where a cognitive radio user exchanges its frequencyof operation.

Spectrum sharingProviding fair spectrum scheduling method.

COGNITIVE RADIO MAIN FUNCTIONS

18\84

SPECTRUM SENSING Spectrum sensing is often considered as a simple detection

problem. However the key challenge is the detection of the weak signal in real environment corrupted by noise and suffering from interference. while minimizing the time spent in sensing.

Spectrum sensing techniques are divided into :

– Non-cooperative detection

• Matched filter detection

• Energy detection

• Cyclostationary feature detection

–Cooperative detection

• Method where information from multiple users are incorporated for primary user detection

• Cooperation with primary network & secondary users 19\84

SPECTRUM SENSING

20\84

1) Matched Filter Detection Maximizes the SNR of the received signal in the presence of

AWGN. Requires a priori knowledge of the primary user signal.

2) Energy Detection Doesn’t need any knowledge of PU signals. RF energy is measured over an observation time to determine

whether the spectrum is occupied or not. It performs poorly under low SNR because of not

differentiating the interference from PU and noise. Can not differentiate signal types but can only determine the

presence of the signal. 21\84

3 )Cyclostationary Feature Detection The idea is to exploit the built-in periodicity of the modulated

signal. A signal is said to be cyclostationary if its autocorrelation is a

periodic function of time with some period. The main advantage of the spectral correlation function is that it

differentiates the noise energy from modulated signal energy. Implementation of cyclostationary detection is more complicated,

and requires longer observation time than energy detectors.4 )Cooperative Detection

Cooperative sensing decreases the probabilities of mis-detection and false alarm considerably. It can solve hidden primary user problem and it can decrease sensing time.

Share spectrum sensing result. Collaborative spectrum sensing is most effective when

collaborating cognitive radios observe independent fading or shadowing. 22\84

PILOT DETECTION

The detection of weak deterministic signals in additive noise. The signal power is confined inside a priori known bandwidth B around central frequency fc .

We assume that activity outside of this band is unknown. In addition, we assume that primary user transmitter sends a pilot signal simultaneously with data. The sensing receiver has a perfect knowledge of the pilot signal and can perform its coherent processing.

23\84

The detection is the test of the following two hypotheses:

H0: Y [n] = W[n] signal absentH1: Y [n] = Xp[n]+ W[n] signal presentn = 1,.........., N , where N is observation interval

Xp[n] is the known pilot data with power θ and W[n] is white Gaussian noise with variance

The optimal detector is the matched filter that projects the received signal in the direction of the pilot.

24\84

Under either hypothesis Y[n] is jointly Gaussian random variable, and since T is a linear combination of jointly Gaussian random variables, it is also Gaussian. Thus,

Then Pd and Pfa can be evaluated as:

If the number of samples used in sensing is not limited, this pilot detector can meet any desired Pd and Pfa simultaneously. The minimum number of samples is a function of the signal to noise ratio

25\84

The theoretical analysis shows that coherent processing given enough samples, arbitrary weak signals can be detected. However, this benefit comes at the cost of perfect synchronization required to demodulate the pilot.

Let consider the sinewave pilot:

Suppose there is a frequency offset between the primary transmitter and sensing receiver:

Then, the decision statistic becomes:

If the sensing interval N becomes comparable or larger than the period of the frequency offset (2π/w), then the decision statistic looses coherent processing gain and eventually becomes equal to zero. Therefore, in the presence of frequency offset the pilot detection has limits on sensing time and detectable signal levels.

26\84

We measured performance for three different frequency offsets.

In the case of perfect synchronization, measurements show complete agreement with the theoretical results )N~1/SNR).

As a result, extremely weak signals measured up to -136dBm can be detected.

Required sensing time vs. input signal level for fixed Pd and Pfa

for sine wave pilot in the presence of frequency offset.

27\84

However, in the presence of frequency offsets coherent processing gains are limited by the processing time.

For example, given a 10Hz frequency offset, if the receiver sense longer than 30ms, it can never meet the desired Pd and Pfa. Similarly, for a 100Hz offset sensing times are limited to 3ms.

Due to sensing time constraints, signal levels below -132dBm and -120dBm can not be detected in the presence of 10Hz and 100Hz frequency offset, respectively.

We say that the receiver hits the SNR wall when the detection of weaker signals is not possible

Required sensing time vs. input signal level for fixed Pd and Pfa for sine wave pilot in the

presence of frequency offset.

28\84

ENERGY DETECTOR In some cases, an optimal detector based on matched

filter is not an option since it would require the knowledge of the pilot data and perfect synchronization for coherent processing. Instead a simple energy detector is adopted, which can be applied to any signal type.

