CN111934839B - An Interference Mitigation and Resource Allocation Method for Underwater Acoustic Soft Frequency Multiplexing Network - Google Patents

Abstract

The invention discloses an interference mitigation and resource allocation method for an underwater acoustic soft frequency reuse network, comprising the following contents: Step 1. In a multi-cell underwater acoustic network, a control node calculates the time delay of its received signal i and interference signal j difference, and design its own data packet length according to the delay difference and the location information of multiple data nodes; step 2, deduce the SINR formula of the center node and edge node; step 3, calculate the average coverage probability, and according to the average coverage probability and water The total bandwidth of the acoustic network system is allocated to the edge area and the central area; step 4, calculate the SINR, construct a linear finite state Markov chain prediction equation, and predict the CSI with propagation delay; step 5, according to the predicted CSI. Adaptive resource allocation for data nodes. The inter-cell interference of the time-varying underwater multi-cell network in the prior art is alleviated, and the inaccuracy of the long-delay feedback CSI used for optimizing the resource allocation within the cell in the prior art is alleviated.

Description

An Interference Mitigation and Resource Allocation Method for Underwater Acoustic Soft Frequency Multiplexing Network

technical field

The invention belongs to the technical field of frequency multiplexing of underwater acoustic networks, and in particular relates to an interference mitigation and resource allocation method for an underwater acoustic soft frequency multiplexing network.

Background technique

In the underwater acoustic network, the single-hop network is easy to implement and control, and can complete the data acquisition task with a small coverage area and a small number of nodes. With the development of research and application of underwater acoustic network, more and more large-scale underwater acoustic network is applied to the exploration practice of ocean. For example, multiple buoys are deployed on the sea surface as control nodes (including underwater acoustic communication equipment), and the control nodes communicate through wireless links on the buoys, and each control node communicates with multiple data nodes (sensor nodes or Vehicle nodes) for underwater acoustic communication to form an integrated communication network at sea and under the sea to complete data tasks with wider coverage and more complex tasks.

In an underwater acoustic network with multiple control nodes as the center of multiple cells, the multiplexing of multiple cells can be realized by means of frequency reuse. In the existing frequency reuse scheme, although FFR (Fractional frequency reuse) can effectively alleviate the interference of edge nodes between cells, the system spectral efficiency is not high; while SFR can not only improve the system spectral efficiency, but also better alleviate the Cell edge nodes are affected by interference. However, how to better alleviate the inter-cell interference problem of the time-varying underwater acoustic network needs further research.

In the multi-cell of the underwater acoustic network, multiple data nodes share the allocated frequency band, and adaptive OFDMA allocates carriers and bits to the data nodes according to the channel state information (CSI), which can further improve the throughput of the whole system. Adaptive OFDMA allocates resources according to the feedback CSI. However, the time-varying characteristics of the underwater acoustic channel make the CSI obtained by the transmitter from the receiver through the feedback link often delayed. Only by predicting the state, can more accurate resource allocation be performed according to CSI. However, how to overcome the inaccuracy of CSI with long-delay propagation feedback requires further research.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide an interference mitigation and resource allocation method for an underwater acoustic soft frequency reuse network, so as to alleviate the inter-cell interference of the time-varying underwater multi-cell network in the prior art, and to optimize the intra-cell resources in the prior art The long-delay feedback CSI used for allocation is inaccurate.

The present invention adopts the following technical solutions: an interference mitigation and resource allocation method for an underwater acoustic soft frequency reuse network, comprising the following contents:

Step 1. In a multi-cell underwater acoustic network containing a single control node and multiple data nodes, the multiple data nodes are divided into central nodes and edge nodes, and the control node is based on the location information of the multiple data nodes. to calculate the time delay difference between the received signal i and the interference signal j, and design its own data packet length according to the time delay difference and the location information of multiple data nodes, the data packet length makes the central node The interference factor β _i =0, the interference factor of the edge node satisfies 0<β _i ≤1;

Step 2, deduce the SINR formula of the central node and the edge node according to the SINR formula of the underwater acoustic network data node, the interference factor of the central node and the interference factor of the edge node;

Step 3, calculate the average coverage probability according to the SINR formula obtained in step 2 and the preset coverage probability threshold, and perform frequency band allocation in the edge area and the central area according to the average coverage probability and the total bandwidth of the underwater acoustic network system;

Step 4, according to the result of the frequency band allocation obtained in the step 3, calculate the SINR as the feedback CSI, construct a linear finite state Markov chain prediction equation, and use the soft mean algorithm to predict the CSI with propagation delay;

Step 5: Perform adaptive resource allocation for data nodes according to the CSI predicted in step 4.

