Intelligent deployment cascade control device based on an FDD-OFDMA indoor small cell in multi-user and interference environments

Abstract

The invention presents an intelligent deployment cascade control (IDCC) device for frequency division duplexing (FDD)-orthogonal frequency division multiplexing access (OFDMA) indoor small cell to enable easy installation, multi-user (MU)service reliability, optimum throughput, power saving, minimum interference and good cell coverage. The proposed IDCC device is designed with a cascade architecture, which mainly contains five units including a resource allocator, a minimum throughput/cell edge CQI converter, an adaptive neural fuzzy inference system (ANFIS) based initial transmit power setting controller (ITPSC) in the first cascade unit, an ANFIS based channel quality index (CQI) decision controller (CQIDC) in the second cascade unit and an ANFIS based self-optimization power controller (SOPC) in the third cascade unit. The SOPC consists of three parts, namely the transmit power adjustment estimator (TPAE), transmission power assignment and self-optimization power controller protection mechanism.

Claims

1. A frequency division duplexing-orthogonal frequency division multiplexing access (FDD-OFDMA) based adaptive neural fuzzy inference system (ANFIS) intelligent deployment cascade control (IDCC) device for indoor small cell operated in the multi-user (MU) and interference environments to self-optimize the service reliability, throughput, minimum transmit power and interference for multimedia call services, comprising: an indoor small cell base station, wherein resources of the indoor small cell base station are allocated to multiple user equipments (UEs); a resource allocator, configured to assigns the average resource blocks (RBs) of small cell for each indoor user according to the total number of indoor users and the setting system bandwidth; a minimum throughput/cell edge channel quality index (CQI) converter, configured to set the cell edge (minimum) channel quality index for each indoor user in accordance with the minimum throughput requirement; an adaptive neural fuzzy inference system based initial transmit power setting controller (ITPSC) in the first cascade unit, configured to adapt the initial power setting for the uth user to the coverage radius of indoor office, the number of the resource blocks and the cell edge channel quality index; an adaptive neural fuzzy inference system based channel quality index decision controller (CQIDC) in the second cascade unit, configured to adapt the best channel quality index to the initial power setting, number of the resource blocks and average path loss (PL) measured by the user equipment; and an adaptive neural fuzzy inference system based self-optimizing power controller (SOPC) in the third cascade unit consists of three parts, namely the transmit power adjustment estimator (TPAE), transmission power assignment and self-optimization power controller protection mechanism; wherein, it can autonomously cascade control the assignments of initial power, the best channel quality index and the minimum transmit power to the transceiver according to the user input parameters including the service reliability, coverage radius and the throughput at the cell edge; the measured average path loss and average signal-to-interference-plus-noise ratio (SINR), so that the present intelligent deployment cascade control device can self-optimize the service reliability of the indoor small cell in the multi-user and interference environments, while maintaining the blocking error rate (BLER) less than 10-1 and minimizing the transmit power and interference power to achieve the design aims of energy saving and interference reducing.

2. The frequency division duplexing-orthogonal frequency division multiplexing access based adaptive neural fuzzy inference system intelligent deployment cascade control device for indoor small cell operated in the multi-user and interference environments of claim 1, wherein the architecture of adaptive neural fuzzy inference system based initial transmit power setting controller unit contains five tiers, a total of three inputs and one output; three input parameters for the u.sub.th user are the coverage radius of indoor office (R.sub.u), the number of resource blocks (nRB.sub.u) and the cell edge channel quality index that is defined as CQI.sub.min,u, the output parameter for the u.sub.th user is an initial minimum transmit power (P.sub.ini,u); each input uses three generalized bell shape membership functions (MFs); each MF contains three levels; the 27 fuzzy inference rules are constructed; a minimum transmit power optimization problem of the adaptive neural fuzzy inference system initial transmit power setting controller is formally formulated as follows: $\mathrm{optimize}{P}_{\mathrm{ini}}=f\left(\stackrel{.fwdarw.}{x}\right),f\left(\stackrel{.fwdarw.}{x}\right)\mathrm{is}\mathrm{the}\mathrm{objective}\mathrm{function};$ $\mathrm{subject}\mathrm{to}\text{:}$ $\stackrel{.fwdarw.}{x}\left\{R_u,{\mathrm{nRB}}_u,{\mathrm{CQI}}_{\min ,u}\right\}$ $0m<R_{u}15m$ $1{\mathrm{nRB}}_{u}100$ $1{\mathrm{CQI}}_{\min ,u}15$ $\underset{u}{\overset{\mathrm{nUE}}{.Math.}}{P}_{\mathrm{ini},u}\left\{20\mathrm{dBm}\right\}.$

3. The frequency division duplexing-orthogonal frequency division multiplexing access based adaptive neural fuzzy inference system intelligent deployment cascade control device for indoor small cell operated in the multi-user and interference environments of claim 2, wherein, for the adaptive neural fuzzy inference system based initial transmit power setting controller in the multi-user environments of claim 2, wherein the training data is generated from the simulation results of the transceiver blocking error rate to train the premise and consequent parameters of the initial transmit power setting controller; the minimum transmit power (dBm) training data of the initial transmit power setting controller for the u.sub.th user is given by:
P.sub.ini,u=P.sub.min,u(CQI.sub.min,u)+L.sub.tG.sub.t+PL(R.sub.u)+FM(SR.sub.u)G.sub.r+L.sub.r; wherein P.sub.rmin,u(CQI.sub.min,u)is the receiver sensitivity of the cell edge channel quality index (CQI.sub.min,u) for the u.sub.th user; L.sub.t denotes the cable loss in dB; G.sub.t and G.sub.r are the antenna gains in dBi of the femtocell and the user equipment, respectively; PL(R.sub.u) denotes the maximum path loss between a femtocell and the u.sub.th user at the cell edge, L.sub.r in dB is the body loss of the user equipment; FM(SR.sub.u) denotes fade margin in dB corresponding to the SRset by the u.sub.th user; the receiver sensitivity of the given cell edge channel quality index (CQI.sub.min,u) for the u.sub.th user is obtained by
P.sub.rmin,u(CQI.sub.min,u)=P.sub.N,u+SNR.sub.th(CQI.sub.min,u); wherein SNR.sub.th(CQI.sub.min,u)denotes the SNR threshold of the receiver for different CQI.sub.min,u,which is generated from the performance simulations using the transceiver; the training data for minimum transmit power is generated for the service reliability of 90%, different coverage radius (2.5, 5, 7.5, 10, 12.5 and 15 meters), different resource block (1100) and cell edge channel quality index (115).

4. The frequency division duplexing-orthogonal frequency division multiplexing access based adaptive neural fuzzy inference system intelligent deployment cascade control device for indoor small cell operated in the multi-user and interference environments of claim 1, wherein the architecture of the adaptive neural fuzzy inference system based channel quality index decision controller unit contains five tiers, a total of three inputs and one output for the u.sub.th user; there are three input parameters for the u.sub.th user including path loss (PL.sub.u.) between orthogonal frequency division multiplexing (OFDM) transmitter and receiver, initial power setting (P.sub.ini,u) and number of the resource blocks; in the interference free environments, the channel quality index decision controller unit adapts the best channel quality index (CQI.sub.best,u) to the changing initial power setting, number of the resource blocks and the measured average path loss; the Gaussian shape membership function of each input parameter is divided into three levels; there are 27 fuzzy inference rules; an optimization problem of channel quality index the best channel quality index of the adaptive neural fuzzy inference system based channel quality index decision controller is formally formulated as follows: in the interference free environments, BLER0.1; optimize CQI.sub.best,u =(x) at the u.sub.thuser, (x) is the objective function: subject to :
x{PL.sub.u, P.sub.ini,u, nRB.sub.u}
30dB PL.sub.u70dB
75dBm P.sub.ini,u 20dBm
1nRB.sub.u100
CQI.sub.best,u {115 }.