Spectrum sensing is required by CR users both before and during the use of licensed spectrum bands.

The concept of energy detection mechanism is quite simple. The detector computes the energy of the received signal and compares it to certain threshold value to decide whether the desired signal is present or not. 29\84

Time domain representation of energy detection mechanism

Frequency domain representation of energy detection mechanism

ENERGY DETECTOR STRUCTURE

30\84

PHYSICAL SYSTEM TO IMPLEMENT ENERGY DETECTOR

31\84

The oscilloscope comprises of the analog to digital converter. So, it can be used as a sampling unit. However if the RF signal be directly fed to the oscilloscope, the sampling frequency is supposed to be very high. So, the received signal is first subjected to the spectrum analyzer.

The spectrum analyzer contains the frequency converting unit i.e. the mixer which generates the intermediate frequency (IF) signal. This signal will then be fed to the oscilloscope so that the sampling frequency needed is practically feasible.

the spectrum analyzer comprises of the attenuator which prevents overloading in case of strong signals.

the band pass filter selects the signal component with specific center frequency and specific bandwidth.

The low noise amplifier amplifies the signal so that the spectrum analyzer can detect it.

32\84

SINGLE USER SENSING ANALYSIS

Two hypotheses are considered: •Hypothesis 0 (H0): the primary user is inactive

H0 : x[n] = w[n], )White space) •Hypothesis 1 (H1): the primary user is active

H1 : x[n] = s[n] + w[n], )occupied)Where n = 0, 1, 2, ....,N − 1, where N is the observation interval , w[n] is the noise and s[n] is the primary signal required to detect.

•If H0 is true then the decision value will be less than the threshold γ On the other hand, if H1 is true then the received signal has both

signal and noise, the decision value will be larger than the threshold γ.

The threshold value is chosen so as to control the parameters such as probability of false alarm and probability of detection.

33\84

If the number of samples used in sensing is not limited, an energy detector can meet any desired Pd and Pfa simultaneously. The minimum number of samples is a function of the signal to noise ratio SNR=

In the low SNR << 1 regime, number of samples required for the detection, that meets specified Pd and Pfa, scales as O(1/SNR2). This inverse quadratic scaling is significantly inferior to the optimum matched filter detector whose sensing time scales as O(1/SNR).

Unfortunately, an increased sensing time is not the only disadvantage of the energy detector. More importantly, there is a minimum SNR below which signal cannot be detected. This minimum SNR level is referred to SNRwall.

34\84

NOISE VARIANCE UNCERTAINTY

There, we have made two very strong assumptions. First, we assumed that noise is white, additive and Gaussian, with zero mean and known variance. However, noise is an aggregation of various sources. Second, we assumed that noise variance is precisely known to the receiver, so that the threshold can be set accordingly.

However, this is practically impossible as noise could vary over time due to temperature change, ambient interference, filtering, etc. Even if the receiver estimates it, there is a resulting estimation error due to limited amount of time. Therefore, our model needs to incorporate the measure of noise variance uncertainty.

35\84

How does the noise uncertainty affect detection of signals in low SNR?

Essentially, setting the threshold too high based on the wrong noise variance, would never allow the signal to be detected. If there is a x dB noise uncertainty, then the detection is impossible below:

For example, if there is a 0.03 dB uncertainty in the noise variance, then the signal in -21 dB SNR cannot be detected using the energy detector.

36\84

a.QPSK sensing

Measurements show that theoretical expectation is observed & when the signal becomes too weak, increasing the number of averages does not improve the detection.

REQUIRED SENSING TIME VS. SIGNAL INPUT LEVEL FOR FIXED PD AND PFA

37\84

a. QPSK sensing

This result is expected and is explained by the SNRwall existence. The limit happens at -110dBm (SNRwall=-25dB). From the theoretical analysis, we know that SNRwall=-25dB corresponds to less than 0.03 dB of noise uncertainty

REQUIRED SENSING TIME VS. SIGNAL INPUT LEVEL FOR FIXED PD AND PFA

38\84

we notice that increasing the FFT size improves the slope of the scaling law. For 1024 pt. FFT we obtained

For 256 pt. FFT we observe scaling law of

Although there is an improved sensing time for sinewave sensing by increasing FFt size, the non-coherent processing in this implementation makes it sensitive to noise uncertainty. In the case of 1024 pt FFT, Signals below -128 dBm cannot be detected, resulting in the SNRwall=-25dB.

b- Sine wave sensing

39\84

The SNRwall for 256 pt. FFT happens for signal levels lower than -122 dBm.