Further, in the step 1, in a cell containing a single control node and multiple data nodes, the area less than the optimal distance threshold is divided into a central area, otherwise it is an edge area, and the data node in the central area is the central node. , the data nodes in the edge area are edge nodes;

The time delay difference τ _i between the received signal i and the interference signal j is:

In formula (1), t _j,send and t _i,send are the time when the control node of the interfering cell and the control node of the target cell send signals respectively, (x _i ,y _i ,z _i ), (x _j ,y _j ,z _j ) are the position coordinates of the target cell data node i and the interfering cell control node j, respectively, d _oi is the distance between the center of the target cell and the data node in the cell, and d _ij is the distance between the target cell data node and the interfering cell control node The distance of , v=1500m/s is expressed as the propagation speed of the acoustic signal in seawater.

Further, in the step 2, the SINR formula of the central node and the edge node is:

In formula (2), P _i represents the effective power obtained by the data node i from the control node of the target cell, P _j represents the interference power obtained by the data node i from the control node of the interfering cell, A _i (l, f), A _j (l , f) are the transmission attenuation of the control node and the interference node at node i respectively, Δf represents the frequency bandwidth of each carrier in the underwater acoustic channel, I _j is the received interference from adjacent cells at the same frequency, α is the cell edge The transmission power ratio between the node and the central node, l _i and l _j are respectively the distance between the control node of the target cell and the control node of the interfering cell and the data node i, k is the expansion factor, and k=1 in the case of cylindrical propagation, spherical surface k=2 during propagation, h _i ( _li ,f) and h _j (l _j ,f) are the channel gains obtained by data node _i from the control node of the target cell and the control node of the interfering cell, respectively, (li ) ^-k ( a(f)) ^-li and (l _j ) ^-k (a(f)) ^-lj are the transmission attenuation of the control node of the target cell and the control node of the interfering cell at the data node i respectively, the power spectral density of the ocean noise N( f) Calculated according to the Wenz model, the seawater absorption loss coefficient a(f) is calculated according to the Thorp formula:

为数据节点的瞬时SINR大于T_FR的概率：Further, in the step 3, the average coverage probability under different frequencies is calculated according to the SINR and the preset coverage probability threshold T _FR , and the coverage probability

is the probability that the instantaneous SINR of the data node is greater than _TFR :

分别表示为数据节点i与干扰j₁、j₂之间的距离；In Equation (3), T _FR can be reasonably designed according to its influence on the frequency bandwidth of the edge region, or obtained through the inversion of the complementary cumulative distribution function based on the data node load, Ej[ ] is the random frequency in the underwater multi-cell network. The expectation of interfering cell j, j ₁ , j ₂ are expressed as the interference transmitted to the center node and edge node of the cell, respectively,

are respectively expressed as the distance between data node i and interference j ₁ , j ₂ ;

进行边缘区域和中心区域的频带分配；According to the total system bandwidth B _total , the average coverage probability

Carry out frequency band allocation in the fringe area and the center area;

In formula (4), B _total is the total bandwidth of the system, B _edge and B _int are the frequency bandwidths of the cell edge area and central area, respectively; for broadband systems, the average coverage probability used for frequency band allocation is further taken as the average of multiple carrier frequencies.

Further, the specific method of step 4 is,

First, convert CSIγ into a finite number of channel states C(m), C(m) ∈ [0, S-1], where m is time and S is the number of channel states; CSI is divided into finite numbers according to the v _s threshold For the discrete values of the FSMC state, adopt the equal probability method to select v _s , so that the stationary probability π _s of each FSMC state is equal and 1/S;

In formula (5), σ ² is the Rayleigh fading channel gain variance; set v ₀ =0, v _S+1 =∞, find the value of each threshold v _s (s=1,2,...,S), CSI is divided into [0,v ₁ ),[v ₁ ,v ₂ ),...,[v _s ,∞); when CSI falls in the interval [v _s ,v _s+1 ], define C(m)=s;

Next, solve the linear correlation coefficient ψ _l ; map the training sequence T(m) to different state regions by quantizing the state label (C(m)=s), and then use the Yule-Walker equation to calculate T(m) ψ _l ; update and record ψ _l using equation (6):

ψ _l (m,s)=ψ _l (m,C(m+1)=s|C(m),...,C(m-L+1)) (6),

In formula (6), ψ _l (m,s) represents the channel state from C(m-L+1), C(m-L+2),...,C(m) to C(m+1)=s When , the lth linear correlation coefficient; use multiple sets of T(m) to obtain multiple temporary linear coefficients, and then average them;

Again, a state transition probability matrix P(m) is established; where the elements of p _q,w (m) represent the probability of CSI transitioning from state q to state w, p _q,w (m)=Pr(C(m)= q|C(m-1)=w), the dimension is S×S; in a small-scale fading channel, it is assumed that from time m-1 to m, the FSMC state only occurs between the current or adjacent states, in the state of The CSI of q can be transferred to the adjacent state (q-1)/(q+1), or remain in the original q state;

p _q,w (m) is approximately expressed as:

Define p _0,0 , p _S-1,S-1 ,

In formula (8), T _s is the symbol period, and f _d is the maximum Doppler frequency shift;

Finally, according to the feedback delay CSI, the soft average algorithm is used to substitute ψ _l (m, s) and P (m) into formula (10) to predict the CSI of m+1 after the delay t _m time slot;

In formula (9), p(s)=P(C(m+1)=s|C(m), C(m-1), C(m-L+1)) is expressed as H(m+1) ) is the conditional probability of maintaining the C(m+1) state.