5. The frequency division duplexing-orthogonal frequency division multiplexing access based adaptive neural fuzzy inference system intelligent deployment cascade control device for indoor small cell operated in the multi-user and interference environments of claim 4, wherein, for the adaptive neural fuzzy inference system based channel quality index decision controller, the training data is used to train the premise and consequent parameters of the channel quality index decision controller; the training data of the best channel quality index (CQI.sub.best,u) at the u.sub.th user's location of the indoor office in the interference free environment is determined by the following rules: ${\mathrm{CQI}}_{\mathrm{best},u}=\{\begin{array}{cc}\mathrm{CQI\_i},& \mathrm{if}{\mathrm{SNR}}_{\mathrm{th}}\left(\mathrm{CQI\_i}\right){\mathrm{SNR}}_u<{\mathrm{SNR}}_{\mathrm{th}}\left(\mathrm{CQI\_i}+1\right),i=114\\ \mathrm{CQI\_}15,& \mathrm{if}{\mathrm{SNR}}_u{\mathrm{SNR}}_{\mathrm{th}}\left(\mathrm{CQI\_}15\right)\\ \mathrm{CQI\_}1,& \mathrm{if}{\mathrm{SNR}}_u<{\mathrm{SNR}}_{\mathrm{th}}\left(\mathrm{CQI\_}1\right)\end{array}$ the signal-to-noise-power ratio (SNR) is estimated by:
SNR.sub.u=P.sub.r,u(W)/P.sub.N,u(W); wherein the average received power P.sub.r,uat the u.sub.th user in the interference free environment is given as:
P.sub.r,u=P.sub.ini,uL.sub.t+G.sub.tPL.sub.u+G.sub.rL.sub.r; wherein PL.sub.u denotes the measured average path loss between a femtocell eNB and an UE in the given cell; P.sub.N,u is the noise power for the u.sub.th user; the training data of the channel quality index decision controller is generated from the simulation results of the transceiver blocking error rate for different measured average path loss (30 dB70 dB), the resource block (1100) and initial minimum transmit power (75 dBm20 dBm).

6. The frequency division duplexing-orthogonal frequency division multiplexing access based adaptive neural fuzzy inference system intelligent deployment cascade control device for indoor small cell operated in the multi-user and interference environments of claim 1, wherein the self-optimizing power controller unit consists of three parts, namely a transmit power adjustment estimator, a transmission power assignment and an self-optimization power controller protection mechanism; wherein the adaptive neural fuzzy inference system based transmit power adjustment estimator in the interference environment primarily adapts the transmit power to the requested throughput at the cell edge (corresponding to the cell edge channel quality index), the best channel quality index and measured average signal-to-interference-plus-noise ratio and estimates the amount of minimum transmit power adjustment needs for each user; wherein the transmission power assignment adjusts the power for each indoor user when the sum of total transmission power to all indoor users doesn't exceed the maximum transmit power limit of the eNB; wherein a protection mechanism for self-optimizing power controller is used to prevent the co-channel interference from the moving users of adjacent cells.

7. The frequency division duplexing-orthogonal frequency division multiplexing access based adaptive neural fuzzy inference system intelligent deployment cascade control device for indoor small cell operated in the multi-user and interference environments of claim 6, wherein the adaptive neural fuzzy inference system based transmit power adjustment estimator in the self-optimizing power controller unit contains five tiers; wherein the transmit power adjustment estimator unit accepts three inputs and generates the optimizing minimum transmit power; wherein three inputs for the u.sub.th user including cell edge channel quality index (CQI.sub.min,u), best channel quality index (CQI.sub.best,u) and average measured signal-to-interference-plus-noise ratio (SINR.sub.u)and one power adjustment output for the u.sub.th user; the generalized bell shape membership function of each input parameter is divided into three levels; there are 27 fuzzy inference rules; an optimization problem of transmit power adjustment estimator the minimum transmit power of the adaptive neural fuzzy inference system based transmit power adjustment estimator in the self-optimizing power controller unit is formally formulated as follows: $\mathrm{in}\mathrm{the}\mathrm{interference}\mathrm{environment},\mathrm{BLER}0.1,\mathrm{optimize}{P}_u=f\left(\stackrel{.fwdarw.}{x}\right)\mathrm{at}\mathrm{the}{u}_{\mathrm{th}}\mathrm{user},f\left(\stackrel{.fwdarw.}{x}\right)\mathrm{is}\mathrm{the}\mathrm{objective}\mathrm{function}\text{:}$ $\stackrel{.fwdarw.}{x}\left\{{\mathrm{CQI}}_{\min ,u},{\mathrm{CQI}}_{\mathrm{best},u}\mathrm{and}{\stackrel{\_}{\mathrm{SINR}}}_u\right\}$ $1{\mathrm{CQI}}_{\min ,u}15$ $1{\mathrm{CQI}}_{\mathrm{best},u}15-25\mathrm{dB}{\stackrel{\_}{\mathrm{SINR}}}_{u}45\mathrm{dB}$ $\underset{u}{\overset{\mathrm{nUE}}{.Math.}}{P}_u+P_u\left\{20\mathrm{dBm}\right\},P_u\mathrm{is}\mathrm{the}\mathrm{transmit}\mathrm{power}\left(\mathrm{dBm}\right)$ $\mathrm{at}\mathrm{the}\mathrm{last}\mathrm{instant}.$

8. The frequency division duplexing-orthogonal frequency division multiplexing access based adaptive neural fuzzy inference system intelligent deployment cascade control device for indoor small cell operated in the multi-user and interference environments of claim 7, wherein, for the transmit power adjustment estimator, the training data is generated to train the premise and consequent parameters of the transmit power adjustment estimator; for the purpose of satisfying the requirements of blocking error rate 10% and the SR.sub.u,the target threshold of the signal to interference plus noise ratio (SINR.sub.th,u) at the u.sub.th user is defined as;
SINR.sub.th,u=max{SNR.sub.th(CQI.sub.min,u)+FM(SR.sub.u),SNR.sub.th(CQI.sub.best,u)}(dB); The output power adjustment (P.sub.u) at the u.sub.th user is given by subtracting the SINR.sub.th,ufrom the measured average signal-to-interference-plus-noise ratio at the u.sub.th user; using the defined SINR.sub.th,uthe training data of the transmit power adjustment value P.sub.u is generated for service reliability (90%), cell edge channel quality index (115), measured average signal-to-interference-plus-noise ratio (25 dB45 dB) and the best channel quality index (115).

9. The frequency division duplexing-orthogonal frequency division multiplexing access based adaptive neural fuzzy inference system intelligent deployment cascade control device for indoor small cell operated in the multi-user and interference environments of claim 7, wherein, for the protection mechanism of the self-optimizing power controller , the intelligent deployment cascade control device inputs the average path loss measured from the user equipment, and then by the indoor path loss model to estimate the distance (d) between the user equipment and the eNB; if the moving user equipment is inside the coverage range of the radius, then the transmission power assignment of the self-optimizing power controller unit is initiated to transmit the minimum power to the moving user equipment of the adjacent cells; otherwise, the transmission power assignment of the self-optimizing power controller unit is disabled to the moving user equipment of the adjacent cells.