As expected, there is a 6dB improvement in SNRwall going from 256 pt. to 1024 pt. FFT as a consequence of 4 times longer coherent processing time.

b- Sine wave sensing

40\84

FALSE ALARM & MISS DETECTION ERROR

A high Pf results in low spectrum utilization since the false-alarms increase the number of missed opportunities (white spaces).

A high Pm results in increasing the interference inflicted on the primary licensee.

41\84

FALSE ALARM AND DETECTION PROBABILITIES OVER AWGN CHANNELS The probability of false alarm is the percentage of white spaces

falsely declared occupied (i.e. the percentage of missed opportunities).

= Pr {Y > Vth|H0}= exp(-Vth²/ 2No)

The probability of detection determines the level of interference protection provided to the primary licensee.

= Pr {Y > Vth|H1}= Q (A/ √No, Vth/ √No )

probability of miss detection = 1-

42

42\84

When Vth is large, decrease which is good but also decrease which is bad.

When Vth is small, increase which is bad but also increase which is good.

43\84

44\84

SENSING-THROUGHPUT TRADEOFF

Objective: Determination of the optimal T for each frame such that the achievable throughput of the secondary network is maximized while the primary users are sufficiently protected.

45\84

CYCLOSTATIONARY FEATURE DETECTOR

These modulated signals are characterized as cyclostationarity since their mean and autocorrelation exhibit periodicity (digital modulated signals have non-random components). This particular property is called cyclostationarity which means that the statistical parameters of the signal vary periodically in time.

On the other hand, noise is not cyclostationary, therefore noise can be distinguished from signal by analyzing the cyclostationary properties of the received samples.

The main advantage of the spectral correlation function is that it differentiates the noise energy from modulated signal energy, which is a result of the fact that the noise is a wide-sense stationary signal with no correlation, while modulated signals are cyclostationary with spectral correlation due to the embedded redundancy of signal periodicity.

46\84

The signal is said to be wide-sense cyclostationary with period if is cyclic in with cycle i.e.,

Cyclostationary process has periodic autocorrelation function.

we define the autocorrelation function as :

Where, E[.] represents statistical expectation operator. Since is periodic, it has Fourier series representation.

47\84

Where sum is taken over integer multiple of fundamental cycle frequencies , .

= m/T0 and m is an integer. The term is known as cyclic auto- correlation

function, which is defined as:

Consider a time series of length T, the expectation in the definition of auto-correlation can be replaced by time average so that for a deterministic time series , we define the cyclic autocorrelation function as:

48\84

The spectral correlation is an important feature of the cyclostationary signals. If the signal x(t) exhibits cyclostationary with cyclic frequency , in time domain, then it also exhibits spectral correlation at shift in frequency domain.

So the SCD could be measured by the normalized correlation between two spectral components of x(t) at frequencies

(f+ /2) and (f- /2) over an interval T as given by:

49\84

In order to estimate the power in a frequency band, we simply pass the signal x(t) into a narrow disjoint band pass filter and measure the average power of the output. We can estimate the signal's PSD. That is, at any particular frequency f, the PSD of x(t) is given by:

50\84

Where is the impulse response of an ideal band pass filter with center frequency f and bandwidth B. For estimating the SCD, we pass the frequency translated signals and

through same set of band pass filters and then measure the temporal correlation of the filtered signals. The estimated SCD is given by the equation.

Implementation using FFT

51\84

Frequency resolution: In order to resolve features need to have sufficient resolution in f and α spectral resolution in f can be increased by

Cycle resolution: Cycle resolution depends on the total observation interval Increases the resolution in α by smoothing and

52\84

53/84

Example: Cycle Resolution Improvement

BPSK at carrier

Model

Hypothesis testing: Is the primary signal out there?

x(n) is primary user modulated signal with known

w(n) is noise with zero mean and unknown power N0 that could vary over time.

Spectral correlation function of y(n):

Noise is not cyclostationary process thus =0 for α≠0.

)()( :0 nwnyH

54

54\84

DETECTION METHOD

The conventional energy detection corresponds to testing

the energy levels obtained from at α=0 for the presence and absence of signal where as the signal feature detection based on spectral correlation of cyclostationarity is based on scanning of a peak cyclic spectrum magnitude of the signals at one of their cyclic frequencies. If the peak cyclic spectrum magnitude is found the signal is present, Otherwise the signal is absent.

55\84

SPECTRAL CORRELATION OF DIFFERENT SIGNALS

Peak value can be seen in both the zero cyclic frequency and unique cyclic frequencies. The peaks in zero cyclic frequency can be used in the energy detection while the peaks in unique cyclic frequencies can be used in our proposed detection method.