Further, the specific method of step 5 is:

First, according to the predicted CSI obtained in step 4, the proportional fairness algorithm is used to consider the instantaneous transmission rate and average transmission rate of the data node to determine the priority of the subcarrier used by the data node at the data scheduling moment, and assign the subcarrier resources according to the priority. Allocated to each data node;

Secondly, the adaptive modulation Chow algorithm is used to optimize the number of bits loaded on each data node subcarrier, and under the constraints of the system target bit error rate BER and total power, the channel capacity on each subcarrier is maximized;

其中n为子载波数，Γ＝-ln(5×BER)/1.6。Finally, the loading power is calculated according to the bit allocation result bit _n

where n is the number of subcarriers, Γ=-ln(5×BER)/1.6.

The beneficial effects of the present invention are as follows: an interference mitigation and resource allocation (T-SFR) method for underwater acoustic soft frequency reuse network based on time delay difference is disclosed. Expand on two fronts. Between cells, the T-SFR scheme uses the time delay difference between different channels to mitigate interference, and allocates frequency bands based on the coverage probability of the underwater acoustic channel. The verification shows that the proposed T-SFR scheme has better performance than the traditional SFR scheme. Higher SINR can alleviate inter-cell interference, and at the same time have higher spectral efficiency. In the cell, in view of the influence of the feedback channel state information (CSI) delay brought by the time-varying characteristics of the underwater acoustic channel, the present invention combines the linear function and the Markov chain model to construct a Linear Finite State Markov Chain (LFSMC) predictor , to predict more accurate CSI for adaptive resource allocation and further improve system throughput. Through inter-cell interference mitigation and intra-cell adaptive resource allocation, the system throughput of the optimized underwater acoustic soft frequency reuse network is better than that of the T-SFR system without LFSMC predictor and the throughput of SFR system without LFSMC predictor quantity.

Description of drawings

Fig. 1 is the application scene diagram of underwater acoustic multi-cell network;

Fig. 2 is the relationship between the position of the cell data node and the system capacity under different frequency reuse factors in the embodiment of the present invention;

Fig. 3 is the geometric model diagram of underwater acoustic SFR network of the present invention;

4 is an analysis diagram of the interference situation of each area data node in cell 1 in the embodiment of the present invention;

Fig. 5 is the SINR performance comparison diagram of edge nodes of different multiplexing schemes under the frequency change in the embodiment of the present invention;

Fig. 6 is the SINR performance comparison diagram of the edge node under different frequencies in the embodiment of the present invention;

7 is a diagram showing the relationship between cell radius and system spectral efficiency under different frequency reuse schemes in an embodiment of the present invention;

8 is a graph showing the relationship between the average coverage probability and the coverage probability signal-to-interference-noise ratio threshold at different frequencies in an embodiment of the present invention;

9 is a comparison diagram of sub-band widths of edge nodes under different interference factors in an embodiment of the present invention;

FIG. 10 is a system bit error rate performance diagram under a fixed modulation mode in an embodiment of the present invention;

Fig. 11 is a fixed adaptive modulation system (AM), a linear adaptive modulation system (LS-AM), an adaptive modulation system based on Markov chain prediction (MC-AM), and a prediction based on an LFSMC predictor in an embodiment of the present invention The throughput performance comparison chart of the four adaptive modes of the adaptive modulation system of (LFSMC-AM);

FIG. 12 is a total throughput performance diagram of the entire network system under different frequency reuse schemes in an embodiment of the present invention;

FIG. 13 is a method flowchart of an interference mitigation and resource allocation method for an underwater acoustic soft frequency reuse network according to the present invention.

Detailed ways

The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

The present invention provides an interference mitigation and resource allocation method for an underwater acoustic soft frequency multiplexing network, as shown in FIG. 13 , including the following contents:

Step 1. As shown in the schematic diagram of the underwater acoustic multi-cell network in Fig. 1, in a cell with a single control node and multiple data nodes, through system simulation, when the frequency reuse factor is 1 and the capacity of the two schemes is equal to 3, we get The optimal distance threshold, the area smaller than the optimal distance threshold is divided into the central area, otherwise it is the edge area, correspondingly, the data nodes are divided into central nodes and edge nodes.