10. The frequency division duplexing-orthogonal frequency division multiplexing access based adaptive neural fuzzy inference system intelligent deployment cascade control device for indoor small cell operated in the multi-user and interference environments of claim 1, wherein the device is based on the frequency division duplexing-orthogonal frequency division multiplexing access method, and uses the adaptive neural fuzzy inference system architecture to adapt the initial power setting to the requested resource block, throughput at the cell edge and coverage radius in the interference free environment; to adapt the best channel quality index to the initial setting power, number of the resource blocks and average path loss measured by user equipment in the interference free environment; to adapt the transmit power assignment to the requested throughput at the cell edge, the best channel quality index and measured average signal-to-interference-plus-noise ratio in the interference environment; the present intelligent deployment cascade control device is designed to self-optimize the signal-to-interference-plus-noise ratio and throughput service reliability of the indoor small cell in the multi-user and interference environments, while maintaining the blocking error rate less than 10.sup.1 and minimizing the transmit power and interference power to achieve the aims of energy saving and interference reducing.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The present disclosure will become more fully understood from the detailed description given herein below and the accompanying drawings which are given by way of illustration only, and thus are not limitative of the present disclosure and wherein:

(2) FIG. 1 is a block diagram of intelligent deployment cascade control (IDCC) device for FDD-OFDMA indoor small cell.

(3) FIG. 2 shows the architecture of an ANFIS based initial transmit power setting controller (ITPSC) unit.

(4) FIG. 3 shows the architecture of an ANFIS based channel quality index (CQI) decision controller (CQIDC) unit.

(5) FIG. 4 shows the architecture of an ANFIS based transmit power adjustment estimator in SOPC unit.

(6) FIG. 5 illustrates the BLER performance of DL OFDM transceiver in IOA channel for CQI=1, 2 . . . , 15.

(7) FIG. 6 illustrates average throughput and system capacity of DL OFDM transceiver in IOA channel for CQI=1, 2 . . . , 15.

(8) FIG. 7 shows the scenario of laboratory with femtocell eNB and measurement points.

(9) FIG. 8 shows the modified path loss model obtained from the field trial in the laboratory.

(10) FIG. 9 shows a set of training data for ITPSC unit: The minimum transmit power corresponding to coverage radius=5 m, the cell edge CQI=3, 7, 10 and number of resource block=1 to 100.

(11) FIG. 10(a) shows the initial coverage radius (R) membership functions of ITPSC unit.

(12) FIG. 10(b) shows the initial number of resource block (nRB) membership functions of ITPSC unit.

(13) FIG. 10(c) shows the initial cell edge CQI (CQI.sub.min) membership functions of ITPSC unit.

(14) FIG. 11(a) shows the learned coverage radius (R) membership functions of ITPSC unit.

(15) FIG. 11(b) shows the learned number of resource block (nRB) membership functions of ITPSC unit.

(16) FIG. 11(c) shows the learned cell edge CQI (CQI.sub.min) membership functions of ITPSC unit.

(17) FIG. 12 shows the root mean square error of ITPSC unit.

(18) FIG. 13 shows a set of training data for CQIDC unit: The best CQI corresponding to the initial transmit power of ITPSC=40 dBm, 20 dBm, 0 dBm, average path loss=30 dB to 70 dB and number of resource block=1 to 100.

(19) FIG. 14(a) shows the initial average path loss (PL) membership functions of CQIDC unit.

(20) FIG. 14(b) shows the initial transmit power initialization (P.sub.ini) membership functions of CQIDC unit.

(21) FIG. 14(c) shows the initial number of resource block (nRB) membership functions of CQIDC unit.

(22) FIG. 15(a) shows the learned average path loss (PL) membership functions of CQIDC unit.

(23) FIG. 15(b) shows the learned transmit power initialization (P.sub.ini) membership functions of CQIDC unit.

(24) FIG. 15(c) shows the learned number of resource block (nRB) membership functions of CQIDC unit.

(25) FIG. 16 shows the root mean square error of CQIDC unit.

(26) FIG. 17 shows a set of training data for the ANFIS based TPAE in the SOPC unit: The optimum power adjustment corresponding to CQI.sub.min=6, the best CQI=7, 10, and different average measured SINR.

(27) FIG. 18(a) shows the initial cell edge CQI (CQI.sub.min) membership functions of the ANFIS based TPAE in the SOPC unit.

(28) FIG. 18(b) shows the initial best CQI (CQI.sub.best) membership functions of the ANFIS based TPAE in the SOPC unit.

(29) FIG. 18(c) shows the initial measured average SINR (SINR) membership functions of the ANFIS based TPAE in the SOPC unit.

(30) FIG. 19(a) shows the learned cell edge CQI (CQI.sub.min) membership functions of the ANFIS based TPAE in the SOPC unit.

(31) FIG. 19(b) shows the learned best CQI (CQI.sub.best) membership functions of power increment estimator unit.

(32) FIG. 19(c) shows the learned measured average SINR (SINR) membership functions of the ANFIS based TPAE in the SOPC unit.

(33) FIG. 20 illustrates the root mean square error of the ANFIS based TPAE in the SOPC unit.

(34) FIG. 21 shows the flow chart of power allocation algorithm.

(35) FIG. 22 shows the simulation flow chart of IDCC.

(36) FIG. 23(a) shows the CCDF of SINR measurement for different UEs using fixed power and the IDCC in interference free environment.

(37) FIG. 23(b) shows the CCDF of SINR measurement for different UEs using fixed power and the IDCC in interference power=90 dbm environment.

(38) FIG. 23(c) shows the CCDF of SINR measurement for different UEs using fixed power and the IDCC in interference power=80 dbm environment.

(39) FIG. 24(a) shows the CCDF of throughput for different UEs using fixed power and the IDCC in interference free environment.

(40) FIG. 24(b) shows the CCDF of throughput for different UEs using fixed power and the IDCC in interference power=90 dbm environment.

(41) FIG. 24(c) shows the CCDF of throughput for different UEs using fixed power and the IDCC in interference power=80 dbm environment.

(42) FIG. 25 shows the average throughput of DL transceiver for different UEs using fixed power and the IDCC in the MU interference environment.

(43) FIG. 26 shows the average transmit power of DL OFDM transceiver for different UEs using the IDCC in the MU interference environment.

DETAILED DESCRIPTION

(44) For your esteemed members of reviewing committee to further understand and recognize the fulfilled functions and structural characteristics of the invention, several preferable embodiments cooperating with detailed description are presented as the follows.

(45) The invention presents an adaptive neural fuzzy inference system (ANFIS) based intelligent deployment cascade control (IDCC) device for FDD-OFDMA indoor small cell operated in the multi-user (MU) and interference environments to self-optimize the MU service reliability (SR), average throughput, transmit power and interference for multimedia call services.

(46) The principal structure of the present invention is an ANFIS based IDCC device as shown in FIG. 1; which mainly contains five units including a resource allocator, a minimum throughput/cell edge CQI converter, an adaptive neural fuzzy inference system (ANFIS) based initial transmit power setting controller (ITPSC) in the first cascade unit, an ANFIS based channel quality index (CQI) decision controller (CQIDC) in the second cascade unit and an ANFIS based self-optimization power controller (SOPC) in the third cascade unit. The SOPC consists of three parts, namely the transmit power adjustment estimator (TPAE), transmission power assignment and self-optimization power controller protection mechanism.

(47) In order to complete the intelligent deployment of small cells, the present invention is to use adaptive network architecture established by Jjh Shing Roger Jang in 1993, known as ANFIS, which is a fuzzy inference system. By using a hybrid learning method, the weights of ANFIS controller are adjusted to the appropriate value. The user inputs the parameters including the service reliability, coverage radius and the throughput at the cell edge. The user equipment (UE) measures the reference signal received power (RSRP) and sends back the estimated average path loss (PL) and signal-to-interference-plus-noise ratio (SINR) to the IDCC device. The proposed IDCC device is design to self-optimize the minimum transmit power of the indoor small cell in the multi-user (MU) and interference environments, while maintaining the blocking error rate (BLER) of the transceiver less than 10-1, and satisfying the requirements of average throughput and service reliability for the UE.