56\84

Unique cyclic frequency searching: The generated spectral correlation function of the received signal is used to search for the unique cyclic frequencies. The method is to detect the peak values in the frequency-cyclic frequency plane. The primary signals SCF as shown exhibit peaks at the unique cyclic frequency and zero cyclic frequency.

Detection decision: The detection decision is based on the searching results of unique cyclic frequency searching stage. As we have explained, the noise does not exhibit cyclostationarity. Therefore, if no unique cyclic frequencies are found, it means that there is no signal in the detected band. Otherwise, the band is used by the primary users.

57\84

When Eb/N0 equals to -5dB the peak values of the primary user signals in unique cyclic frequencies can obviously be seen. The peaks in power spectral density can also be detected by the radiometry based method at cyclic frequency equals to zero.

58\84

In the lower Eb/N0 levels, take -10dB for example, the peaks in the spectral correlation whelmed by the noise. Although the spectral correlation of the noise is zero when the cyclic frequency does not equal to zero, the peaks are whelmed by the noise due to the cross-spectral correlation between the signals and noise. Therefore, the spectral correlation based peak detection is used for the low noise level environment.

59\84

In order to maintain the noise robust advantage of the detection method, the contour figure of the spectral correlation function which describes the visibility of the cyclostationary signals among noise is used for searching the unique patterns of different primary user signals.

The contour figure of the AM modulated signal in the noise free environment. Four clear point of the signal are seen.

60\84

The unique pattern is the intersection point which denotes the unique cyclic frequency of the AM modulated signal.

The dark line is caused by the cross-spectral correlation effect. The intersection points of these dark lines denote the unique cyclic frequency which distinguished the AM modulated signal from the noise. The back ground areas are caused by the noise. As the noise level increasing, the background areas get darker and the visibility of the lines and intersection points decreases. 61\84

HYBRID DETECTOR To increase accuracy and optimize the detection

probability of cognitive radio user a hybrid sensing algorithms is proposed. Actually this algorithm is the cascading result of the Energy based sensing and Cyclostationary feature detection.

Energy detector is a good solution to detect free bands because no a priori information is needed. On the contrary, cyclostationary detection is robust, but computationally expensive. Since these two methods are complementary we have proposed our iterative hybrid architecture (HSD) which permits to detect quickly with minimum a priori information free bands by taking advantage of both methods.

62\84

DECISION RULE OF THE E-HSD ALGORITHM

63\84

It is assumed that No is constant in time. Let Xi be the energy of the N received samples during an observation time T after the ith iteration, ξ a variable threshold that is first initialized at + ∞ .

1- At the beginning of the sensing the energy detector calculates the energy X of the received samples after an observation time T .

2- If X falls inside the interval [0,ξ ], the energy detector cannot make a direct decision of type signal present or signal absent.

3- The cyclostationary block will be used to make the decision

if the signal present (or the signal absent).

4- The calculated value X is then saved in a buffer called buffer2 of size N2 (or buffer1 of size N1 ).

64\84

5- The algorithm continues in the same way except when buffer2 is full. In this case, the adaptation stage starts to modify the value of the threshold ξ according to the average of buffer2 and then the oldest value in the buffer will be replaced by the new calculated one ( Xi after the ith iteration).

6- At any time, if the calculated value X is greater than ξ the adaptation stage will take automatic decision of type signal present avoiding the use of the cyclic test.

7- When buffer1 is full, its mean μ1 will be calculated which will be used to determine No (No= μ1 / BT), then using No we can estimate ξ that guarantees the desired false alarm, from the equation .

8- It should be noted that N1 is big enough (bigger than N2 ) in order to make a good estimation of No. 65\84

The same probability of false alarm = 0.1 is imposed on all three sensing schemes.

the detection performance versus SNR for the three sensing schemes for sensing times T1 = 2 ms and T2 = 18 ms. As we can see, for an SNR that is less than −12 dB, the two-stage sensing scheme performs better than either energy detection or cyclostationary detection.

66\84

In this Hybrid sensing algorithm the energy based detector is used to verify whether primary user is present or not. Here Cyclostationary algorithm is used just to get the features (modulation, operating frequency, no. of signal) of primary user when primary is present and is used as detector when energy detector is not sure about the presence or absence of primary user.

67\84

68\84

RESOURCES SCHEDULING

The procedure for implementing the Cognitive femtocell access control is illustrated in five recurring steps: Environment Sensing, User Classification, Reservation Determination, Resource Allocation, and Dynamic Reconfiguration.