A single control node calculates the delay difference between the received signal i and the interference signal j according to the position information of all data nodes in the cell, and designs the data packet length according to the delay difference and the position of the data node, so that the interference factor of the central node β _i =0, The interference factor of the edge node is 0 < β _i ≤ 1;

Step 2, deduce the SINR formula of the center node and the edge node according to the existing underwater acoustic network data node SINR formula, and the interference factor of the above-mentioned center node and edge node;

Step 3, calculate the average coverage probability under different frequencies according to the SINR and the preset coverage probability threshold, and perform frequency band allocation in the edge area and the central area according to the obtained average coverage probability and the total system bandwidth;

Step 4. According to the result of the above frequency band allocation, calculate the SINR as the feedback CSI, construct the linear finite state Markov chain prediction equation under the existing underwater acoustic fading channel model, and use the soft mean algorithm to predict the CSI with propagation delay. ;

Step 5: Perform adaptive resource allocation for data nodes according to the predicted CSI.

Wherein, in the step 1, in a cell containing a single control node and multiple data nodes, through system simulation, the optimal distance threshold is obtained when the frequency reuse factor is 1 and the capacity of the two schemes is equal to 3, The area smaller than the optimal distance threshold is divided into the central area, otherwise it is the edge area, and correspondingly, the data nodes are divided into central nodes and edge nodes.

The control node calculates the time delay difference τ _i between the received signal i and the interference signal j according to the location information of all data nodes:

In formula (1), t _j,send and t _i,send are the time when the control node of the interfering cell and the control node of the target cell send signals respectively, (x _i ,y _i ,z _i ), (x _j ,y _j ,z _j ) are the position coordinates of the target cell data node i and the interfering cell control node j, respectively, d _oi is the distance between the center of the target cell and the data node in the cell, and d _ij is the distance between the target cell data node and the interfering cell control node The distance of , v=1500m/s is expressed as the propagation speed of the acoustic signal in seawater.

The design data packet length is T _p , and the interference factor β _i of the data node is defined as the degree to which a data packet is affected by interference, expressed as a percentage;

当τ_i,j＜T_p时，0＜β_i≤1。For the central node, if τ _i ≥ T _p , the reception of the data packet can be completed before the interference comes, at this time β _i =0; for the edge node, the duration of the data packet affected by the interference when it is received is T _p − τ _i , then

When τ _i,j <T _p , 0<β _i ≤1.

In the described step 2, according to the existing underwater acoustic network data node SINR formula, and the interference factor of the above-mentioned center node and edge node, derive the SINR formula of the center node and the edge node;

分别为目标小区控制节点和干扰小区控制节点在数据节点i处的传输衰减，海洋噪声的功率谱密度N(f)根据Wenz模型计算，海水吸收损失系数a(f)根据Thorp公式计算：

In formula (2), P _i represents the effective power obtained by the data node i from the control node of the target cell, P _j represents the interference power obtained by the data node i from the control node of the interfering cell, A _i (l, f), A _j (l , f) are the transmission attenuation of the control node and the interference node at node i respectively, Δf represents the frequency bandwidth of each carrier in the underwater acoustic channel, I _j is the received interference from adjacent cells at the same frequency, α is the cell edge The transmission power ratio between the node and the central node, l _i and l _j are respectively the distance between the control node of the target cell and the control node of the interfering cell and the data node i, k is the expansion factor, and k=1 in the case of cylindrical propagation, spherical surface k=2 during propagation, h _i ( _li ,f) and h _j (l _j ,f) are the channel gains obtained by data node i from the control node of the target cell and the control node of the interfering cell, respectively,

are the transmission attenuation of the target cell control node and the interference cell control node at the data node i respectively, the power spectral density N(f) of the ocean noise is calculated according to the Wenz model, and the seawater absorption loss coefficient a(f) is calculated according to the Thorp formula:

In step 3, the average coverage probability under different frequencies is calculated according to the SINR and the preset coverage probability threshold T _FR , and the coverage probability F (T _FR ) is the probability that the instantaneous SINR of the data node is greater than T _FR :

分别表示为数据节点i与干扰j₁、j₂之间的距离。In Equation (3), T _FR can be reasonably designed according to its influence on the frequency bandwidth of the edge region, or obtained through the inversion of the complementary cumulative distribution function based on the data node load, Ej[ ] is the random frequency in the underwater multi-cell network. The expectation of interfering cell j, j ₁ , j ₂ are expressed as the interference transmitted to the center node and edge node of the cell, respectively,

are expressed as the distance between the data node i and the interference j ₁ , j ₂ , respectively.

进行边缘区域和中心区域的频带分配；According to the total system bandwidth B _total , the average coverage probability

Carry out frequency band allocation in the fringe area and the center area;

In formula (4), B _total is the total bandwidth of the system, B _edge and B _int are the frequency bandwidths of the cell edge area and central area, respectively; for broadband systems, the average coverage probability used for frequency band allocation is further taken as the average of multiple carrier frequencies.

In step 4, the linear finite state Markov chain prediction equation under the existing underwater acoustic fading channel model is constructed, and under the discrete finite number of channel states, the linear correlation coefficient ψ _l and the channel transition probability P(m) are Enter a linear equation to predict the CSI at the moment of data transmission, so that the closer the predicted CSI is to the actual CSI, the better the effect. Specifically take the following steps:

First, convert CSIγ into a finite number of channel states C(m), C(m)∈[0, S-1], where m is time and S is the number of channel states. The CSI is divided into discrete values of a finite number of FSMC states according to the v _s threshold, and v _s is selected by the equal probability method, so that the stationary probability π _s of each FSMC state is equal and 1/S.