(48) The architecture diagram of ANFIS based ITPSC unit is shown in FIG. 2, which contains five layers, a total of three inputs and one output. Three input parameters for the u.sub.th user are the coverage radius of indoor office (R.sub.u), the number of resource blocks (nRB.sub.u) and the cell edge CQI that is defined as CQI.sub.min,u, the output parameter for the u.sub.th user is an initial minimum transmit power (P.sub.ini,u). The ITPSC unit adapts the initial power setting to the changing R.sub.u, nRB.sub.u and CQI.sub.min,u. The generalized bell shape membership function of each input parameter for the u.sub.th user is divided into three levels. There are 27 fuzzy inference rules. The ITPSC unit defines the R.sub.u of less than 5 meters as low (L), it defines 5 meters10 meters as medium (M), it defines 11 meters 15 meters as high (H); it defines the nRB.sub.u of 1 RB25 RBs as low (L), it defines 26 RBs74 RBs as medium (M), it defines 75 RBs100 RBs as high (H); it defines CQI.sub.min,u in 15 for L, it defines CQI.sub.min,u in 610 for M, it defines CQI.sub.min,u in 1115 for H; the output for the u.sub.th user is the initial transmit power setting (P.sub.ini,u), which satisfies the u.sub.th user requests under interference free environments.

(49) The architecture diagram of ANFIS based CQIDC unit is shown in FIG. 3, which contains five layers, a total of three inputs and one output. There are three input parameters for the u.sub.th user including average path loss (PL.sub.u) between orthogonal frequency division multiplexing (OFDM) transmitter and receiver, initial transmit power setting (P.sub.ini,u) and number of resource blocks (nRB.sub.u), the output parameter for the u.sub.th user is the best CQI (CQI.sub.best,u). Under interference free environments, the CQIDC unit adapts the best CQI to the changing PL.sub.u, P.sub.ini,u and nRB.sub.u. The Gaussian shape membership function of each input parameter is divided into three levels. There are 27 fuzzy inference rules. The CQIDC unit defines the average path loss (PL.sub.u) in 30 dB40 dB for L, it defines average path loss (PL.sub.u) in 41 dB60 dB for M, it defines average path loss (PL.sub.u) in 61 dB70 dB for H; it defines initial transmit power setting (P.sub.ini,u) in 75 dBm51 dBm for L, it defines initial transmit power setting (P.sub.ini,u) in 50 dBm4 dBm for M, it defines the initial transmit power setting (P.sub.ini,u) in 3 dBm20 dBm for H; it defines the nRB.sub.u of 1 RB25 RBs as low (L), it defines 26 RBs 74 RBs as medium (M), it defines 75 RBs100 RBs as high (H). The output for the u.sub.th user is the best CQI (CQI.sub.best,u) under interference free environments.

(50) The SOPC consists of three parts, namely the transmit power adjustment estimator (TPAE), transmission power assignment and self-optimization power controller protection mechanism. The power adjustment estimator in the interference environment primarily estimates the amount of minimum transmit power adjustment needs for each user; the transmission power for each user is adjusted when the sum doesn't exceed the maximum transmit power limit. The protection mechanism of the SOPC is used to prevent the co-channel interference from the moving users of adjacent cells.

(51) The architecture diagram of ANFIS based TPAE in the SOPC unit is shown in FIG. 4, which contains five tiers, a total of three inputs for the u.sub.th user including cell edge CQI (CQI.sub.min,u), best CQI (CQI.sub.best,u) and average measured SINR (SINR.sub.u), the output parameter for the u.sub.th user is a power adjustment (P.sub.u). In the interference environment, the ANFIS based TPAE in the SOPC unit adapts output power adjustment P.sub.u to the changing CQI.sub.min,u, CQI.sub.best,u and SINR.sub.u. The SOPC unit will be coordinated with CQIDC unit to set the minimum transmit power for the transceiver, which switches to the corresponding modulation mode and coding rate. The SOPC unit accepts three inputs and generates the optimizing minimum transmit power. The SOPC unit will continue to estimate the average SINR value in the MU interference environments. The transmit power of OFDM transceiver will be adjusted with the changing interference power to maintain the BLER less than 10.sup.1 and to satisfy the requirement of SR for the u.sub.th user; The generalized bell shape membership function of each input parameter is divided into three levels. There are 27 fuzzy inference rules. The ANFIS based TPAE in the SOPC unit defines the CQI.sub.min,u in 15 for L, it defines CQI.sub.min,u in 610 for M, it defines CQI.sub.min,u in 1115 for H; it defines the CQI.sub.best,u in 15 for L, it defines CQI.sub.best,u in 610 for M, it defines the CQI.sub.best,u in 1115 for H; it defines the SINR.sub.u less than 25 dB5 dB for L, it defines the SINR.sub.u in 4 dB25 dB for M, it defines the SINR.sub.u in 26 dB45 dB for H. The signal-to-interference-plus-noise ratio (SINR) is estimated through the reference signal received power (RSRP) measured from the user equipment (UE) and sends back to the IDCC device in e Node B (eNB). The estimated path loss (PL) is obtained by subtracting RSRP from the transmit reference signal.

(52) (A) The Architecture of the ANFIS Controller:

(53) The ANFIS based TPAE in the SOPC unit is used as an example to describe the framework of the ANFIS controller:

(54) Layer 1: The generalized bell shape membership functions are defined as:

(55) $\begin{array}{cc}{A}_{j,n}\left(x_{j,m}\right)=\frac{1}{1+{.Math.\frac{x_{j,m}-c_{j,n}}{a_{j,n}}.Math.}^{2b_{j,n}}},\mathrm{for}n=1,2,3\mathrm{and}j=1,2,3& \left(1\right)\end{array}$

(56) where x.sub.j,m is the m.sub.th input and the premise parameters a.sub.j,n, b.sub.j,n, c.sub.j,n pertaining to the node outputs are updated according to given training data and the gradient descent approach.

(57) Layer 2: The output of node i, denoted by O.sub.2,i, is the product of all the incoming signals for the i.sub.th rule. It is given by:
w.sub.i,m=O.sub.2,i=A.sub.1,p(x.sub.1,m)A.sub.2,q(x.sub.2,m)A.sub.3,r(x.sub.3,m)
for i=1,2,27;p=1,2,3;q=1,2,3;r=1,2,3(2)

(58) Layer 3: The output of node i, denoted by O.sub.3,i, is called the normalized firing strength and calculated as:

(59) $\begin{array}{cc}{O}_{3,i}={\hat{w}}_{i,m}=\frac{w_{i,m}}{\underset{i=1}{\overset{27}{.Math.}}{w}_{i,m}},\mathrm{for}i=127& \left(3\right)\end{array}$

(60) Layer 4: Every node in the fourth layer is an adaptive node with a node function:
O.sub.4,i=.sub.i,mf.sub.i,m=.sub.i,m(.sub.ix.sub.1,m+.sub.ix.sub.2,m+.sub.ix.sub.3,m+.sub.i);
for i=127(4)

(61) where O.sub.4,i is the node output, f.sub.i,m is a crisp output in the consequence, and the .sub.i, .sub.i, .sub.i, .sub.i are the consequent parameters of node i. The 27 fuzzy inference rules of f.sub.i,m are constructed as follows:
R.sub.1: if (x.sub.1,m is A.sub.11) and (x.sub.2,m is A.sub.21) and (x.sub.3,m is A.sub.31) then (output is f.sub.1,m);
R.sub.2: if (x.sub.1,m is A.sub.11) and (x.sub.2,m is A.sub.21) and (x.sub.3,m is A.sub.32) then (output is f.sub.2,m);
R.sub.3: if (x.sub.1,m is A.sub.11) and (x.sub.2,m is A.sub.21) and (x.sub.3,m is A.sub.33) then (output is f.sub.3,m);
R.sub.26: if (x.sub.1,m is A.sub.13) and (x.sub.2,m is A.sub.23) and (x.sub.3,m is A.sub.32) then (output is f.sub.26,m);

(62) . . .
R.sub.27: if (x.sub.1,m is A.sub.13) and (x.sub.2,m is A.sub.23) and (x.sub.3,m is A.sub.33) then (output is f.sub.27,m)(5)

(63) The above 27 fuzzy inference rules are used for determining the assigned data rate to achieve optimization objective.