69\84

1 .ENVIRONMENT SENSING: After the FAP is powered on and the initialization process has

completed, the first step in the Cognitive access policy is the Environment Sensing. The FAP in this step of the procedure will scan the surrounding environment and listen for any UEs requesting access in uplink (UL) channels within its femtocell coverage.

Usually an UE will attempt to access the base station, either the MBS or nearby FAP, of which the UE receives the highest pilot signal power.

70\84

2 .USER CLASSIFICATION: During this stage, all nearby users within the femtocell

coverage are identified. Femto-users refer to active users that are currently connected to the FAP while Macro-users refer to those connected to the MBS.

Femto-user’s unique 15-digit IMSI is examined to see if it is listed on the AAL of the FAP. If the IMSI is indeed listed on the AAL, that Femto-user is then classified as an Owner of the FAP; otherwise, the Femto-user will be classified as a Visitor of the FAP.

71\84

3 .RESERVATION DETERMINATION:

In this step, the Owner density DFAP of the FAP is calculated as the number of Owners over the total number of Femto-users that are currently connected to the FAP. The value of the owner density DFAP will normally be different for each FAP at different time instance.

Furthermore, another parameter called femtocell resource reservation coefficient KFAP is defined to be the percentage of femtocell resources in FAP that are dedicated for Owners.

the higher value of KFAP means more resource is reserved for Owners, or less resource is reserved for Visitors

72\84

4 .RESOURCE ALLOCATION: Once the reservation coefficient KFAP is set based on the

Owner density DFAP, the FAP then allocates the determined portion of femtocell resources and dedicated them for Owners, while the remaining portion is reserved for Visitors.

In the case of WCDMA, the total amount of power is considered as the sum of pilot and data power. Since the pilot power is broadcast for the purpose of defining femtocell coverage, thus only the data power from the FAP is divided between Owners and Visitors.

73\84

Resource blocks consisting of time slots and sub-carrier frequencies should be used as the distribution element in femtocell resource reservation.

74\84

5 .DYNAMIC RECONFIGURATION:

The word cognitive refers to the ability for mobile terminals to observe the surrounding environment, and then react accordingly to achieve better performance.

Therefore, after all femtocell resources are distributed, the FAP utilizing the Cognitive access strategy continues to monitor the surrounding environment and then make appropriate adjustments to both DFAP and KFAP dynamically so that the overall network performance can be maintained in an optimum state.

75\84

OPERATION MODE

With different values of KFAP , the corresponding Cognitive access policy is expected to behave differently. For FAPs that utilize the Cognitive access control, four operation modes: Sleep, Private, Public, and Hybrid, are introduced.

Sleep mode: When no UEs are currently connected, or when the FAP is first powered on . However, once any UE is connected, the FAP will no longer stay in the Sleep mode and will be transited to either the Private mode or the Public mode depending on the user class of that connected user.

Under the Sleep mode, the power level of FAP’s pilot signal is reduced to prevent unnecessary interference with other nearby FAPs.

76\84

Private mode: When one or more UEs are currently connected to the FAP and all are Owners who are listed on the AAL of the FAP.

The femtocell resource reservation coefficient KFAP is set to 1. By doing this, Owners will experience the best femtocell performance from their FAPs.

Public mode: When one or more UEs are currently connected to the FAP but all connected UEs are Visitors who are not on the AAL of the FAP.

Resource reservation coefficient KFAP is set to 0. All available femtocell resources are equally distributed among all connected Visitors.

77\84

Hybrid mode: When two or more UEs are currently connected to the FAP where mixed types of user classes are observed.

value of KFAP for each FAP is dynamically adjusted based on current environment situation.

78\84

RELATIONSHIPS AND TRANSITIONS

the relationships between the resource reservation coefficient K and FAP operation modes.

Owner density D is used as the boundary condition between public and hybrid modes.

79\84

FAP is serving one Owner and three Visitors with 20% resource reservation (K = 0.2), the Owner would only receive 20% of femtocell resources if the FAP is operating in Hybrid mode. However, the Owner will get a higher portion of resource at 25% if the FAP is operating in the Public mode.

In a different case where the FAP is serving two Owners plus three Visitors, the value of K should be set to at least 40%.

For a FAP operating in Hybrid mode, its value of K should be set to a value smaller than 1 but greater than the current Owner density D.

80\84

TRANSITION MAP OF OPERATION MODE

81\84

POWER MEASURMENT When determining whether to connect to the central

MBS or a nearby FAP, each user will detect the received pilot signals and then attempt to connect itself to the base station with the highest received pilot power. The received signal power R is calculated from the transmitted signal power P with corresponding path loss PL and antenna gain G

82\84

SIMULATION PROCESS FLOW

83\84

Thank you