In Equation (5), σ ² is the Rayleigh fading channel gain variance. Set v ₀ =0, v _S+1 =∞, and find the value of each threshold v _s (s=1,2,...,S). Divide the CSI into [0,v ₁ ),[v ₁ ,v ₂ ),…,[v _s ,∞). When the CSI falls in the interval [v _s ,v _s+1 ], C(m)=s is defined.

Next, solve for the linear correlation coefficient ψ _l . The training sequence T(m) is mapped to different state regions using state labels (C(m)=s) and quantized, and T(m) is then calculated using the Yule-Walker equation to calculate ψ _l . Update and record ψ _l using equation (6):

ψ _l (m,s)=ψ _l (m,C(m+1)=s|C(m),...,C(m-L+1)) (6),

In formula (6), ψ _l (m,s) represents the channel state from C(m-L+1), C(m-L+2),...,C(m) to C(m+1)=s When , the lth linear correlation coefficient. Because the ψ _l obtained by different T(m) is different, multiple sets of T(m) are used to obtain temporary multiple linear coefficients, and then average them.

Again, the state transition probability matrix P(m) is established. The p _q,w (m) element is expressed as the probability of CSI transitioning from state q to state w, p _q,w (m)=Pr(C(m)=q|C(m-1)=w), The dimension is S×S. In a small-scale fading channel, assuming that from time m-1 to m, the FSMC state only occurs between the current or adjacent states, the CSI in state q can be transferred to the adjacent state (q-1)/(q+1 ), or remain in the original q state. Usually the Markov process is stationary and the state transition probability is independent of time m. In a Rayleigh fading channel, p _q,w (m) can be approximately expressed as:

Define p _0,0 , p _S-1,S-1

In formula (8), T _s is the symbol period, and f _d is the maximum Doppler frequency shift.

Finally, according to the feedback delay CSI, the soft average algorithm is used to substitute ψ _l (m, s) and P (m) into formula (10) to predict the CSI of m+1 after the delay t _m time slot;

In formula (9), p(s)=P(C(m+1)=s|C(m), C(m-1), C(m-L+1)) is expressed as H(m+1) ) is the conditional probability of maintaining the C(m+1) state.

In step 5, the following steps are taken for adaptive resource allocation:

First, according to the predicted CSI obtained in step 4, the proportional fairness algorithm is used to consider the instantaneous transmission rate and average transmission rate of the data node to determine the priority of the subcarrier used by the data node at the data scheduling moment, and assign the subcarrier resources according to the priority. Allocated to each data node;

其中n为子载波数，Γ＝-ln(5×BER)/1.6；Secondly, the adaptive modulation Chow algorithm is used to optimize the number of bits loaded on each data node sub-carrier, and under the constraints of the system target bit error rate BER and total power, the channel capacity on each sub-carrier is maximized. Finally, the loading power is calculated according to the bit allocation result bit _n

where n is the number of subcarriers, Γ=-ln(5×BER)/1.6;

The Chow algorithm is used to optimize the number of bits loaded on each data node sub-carrier. Data information is transmitted according to optimal bit allocation and power loading to obtain optimal throughput. By predicting the CSI during data transmission, the throughput of the system is improved.

Example

Table 1 Parameter settings

According to the parameters given in Table 1 above, the following 5 steps are used to verify the rationality of the embodiments of the present invention:

1. In a cell containing a single control node and multiple data nodes, as shown in the relationship between the geometrical position of the data node in the cell and the system capacity in Figure 2, the abscissa in the figure represents the distance between the data node and the control node. Assuming that the cell radius is R, the position of 0.72 indicates that the distance between the data node and the control node is 0.72R. The simulation results show that the two schemes of frequency reuse factor of 1 and 3 in the figure both decrease the system capacity as the distance between the data node and the control node increases. And it can be obtained that the reasonable division of the center of the cell and the edge area should take the radius ratio of 0.72 as the boundary point, and the area with a distance less than 0.72R as the center area, otherwise it is the edge area. Correspondingly, the data nodes are divided into center nodes and edge nodes.

According to this method, the 6 data nodes in each cell are divided into 4 central nodes and 2 edge nodes. The control node calculates the time delay difference τ _i between the received signal i and the interference signal j according to the location information of all data nodes in the cell, and designs the data packet length T _p according to the time delay difference τ _i and the location of the data nodes.