(64) Layer 5: The single node in the fifth layer is a fixed node labeled , which computes the overall output O.sub.5 as the summation of all incoming signals.

(65) $\begin{array}{cc}{G}_m=O_5=\underset{i=1}{\overset{27}{.Math.}}{\hat{w}}_{i,m}{f}_{i,m}& \left(6\right)\end{array}$

(66) (B) The Minimum Throughput/Cell Edge CQI Conversion Unit:

(67) In order to satisfy the user requirements of indoor small cell in throughput and blocking error rate (BLER) of less than 10.sup.1, the relationship between the throughput and SINR threshold for the different CQI must be obtained. Therefore, the BLER and throughput of the LTE downlink (DL) transceiver for indoor small cell are simulated to generate the training data for the ANFIS ITPSC. The system parameters are shown in Table 1 and fundamental parameters of the transceiver are shown in Table 2. In the simulation of the present embodiment, for the different channel quality index (CQI), the BLER of 11 SISO-OFDM transceiver is simulated where the system bandwidth is 20 MHz, the indoor office A (IOA) channel is selected as channel model, the least square (LS) channel estimation and minimum mean square error (MMSE) equalizer are used, and the user speed is assumed to be 10 km/hr. The 1000 sub frames are applied for the simulations. The results are shown in FIG. 5, which is used as a training data to define the SINR threshold for BLER=10.sup.1 under different CQI, as shown in Table 3. In each CQI (corresponding to each pair of modulation and code rate mode), the difference between the SINR measurement value and the SINR threshold value is used to control the size of the transmit power adjustment P, which must compliance with the provisions of Table 3, where the interference power is not considered. The physical resource blocks (PRBs) of Table 3 is generated according to the system bandwidth of 20 MHz and three indoor users. When the system bandwidth is 20 MHz, the corresponding total number of physical resource blocks (PRBs) is 100 PRBs.

(68) The resource assignment method of this invention is the orthogonal frequency division multiplexing access (OFDMA) for the frequency division duplexing (FDD) mode of indoor small cell operated in the multi-user (MU) environments. The eNB of the indoor office will perform the resource allocation for each UE with 33 RBs at each time instant. For practical implementation considerations, the system capacity of the downlink (DL) OFDM transceiver formula is modified as [10]:

(69) $\begin{array}{cc}C\left(\mathrm{bps}\right)=\mathrm{BW}.Math.\mathrm{BW\_eff}.Math.\frac{{\mathrm{nRB}}_u}{{\mathrm{nRB}}_{\mathrm{total}}}.Math..Math.{\log }_2\left(1+\mathrm{SINR}/\mathrm{SINR\_eff}\right)& \left(7\right)\end{array}$

(70) where nRB.sub.total is the total number of RBs and nRB.sub.u is the number of RBs assigned for the u.sub.th user; BW and BW_eff are system bandwidth and effective system bandwidth, respectively. The parameter is a correction factor. SINR and SINR_eff are signal to interference plus noise power ratio and effective signal to interference plus noise power ratio, respectively. In this invention, the simulation parameters of DL SISO OFDM transceiver is given in Table 1, where BW=20 MHz, BW_eff=0.83, =0.43 and SINR_eff=2.51199 (4 dB). The average throughput of DL transceiver in IOA channel for CQI=1, 2 . . . , 15 is shown in FIG. 6, where the system capacity is denoted by dotted curve and the simulated average throughput for CQI=1, 2 . . . , 15 are denoted by solid curve. It is observed that the simulated average throughput approximates to the Shannon capacity bound for LTE specifications. Thus, the throughputs corresponding to SINR threshold for the different CQIs are calculated with (7) and listed in Table 3. As can be seen from table, the input throughput requirement settings for the u.sub.th user in the range of 1.992.76 Mbps, 6.767.44 Mbps, 12.9914.13 Mbps correspond to the cell edge CQI (CQI.sub.min,u) in 3, 7, and 10, respectively.

(71) (C) Initial Transmit Power Setting Controller (ITPSC) Unit:

(72) In order to control the initial transmit power of small cell eNB for satisfying the requirements of the u.sub.th user, the BLER performance of the LTE downlink (DL) transceiver is simulated to generate the training data for the ITPSC. This invention considers multi-user system reliability (SR) requirements of indoor small cell in fading environments. The received signal strength P.sub.r at the UE is log-normally distributed. The coverage probability of P.sub.r greater than the receiver sensitivity P.sub.r,min from the femtocell to a UE at the distance d is:

(73) $\begin{array}{cc}\begin{array}{c}{P}_{W_0}\left(d\right)=p\left[P_r{P}_{r,\min }\right]={}_{W_0}^{}p\left(P_r\right){\mathrm{dP}}_r\\ =\frac{1}{2}-\frac{1}{2}\mathrm{erf}\left(\frac{P_{r,\min }-K+10N{\log }_{10}\left(\frac{d}{R}\right)}{\sqrt{2}{}_W}\right)\end{array}& \left(8\right)\end{array}$

(74) where R is the coverage radius, K is the average signal strength (dBm) at the cell edge, KP.sub.r,min (dB) is the fade margin (FM) at the cell edge (d=R) which is used to guarantee the reliability at the cell edge, .sub.W is the standard deviation of received signal strength (dB) and N is the path loss exponent.

(75) The percentage of the UE in a cell of radius R for P.sub.r greater than the receiver sensitivity P.sub.r,min is defined as the service reliability (SR), which is given as:

(76) $\begin{array}{cc}SR=\frac{1}{2}\left\{1+\mathrm{erf}\left(p\right)+\exp \left(\frac{2\mathrm{pq}+1}{q^2}\right)\left[1-\mathrm{erf}\left(\frac{\mathrm{pq}+1}{q}\right)\right]\right\}.,& \left(9\right)\\ \mathrm{where}& \\ p=\frac{FM}{\sqrt{2}{}_W},q=\frac{10N{\log }_{10}e}{\sqrt{2}{}_W},& \left(10\right)\end{array}$

(77) The minimum transmit power of the ITPSC is evaluated by link budget formula for the different SR, coverage radius (R.sub.u) of indoor office, and the cell edge CQI (CQI.sub.min,u) requested by the u.sub.th user. The minimum transmit power in dBm of the ITPSC is given by:
P.sub.ini,u=P.sub.rmin,u(CQI.sub.min,u)+L.sub.tG.sub.t+PL(R.sub.u)+FM(SR.sub.u)G.sub.r+L.sub.r(11)

(78) where P.sub.rmin,u(CQI.sub.min,u) is the receiver sensitivity of the cell edge CQI (CQI.sub.min,u) for the u.sub.th user. L.sub.t denotes the cable loss in dB. G.sub.t and G.sub.r are the antenna gains in dBi of the femtocell and the UE, respectively. PL(R.sub.u) denotes the maximum path loss between a femtocell and the uth user at the cell edge. L.sub.r in dB is the body loss of the UE. FM(SR.sub.u) denotes fade margin in dB corresponding to the SR set by the u.sub.th user. The receiver sensitivity of the given cell edge CQI (CQI.sub.min,u) for the u.sub.th user is obtained by:
P.sub.rmin,u(CQI.sub.min,u)=P.sub.N,u+SNR.sub.th(CQI.sub.min,u)(12)

(79) where SNR.sub.th(CQI.sub.min,u) denotes the SNR threshold of the receiver for different CQI.sub.min,u, which is generated from the performance simulations using the transceiver specification listed in Table 2. The receiver noise power P.sub.N,u in dBm for the u.sub.th user is given as:
P.sub.N,u=NF(dB)+(174)+10 log.sub.10(BW.sub.r,u)(dBm)(13)

(80) where NF is the noise FIG. of the UE receiver and BW.sub.r,u is the receiver bandwidth.
BW.sub.r,u=15 kHz12nRB.sub.u(14)

(81) where nRB.sub.u is the allocated RBs of the u.sub.th user. The SNR thresholds for BLER=0.1 are summarized in Table 3. Using the ITU-R indoor path loss model [12], the path loss between a femtocell eNB and an UE separated by a distance d (m) in a given cell is
PL(d)=20 log.sub.10(f)+10N log.sub.10(d)+L.sub.f(n)28 (dB)(15)

(82) where the carrier frequency f (MHz) is set as 2350 MHz with 20 MHz bandwidth in the experiment. N is the path loss exponent, where the nominal value in the indoor office is set as 3 [12]. L.sub.f(n) (dB) is the penetration loss between the floors, where n is the number of floors. The penetration loss is not considered in the simulations.

(83) In addition, the standard deviation .sub.W of the received shadow fading signal power in the indoor office environment is set as 10 dB.

(84) Experiment Measurements in the Laboratory:

(85) For the purpose of determining the path loss exponent N and the standard deviation .sub.W of the received shadow fading signal in the indoor office environments, the power measurement of small cell eNB (ITRI-SC-CUT3) is performed in the laboratory. The scenario of laboratory is shown in FIG. 7, where the femtocell eNB and measurement positions are illustrated. The length and width of the laboratory are 20.98 meter and 7.30 meter, respectively. The black stars located in the laboratory represent 38 measurement positions, which are distributed at distance of 119 meters from the eNB. 1000 RSRPs are measured at each measurement position.

(86) UE (Samsung Galaxy Note Edge SM-N915G) used with drive test tool report RSRPs for different distances between transmitter and receiver in the laboratory and calculate their standard deviation. Then the path loss model of the laboratory can be obtained by modifying the ITU-R indoor office path loss model. FIG. 8 shows the modified path loss model obtained from the field trial in the laboratory, where circle solid curve is the ITU-R indoor office path loss model; the triangular dotted curve is the measured path loss model; the square solid curve is the modified path loss model of the laboratory. Therefore, the formula of the ITU path loss model is modified as
PL(d)=20 log.sub.10(f)+28 log.sub.10(d)36 (dB)(16)

(87) where the standard deviation .sub.W of the received shadow fading signal and the path loss exponent N are 4.27 dB and 2.8, respectively. Finally, by substituting O and N into (9)(10), the fade margin FM for 90% service reliability is calculated as 2.14 dB.

(88) The training data of the ITPSC is generated from the simulation results of the transceiver BLER, as shown in Table 3. Integrating Table 3 with equations (11), (12), (13), (14) and (16), the minimum transmit power is calculated for the service reliability of 90%, different coverage radius (2.5, 5, 7.5, 10, 12.5 and 15 meters), different number of resource block (1100) and cell edge CQI (115). FIG. 9 shows a set of training data of the ITPSC for coverage radius of 5 meter, cell edge CQI (3, 7, 10) and different number of resource block (1100). For example, when the number of resource block is 50, the minimum transmit power corresponding to different cell edge CQI (3, 7, 10) is (35.33 dBm, 27.33 dBm, 18.33 dBm).

(89) The function of the ITPSC is to set the initial minimum transmit power of the femtocell eNB, which satisfies the requirements of the different RBs (nRB.sub.u), coverage radius (R.sub.u) of indoor office, and the cell edge CQI (CQI.sub.min,u) requested by the u.sub.th user in the interference free environments. Each input uses three generalized bell shape membership functions (MFs), which are defined as:

(90) $\begin{array}{cc}{A}_{j,n}\left(x_{j,m}\right)=\frac{1}{1+{.Math.\frac{x_{j,m}-c_{j,n}}{a_{j,n}}.Math.}^{2b_{j,n}}},\mathrm{for}n=1,2,3\mathrm{and}j=1,2,3& \left(17\right)\end{array}$

(91) where x.sub.j,m is the m.sub.th input and the premise parameters a.sub.j,n, b.sub.j,n, c.sub.j,n pertaining to the node outputs are updated according to given training data and the steepest descent approach. The 27 fuzzy inference rules are constructed in Table 4. A minimum transmit power optimization problem of the ANFIS-ITPSC for the u.sub.th user is formally formulated as follows:

(92) $\begin{array}{cc}\mathrm{Optimize}{P}_{\mathrm{ini}}=f\left(\stackrel{.fwdarw.}{x}\right),f\left(\stackrel{.fwdarw.}{x}\right)\mathrm{is}\mathrm{the}\mathrm{objective}\mathrm{function};\mathrm{subject}\mathrm{to}\text{:}\stackrel{.fwdarw.}{x}\left\{R_u,{\mathrm{nRB}}_u,{\mathrm{CQI}}_{\min ,u}\right\}0m<R_{u}15m1{\mathrm{nRB}}_{u}1001{\mathrm{CQI}}_{\min ,u}15\underset{u}{\overset{\mathrm{nUE}}{.Math.}}{P}_{\mathrm{ini},u}\left\{20\mathrm{dBm}\right\}& \left(18\right)\end{array}$

(93) The premise parameters of three MFs before and after training are shown in FIG. 10 and FIG. 11, respectively. The root mean square error (RMSE) curve of the ITPSC is shown in FIG. 12, which demonstrates that the RMSE converges to 0.9 dBm after 300 epochs.

(94) (D) Channel Quality Index Decision Controller (CQIDC) Unit:

(95) In the real radio channel environment, indoor small cell base station deployment will face co-channel interference of macro cell base station or neighboring small cell, resulting in performance degradation of indoor small cell base station. Therefore, the CQIDC unit in the IDCC device determines the best CQI in interference-free environment to meet the receiver performance of BLER0.1. Further, in interference environments, self-optimizing power system control unit (SOPC) keeps track of the measured SINR to self-optimize the transmit power, enabling the UE to meet the objective needs of the service's reliability and minimum transmit power.

(96) In the interference free environment, in order to determine the best CQI (CQI.sub.best,u) at the u.sub.th user's location of the indoor office, the following formula is used to estimate signal-to-noise-power ratio (SNR). It can be expressed as
SNR.sub.u=P.sub.r,u(W)/P.sub.N,u(W)(19)

(97) where the average received power P.sub.r,u at the u.sub.th user in the interference free environment is given as
P.sub.r,u=P.sub.ini,uL.sub.t+G.sub.tPL.sub.u+G.sub.rL.sub.r(20)

(98) where PL.sub.u denotes the measured average path loss between a femtocell and an UE in the given cell. The noise power P.sub.N,u can be calculated by (13). Then, the best CQI in the interference free environment is determined by the following rules:

(99) $\begin{array}{cc}{\mathrm{CQI}}_{\mathrm{best},u}=\{\begin{array}{cc}\mathrm{CQI\_i},& \mathrm{if}{\mathrm{SNR}}_{\mathrm{th}}\left(\mathrm{CQI\_i}\right){\mathrm{SNR}}_u<{\mathrm{SNR}}_{\mathrm{th}}\left(\mathrm{CQI\_i}+1\right),i=114\\ \mathrm{CQI\_}15,& \mathrm{if}{\mathrm{SNR}}_u{\mathrm{SNR}}_{\mathrm{th}}\left(\mathrm{CQI\_}15\right)\\ \mathrm{CQI\_}1,& \mathrm{if}{\mathrm{SNR}}_u<{\mathrm{SNR}}_{\mathrm{th}}\left(\mathrm{CQI\_}1\right)\end{array}& \left(21\right)\end{array}$

(100) The training data of the CQIDC is generated from the simulation results of the transceiver BLER, as shown in Table 3. Integrating Table 3 with equations (19), (20) and (21), the best CQI is calculated for different measured average path loss (30 dB70 dB), resource block (1100) and initial minimum transmit power (75 dBm20 dBm). FIG. 13 shows a set of CQIDC training data of the best CQI corresponding to different average path loss (30 dB70 dB), when the resource block is 50, the minimum transmit power is (40 dBm, 20 dBm, 0 dBm). It is observed that the best CQI will decrease as the path loss becomes larger, and at the same average path loss, the best CQI increases as the initial transmit power becomes larger.