For the central node, if τ _i ≥ T _p , the data packet reception can be completed before interference comes. At this time, the central data node will not be interfered by adjacent cells, and β _i =0. Figure 3 shows the geometric model diagram of the underwater acoustic multi-cell network from a two-dimensional perspective. It can be seen that the data nodes of the cell are divided into central nodes and edge nodes. Correspondingly, the cells are divided into central and edge regions; and according to Figure 3 - Different power levels in (b) allocate corresponding frequency band resources to data nodes in different areas of each cell, and the power level of cell edge nodes is higher than that of the central node to reduce the impact of interference on edge nodes. Combining Figure 3 and Figure 4 to illustrate: when the central node i is in the five-pointed star position as shown in Figure 3-(a), the data node i will not only receive useful signals from the serving cell control node at do _oi /vs, It will also be interfered by neighboring cells at d _ij /vs; as shown in Figure 4-(a), when τ1 is greater than T _p , the data packet information of this node will not be interfered. When the central node i is in the circular position as shown in Figure 3-(a), as shown in Figure 4-(b), the τ ₂ of the data node i is equal to T _p , at this time, the data packets of this node just do not affected by interference.

因为边缘节点距离小区中心较远，节点在未完全接收到数据包时，就可能已经收到来自相邻小区的干扰，使τ_i,j＜T_p，所以0＜β_i≤1。结合图3和图4举例说明：当边缘节点i处于如图3-(a)中菱形位置时，此节点的τ₃小于T_p，对应图4-(c)，此类边缘节点的数据包信息会有一部分会受到干扰影响，此时0＜β_i＜1；当边缘节点i位于如图3-(a)的三角形位置时，如图4-(d)所示，该数据节点的τ₄小于T_p，甚至τ₄＝0，此时β_i＝1，该边缘节点会严重受到干扰影响。For edge nodes, the duration of the packet being affected by interference when it is received is T _p -τ _i , then

Because the edge node is far from the center of the cell, when the node does not fully receive the data packet, it may have received interference from the adjacent cell, so that τ _i,j <T _p , so 0<β _i ≤1. Combining Figure 3 and Figure 4 to illustrate: when the edge node i is in the diamond-shaped position as shown in Figure 3-(a), the τ ₃ of this node is less than T _p , corresponding to Figure 4-(c), the data packets of such edge nodes A part of the information will be affected by interference, at this time 0 < β _i <1; when the edge node i is located in the triangle position as shown in Figure 3-(a), as shown in Figure 4-(d), the τ of the data node ₄ is less than T _p , even if τ ₄ =0, at this time β _i =1, the edge node will be seriously affected by interference.

2. According to the existing underwater acoustic network data node SINR formula, and the above-mentioned interference factors of the center node and edge node, the SINR of the center node and the edge node is derived by formula (2). Figure 5 compares the SINR performance of edge nodes with different reuse schemes under frequency variation. The results show that under the theoretical and simulated channels, the SINR of the T-SFR scheme is the best, followed by SFR, and the last is FFR, and with the increase of f, the SINR presents a non-monotonic change, and has a certain frequency at a certain frequency. Therefore, the system can provide the optimal operating frequency for edge nodes to maximize SINR. Figure 6 analyzes the SINR performance of the edge node when the frequency changes under different interference factors β _i . It can be seen that with the increase of β _i , the SINR decreases nonlinearly; and when the operating frequency f is equal to the optimal operating frequency 11kHz When , the SINR performance of edge nodes is optimal. Figure 7 compares the spectral efficiency under different frequency reuse schemes. The simulation results show that the spectral efficiency has the best performance in a fixed cell radius, and the spectral efficiency of the three schemes has the following relationship: T-SFR>SFR>FFR.

3. According to the SINR of the above-mentioned edge nodes and the preset coverage probability threshold, use formula (3) to calculate the average coverage probability under different frequencies, and according to the obtained average coverage probability, the total system bandwidth B _total is in the range of 9kHz to 15kHz. The band allocation B _edge in the edge region is derived, and then the band allocation in the central region is obtained according to the known B _edge . Figure 8 simulates the average coverage probability at different frequencies. It can be seen that with the increase of the interference factor β _i and the SINR threshold, the coverage probability decreases. Figure 9 simulates the allocated bandwidth for edge regions at different SINR thresholds. The results show that with the increase of the SINR threshold and β _i , the frequency bandwidth to which the edge region is allocated is larger. In the underwater acoustic network, the rate requirements of edge nodes are met by determining β _i and an appropriate SINR threshold.

4. According to the frequency band allocation result, calculate the SINR as the feedback CSI required for adaptive resource allocation, and construct a linear finite state Markov chain prediction equation according to the existing underwater acoustic fading channel model. A few times, the soft average algorithm in formula (9) is used to predict the CSI at the moment of data transmission.

5. According to the predicted CSI, the 512 subcarrier resources are first allocated to the data nodes at the data scheduling time using the proportional fairness algorithm, then the adaptive modulation Chow algorithm is used to optimize the number of bits loaded on the subcarriers, and finally the loading power is calculated according to the bit allocation result. , and then complete the adaptive resource allocation of data nodes. Figure 10 shows the system bit error rate performance under four fixed modulation modes. From the figure, it can be concluded that under a certain bit error rate, there is a modulation mode in a certain SINR threshold interval that makes the number of loaded bits the most. For example, when the bit error rate BER is constrained to be 10 ^-3 , Table 2 shows the optimal modulation modes (corresponding to the number of bits) corresponding to different SINR threshold intervals for adaptive resource allocation.

Table 2 Modulation switching threshold

By predicting the CSI during actual data transmission, the throughput of the system can be improved. Figure 11 compares the system throughput performance of several adaptive schemes, specifically the following four schemes: Adaptive Modulation without Prediction (AM), Adaptive Modulation with Linear Prediction (LS-AM), and Adaptive Modulation with Markov Chain Prediction (MC-AM) and the proposed LFSMC Adaptive Modulation (LFSMC-AM). Due to insufficient consideration of the channel delay, the non-prediction AM based on direct feedback CSI has the worst performance; the throughput performance of MC-AM is better than that of LS-AM, which shows the excellent performance of the Markov chain model; The CSI prediction method combines the advantages of linear functions and Markov chain models, so the throughput performance of LFSMC-AM is optimal in several ways.

According to the parameters and methods described in the above steps, an adaptive resource allocation simulation experiment was carried out on the time-varying underwater acoustic network. Taking the underwater acoustic network of 7 cells as an example, the FFR, SFR, T The throughput performance of the three frequency reuse schemes of -SFR, the performance of each scheme is shown in Figure 12. It can be seen from the figure that the T-SFR-LFSMC scheme is better than other schemes in throughput performance in the central area and the edge area. Using higher transmit power to alleviate interference, and using the T-SFR time delay difference idea to reduce the impact of partial interference in adjacent cells, on the one hand, the LFSMC predictor is used to obtain more accurate CSI, which improves the adaptive resource allocation in the cell performance. Since the total throughput is the sum of the throughput of the central area and the edge area, the total throughput performance under the T-SFR-LFSMC scheme is the best. The T-SFR system with LFSMC predictor is better than the T-SFR system without predictor. The throughput is improved by 6.2%, which is 35% higher than the SFR system without predictor.

The present invention specifically analyzes the effect from the following two aspects: between cells, the T-SFR scheme utilizes the time delay difference between different channels for interference mitigation, and performs frequency band allocation based on the coverage probability of the underwater acoustic channel. It can maintain good spectral efficiency while alleviating inter-cell interference. Through the proposal of the T-SFR scheme, the SINR is increased by 0.9dB compared with the traditional SFR, and the spectral efficiency is improved by about 10% on average compared with the traditional SFR. In the cell, in view of the influence of the feedback channel state information (CSI) delay brought by the time-varying characteristics of the underwater acoustic channel, the present invention combines the linear function and the Markov chain model to construct a Linear Finite State Markov Chain (LFSMC) predictor , to predict more accurate CSI for adaptive resource allocation and further improve system throughput. With the implementation of the present invention, through the interference mitigation between cells and the adaptive resource allocation within the cells, the system throughput of the optimized underwater acoustic soft frequency reuse network is significantly improved. The example simulation results show that the T of the LFSMC predictor is The -SFR system has a 6.2% improvement in throughput over the T-SFR system without a predictor and a 35% improvement in throughput over the SFR system without a predictor.

Claims (4)

1. An interference mitigation and resource allocation method for an underwater acoustic soft frequency reuse network is characterized by comprising the following contents:

step 1, in a multi-cell underwater acoustic network containing a single control node and a plurality of data nodes, dividing the plurality of data nodes into a central node and an edge node, calculating the time delay difference of a received signal i and an interference signal j by the control node according to the position information of the plurality of data nodes, and designing the length of a data packet of the control node according to the time delay difference and the position information of the plurality of data nodes, wherein the length of the data packet enables the interference factor beta of the central node to be larger than the length of the data packet _i 0, the interference factor of the edge node satisfies 0 < beta _i ≤1；

Wherein, for the central node, let τ _i ≥T _p The reception of the data packet can be completed before the interference comes, in this case beta _i 0; for the edge node, the data packet is affected by interference when received for a duration T _p -τ _i Then, then

When tau is _i ＜T _p When 0 < beta _i Less than or equal to 1; the data packet length is T _p The control node calculates the time delay difference tau of the received signal i and the interference signal j according to the position information of all the data nodes _i ；

Step 2, deducing SINR formulas of the central node and the edge nodes according to an SINR formula of a data node of the underwater acoustic network, the interference factor of the central node and the interference factor of the edge nodes;

step 3, calculating an average coverage probability according to the SINR formula obtained in the step 2 and a preset coverage probability threshold, and performing frequency band allocation of an edge area and a central area according to the average coverage probability and the total bandwidth of the underwater acoustic network system;

in the step 3, according to the SINR and a preset coverage probability threshold T _FR Calculating the average coverage probability at different frequencies