(101) The function of the CQIDC is to determine the best CQI of the femtocell at the u.sub.th user's location of indoor office, which satisfies the receiver performance of BLER0.1 in the interference free environments. Each input uses three Gaussian MFs, which are defined as

(102) 0$\begin{array}{cc}{A}_{j,n}\left(x_{j,m}\right)=e^{\frac{-{\left(x_{j,m}-b_{j,n}\right)}^2}{2a_{j,n}^2}},\mathrm{for}n=1,2,3\mathrm{and}j=1,2,3& \left(22\right)\end{array}$

(103) where x.sub.j,m is the m.sub.th input and the premise parameters a.sub.j,n, b.sub.j,n pertaining to the node outputs are updated according to given training data and the steepest descent approach. The 27 fuzzy inference rules are constructed in Table 5. The output of CQIDC is the best CQI of the femtocell at the u.sub.th user's location of indoor office. An optimization problem of the best CQI of the ANFIS-CQIDC is formally formulated as follows:

(104) In the interference free environments, BLER 0.1,

(105) optimize CQI.sub.best,u =(x) at the u.sub.th user,

(106) (x) is the objective function;

(107) subject to :
x {PL.sub.u, P.sub.ini,u, nRB.sub.u}
30dBPL.sub.u70dB
75dBm P.sub.ini,u 20dBm
1nRB.sub.u100
CQI.sub.best,u {115 }(23)

(108) The premise parameters of three MFs before and after training are shown in FIG. 14 and FIG. 15, respectively. The root mean square error (RMSE) curve of the CQIDC is shown in FIG. 16, which demonstrates that the RMSE converges to 0.46 after 300 epochs.

(109) (E) Self-Optimizing Power Control (SOPC) Unit:

(110) The SOPC consists of three parts, namely the transmit power adjustment estimator (TPAE), transmission power assignment and self-optimization power controller protection mechanism. The TPAE in the interference environment primarily estimates the amount of minimum transmit power adjustment needs for each user; the transmission power for each user is adjusted when the sum of total transmission power to all indoor users doesn't exceed the maximum transmit power limit of the eNB. The protection mechanism of the SOPC is used to prevent the co-channel interference from the moving users of adjacent cells.

(111) The ANFIS based TPAE of the SOPC unit adapts output power adjustment value P.sub.u at the u.sub.th user's location to the changing cell edge CQI (CQI.sub.min,u) set by user demand, the best CQI (CQI.sub.best,u) and measured average SINR (SINR.sub.u), so that the femtocell in the interference environment not only can still meet the demands set by the UE and communication quality, but also self-optimizing the minimum transmit power of eNB, thereby reducing the co-channel interference to the neighboring cells.

(112) For the purpose of satisfying the requirements of BLER10% and the SR of 90%, the threshold of the signal to interference plus noise ratio (SINR.sub.th,u) at the) u.sub.th user is defined as
SINR.sub.th,u=max{SNR.sub.th(CQI.sub.min,u)+FM(SR.sub.u),SNR.sub.th(CQI.sub.best,u)}(dB)(24)

(113) The output power adjustment (P.sub.u) at the u.sub.th user is given as
P.sub.u=SINR.sub.th,uSINR.sub.u(dB)(25)

(114) where SINR.sub.u is the measured average SINR at the u.sub.th user.

(115) The training data of ANFIS based TPAE of the SOPC unit is generated from the simulation results of the single input single output (SISO) transceiver BLER, as shown in Table 3. The fundamental specification of the SISO transceiver is listed in Table 2. Integrating Table 2, Table 3 with equations (24) and (25), the adjustment value of the minimum transmit power is calculated for service reliability (90%), cell edge CQI (115), measured average SINR (25 dB45 dB) and the best CQI (115). FIG. 17 shows a set of SOPC training data for the transmit power adjustment estimator corresponding to the different measured average SINR when the service reliability is 90%, cell edge CQI (CQI.sub.min,u) is 6 for two different best CQI (CQI.sub.best,u)(7, 10). It is observed that the transmit power adjustment will decreases as the measured average SINR increases in order to achieve the aim of minimum transmit power. For example, the SNR threshold for CQI.sub.best,u=7 is 13 dB. According to (20), the SINR threshold (SINR.sub.th,u) is equal to 14.14 dB, which is the sum of SNR threshold (12 dB) corresponding to CQI.sub.min,u=6 and FM(SR.sub.u)(2.14 dB) corresponding to SR.sub.u=90%. The SNR threshold for CQI.sub.bes,u=10 is 22 dB, which is greater than 14.14 dB. Thus, the SINR threshold (SINR.sub.th,u) is equal to 22 dB for CQI.sub.best,u=10. FIG. 13 shows P.sub.u=0 dB when the measured average SINR is 14.14 dB and 22 dB corresponding to CQI.sub.best,u=7 and 10, respectively.

(116) The function of ANFIS based TPAE of the SOPC unit is to determine the minimum transmit power of the femtocell eNB, which satisfies the receiver performance of BLER0.1 in the interference environments. Each input uses three generalized bell shape MFs, which are defined in (1). The 27 fuzzy inference rules are constructed in Table 6. Optimization problem of the minimum transmit power of the ANFIS based TPAE in the SOPC unit is formally formulated as follows:

(117) $\begin{array}{cc}\mathrm{In}\mathrm{the}\mathrm{interference}\mathrm{environment},\mathrm{BLER}0.1,\mathrm{optimize}{P}_u=f\left(\stackrel{.fwdarw.}{x}\right)\mathrm{at}\mathrm{the}{u}_{\mathrm{th}}\mathrm{user},f\left(\stackrel{.fwdarw.}{x}\right)\mathrm{is}\mathrm{the}\mathrm{objective}\mathrm{function}\text{:}\stackrel{.fwdarw.}{x}\left\{{\mathrm{CQI}}_{\min ,u},{\mathrm{CQI}}_{\mathrm{best},u}\mathrm{and}{\stackrel{\_}{\mathrm{SINR}}}_u\right\}1{\mathrm{CQI}}_{\min ,u}151{\mathrm{CQI}}_{\mathrm{best},u}15-25\mathrm{dB}{\stackrel{\_}{\mathrm{SINR}}}_{u}45\mathrm{dB}\underset{u}{\overset{\mathrm{nUE}}{.Math.}}{P}_u+P_u\left\{20\mathrm{dBm}\right\},P_u\mathrm{is}\mathrm{the}\mathrm{last}\mathrm{transmit}\mathrm{power}\left(\mathrm{dBm}\right)& \left(26\right)\end{array}$

(118) The premise parameters of three MFs before and after training are shown in FIG. 18 and FIG. 19, respectively. The root mean square error (RMSE) curve of the TPAE is shown in FIG. 20, which demonstrates that the RMSE converges to 0.86 dB after 300 epochs.

(119) Power Assignment Algorithm

(120) For the purpose of controlling the sum of individual transmission power to all indoor users less than the maximum transmit power limit of the indoor small cell eNB, a power assignment algorithm is proposed in FIG. 21. First, the previous transmit power P.sub.u(m1) of the u.sub.th user is compensated with the amount P.sub.u estimated by TPAE, and the sum of individual transmission power to all indoor users is calculated. Then the total transmitting power P.sub.total(m) at the m.sub.th time instant is compared with the maximum transmission power P.sub.max of the indoor small cell eNB. If the total transmitting power P.sub.total(m) at the m.sub.th time instant is less than P.sub.max, the power P.sub.u(m) of the u.sub.th user is assigned for each user at the m.sub.th time instant.

(121) If the total transmitting power P.sub.total(m) at the m.sub.th time instant is greater than P.sub.max, then the feedback loop of the step power adjustment is performed, wherein the previous transmit power P.sub.u(m1) of the u.sub.t, user at the (m1).sub.th time instant is temporarily stored in P.sub.tmp,u(i) for u=1, 2 . . . nUE. In the feedback loop of the step power adjustment, P.sub.tmp,u(i) will be decreased by when P.sub.u>0; P.sub.tmp,u(i) will be increased by when P.sub.u<0 and is assumed to be 0.1 dB. After each step power adjustment loop, the sum of P.sub.tmp,u(i+1) at the (i+1).sub.th loop for all users is compared the maximum transmission power P.sub.max of the indoor small cell eNB. If the total transmitting power P.sub.total(i+1) at the (i+1).sub.th loop is greater than P.sub.max, the transmit power P.sub.u(m) is equal to P.sub.tmp,u(i) at the i.sub.th loop and assigned to the u.sub.th user at the m.sub.th time instant. If the total transmitting power P.sub.total(i+1) at the (i+1).sub.th loop is less than P.sub.max, i is increased by one and feedback to the loop of the step power adjustment.

(122) (F) Protection Mechanism of the SOPC:

(123) The protection mechanism of the SOPC is included in the IDCC device to prevent the co-channel interference from the moving users of adjacent cells. The SODCC device inputs the average path loss measured from the UE, and then by the indoor path loss model of (16) to estimate the distance (d) between the UE and the eNB (base station). If the moving UE is inside the coverage range of the radius (R), then the SOPC unit is initiated to transmit the minimum power to the moving UE of the adjacent cells. Otherwise, he SOPC unit is disabled to the moving UE of the adjacent cells.

(124) (G) Experimental Results:

(125) FIG. 22 shows the simulation procedure to verify the service reliability. In the experiment, the SINR are measured at uniformly distributed UE positions in an indoor office with coverage radius R. The SINR are measured 1000 times at each UE position. The total number of measurements is defined as:

(126) $\begin{array}{cc}{m}_R=\underset{r=1}{\overset{R}{.Math.}}7r1000& \left(27\right)\end{array}$

(127) On the circumference of radius r=1 meter, the SINRs are measured at 7 uniformly distributed positions; on the circumference of the radius r=2, 3, 4, 5 meters, the corresponding uniformly distributed positions are 14, 21, 28, 35, respectively. When the coverage range of indoor office is set as 5 meters, the total number of positions to measure the SINR in an indoor office is 105. The total number of measurement positions increases with the coverage range of femtocell in the indoor office.

(128) The complementary cumulative distribution function (CCDF) of the measured SINR can be expressed as
F(SINR.sub.th)=P(measured SINR>SINR.sub.th)(28)

(129) The CCDF has the same meaning with the system reliability, which is defined as the percentage of the UE locations within a eNB coverage area of radius R for which the measured SINR exceeds a given SINR.sub.th. FIG. 18 shows the CCDF of the measured SINR for coverage radius of 5 meter, (90%) and three different cell edge CQIs 3, 7 and 10 corresponding to the minimum throughput requirements of three users UE1, UE2 and UE3, 2.76 Mbps, 7.44 Mbps and 7.44 Mbps, respectively, in the interference free environments.

(130) The SINR service reliabilities of the SOPC for coverage radius of 5 meter, service reliability 90% and different cell edge CQI CQI.sub.min=3, 7, 10 in the interference environments are also verified with FIG. 23(a), (b), (c) for interference power-100 dBm, 90 dBm and 80 dBm, respectively, where the solid curves denote for combining ITPSC and SOPC and the dot curves denote for using the ITSPC only. By means of Table 3, it can be found that the SINR.sub.th values corresponding to cell edge CQI 3, 7, and 10 are 5 dB, 13 dB, and 22 dB, respectively. The throughput service reliabilities for coverage radius of 5 meter and different cell edge CQI CQI.sub.min=3, 7, 10 in the interference environments are also verified with FIG. 24(a), (b), (c) for interference power 100 dBm, 90 dBm and 80 dBm, respectively, where the solid curves denote for combining ITPSC and SOPC and the dot curves denote for using the ITSPC only. Then (7) is used to calculate the throughputs of UE1, UE2 and UE3, which are 2.76 Mbps, 7.44 Mbps and 14.13 Mbps corresponding to cell edge CQI 3, 7, and 10, respectively. For example, as shown in FIGS. 23(a), (c) and 24(a), (c), both the CCDF of the measured SINR in 22 dB and the CCDF of the throughput in 21.4 Mbps are about 91% when the interference power is small (100 dBm) for both cases of combining ITPSC and SOPC method and using the ITPSC only. Increasing the interference power to 80 dBm, for example, when using the ITPSC only, the CCDF of the measured SINR value in 22 dB and the CCDF of the throughput in 14.13 Mbps for UE3 is dropped to about 15%, while the SOPC is still able to maintain about 89%. With the increase of interference power, it is observed that using the ITPSC only is unable to maintain the service reliability requirements set by the user, but using the SOPC is still able to maintain the requested service reliability.

(131) FIG. 25 and FIG. 26 show average throughput and average transmit power, respectively, for comparing the IDCC with fixed transmit power in the different interference environments. FIG. 25 shows using the fixed transmit power the average throughputs of UE1, UE2, UE3 and total average throughput for 100 dBm interference power are 10.86 Mbps, 17.38 Mbps, 24.08 Mbps and 52.45 Mbps, respectively. When the interference power increases to 80 dBm, the average throughputs of UE1, UE2 UE3 and total average throughput reduce to 2.97 Mbps, 6.93 Mbps, 12.41 and 22.32 Mbps, respectively. Using the present IDCC, the average throughputs of UE1, UE2, UE3 and total average throughput for 100 dBm interference power are 10.48 Mbps, 17.38 Mbps, 24.59 Mbps and 52.45 Mbps, respectively. When the interference power increases to 80 dBm, the average throughputs of UE1, UE2 UE3 and total average throughput reduce to 11.48 Mbps, 17.31 Mbps, 22.97 Mbps and 51.75 Mbps, respectively. The average throughputs of UE1, UE2, UE3 and total average throughput almost remain the same values when the IDCC is used. FIG. 26 shows using the present IDCC, the average transmit power increases as the interference power increases. Using the present IDCC, the average transmit power of UE1, UE2, UE3 and total average transmit power for 100 dBm interference power are 39.88 dBm, 29.85 dBm, 22.36 dBm and 21.58 dBm, respectively. When the interference power increases to 80 dBm, the average transmit power of UE1, UE2 UE3 and total average transmit power increase to 24.06 dBm, 13.56 dBm, 5.896 dBm and 5.154 dBm, respectively. The average transmit power of UE1, UE2, UE3 and total average transmit power adjust with the interference environments adaptively to satisfy the requirements of the service reliability when the IDCC is used.

(132) Thus the simulation results show that the present FFD-OFDMA based IDCC device for indoor small cell operated in the MU and interference environments to self-optimize the service reliability, throughput at the cell edge, minimum transmit power and interference for multimedia call services. Thus the IDCC device can achieve the goals of saving power consumption and reducing co-channel interference. In this embodiment of the simulation, the basic OFDM transceiver parameters listed in Table 2 is a single antenna mode (SISO), the present invention is also applicable to multi-antenna mode (MIMO) and other different channel environments.

(133) While the preferred embodiment of the invention has been set forth for the purpose of disclosure, modifications of the disclosed embodiment of the invention as well as other embodiments thereof may occur to those skilled in the art. Accordingly, the appended claims are intended to cover all embodiments which do not depart from the spirit and scope of the invention.