Instantaneous SINR for data nodes greater than T _FR Probability of (c):

in the formula (3), T _FR Can be reasonably designed according to its influence on the bandwidth of the edge area, or obtained by inverting a complementary cumulative distribution function based on the data node load, Ej [ ·]Expectation of random interference cell j in underwater multi-cell network ₁ 、j ₂ Expressed as interference transmitted to cell centre nodes and edge nodes, respectively,/ _j1 、l _j2 Denoted as data node i and interference j, respectively ₁ 、j ₂ The distance between them; alpha is the transmission power ratio of the cell edge node to the central node; beta is a _i An interference factor for the received signal; p _i Representing the effective power obtained by the data node i from the target cell control node; p _j Representing the interference power obtained by the data node i from the interference cell control node; a. the _i (l, f) controlling the transmission attenuation of the node at the node i for the target cell, and particularly expanding to

l _i Expressed as the distance between the target cell control node and the data node i; k is a spreading factor; n (f) is the power spectral density of the ocean noise; a (f) is the seawater absorption loss coefficient;

according to the total bandwidth B of the system _total Average probability of coverage

Performing frequency band allocation of the edge region and the central region;

in the formula (4), B _total Is the total bandwidth of the system, B _edge And B _int Respectively the frequency bandwidth of the edge area and the central area of the cell; for a broadband system, the average coverage probability used by frequency band allocation is further taken as the average value of multiple carrier frequencies;

step 4, calculating SINR as the CSI fed back according to the result of the frequency band allocation obtained in the step 3, constructing a linear finite state Markov chain prediction equation, and predicting the CSI with propagation delay by adopting a soft mean algorithm;

the specific method of the step 4 is that,

first, CSI gamma is converted into a limited number of channel states C (m), C (m) epsilon [0, S-1]Where m is time and S is the number of channel states; according to v _s Threshold value divides CSI into discrete values of finite FSMC states, and equal probability method is adopted to select v _s Make the stationary probability of each FSMC state pi _s Equal and 1/S;

in the formula (5), σ ² Is the rayleigh fading channel gain variance; setting v ₀ ＝0，v _S+1 Finding each threshold v ∞ _s (S-1, 2, …, S) value, partitioning CSI into [0, v ₁ ),[v ₁ ,v ₂ ),…,[v _s Infinity); when CSI falls in the interval [ v ] _s ,v _s+1 ]Definitions c (m) ═ s;

secondly, solving the linear correlation coefficient psi _l (ii) a Mapping a training sequence T (m) to different state regions by using a state label (C (m) ═ s) through a quantization method, and reusing a Yule-Walker partyProgram calculation psi _l (ii) a Update and record psi using equation (6) _l ：

ψ _l (m,s)＝ψ _l (m,C(m+1)＝s|C(m),...,C(m-L+1)) (6)，

In formula (6), phi _l (m, s) represents the L-th linear correlation coefficient when the channel status is from C (m-L +1), C (m-L +2), …, C (m) to C (m +1) ═ s; obtaining a plurality of temporary linear coefficients by adopting a plurality of groups of T (m), and then calculating the average value of the linear coefficients;

thirdly, establishing a state transition probability matrix P (m); wherein p is _q,w (m) element represents the probability, p, for the CSI to transition from state q to state w _q,w (m) ═ Pr (C (m) ═ q | C (m-1) ═ w) with dimensions sxs; in a small-scale fading channel, assuming that the FSMC state only occurs between the current state or adjacent states from time m-1 to m, the CSI in the state q can be transferred to the adjacent state (q-1)/(q +1) or be kept in the original q state;

p _q,w (m) is approximately expressed as:

definition of p _0,0 ，p _S-1,S-1 ，

In the formula (8), T _s Is a symbol period, f _d Is the maximum doppler shift;

finally, according to the feedback delay CSI, utilizing soft mean algorithm to make psi _l (m, s) and P (m) are substituted into equation (10) to predict the delay t _m CSI of m +1 after the time slot;

in formula (9), P(s) ═ P (C (m +1) ═ s | C (m), C (m-1), C (m-L +1)) represents a conditional probability that H (m +1) holds the C (m +1) state;

and 5, performing adaptive resource allocation of the data nodes according to the CSI obtained by prediction in the step 4.

2. The method according to claim 1, wherein in step 1, in a cell containing a single control node and multiple data nodes, a region smaller than the optimal distance threshold is divided into a central region, and vice versa, the data nodes in the central region are central nodes, and the data nodes in the edge region are edge nodes;

the time delay difference tau of the receiving signal i and the interference signal j _i Comprises the following steps:

in the formula (1), t _j,send 、t _i,send Time of sending signal for interfering cell control node and target cell control node respectively, (x) _i ,y _i ,z _i )、(x _j ,y _j ,z _j ) Position coordinates, d, of a target cell data node i and an interfering cell control node j, respectively _oi Is the distance between the center of the target cell and the data node in the cell, d _ij And the distance between the data node of the target cell and the control node of the interference cell is represented as the propagation speed of the acoustic signal in the seawater at the speed of 1500 m/s.

3. The method according to claim 1 or 2, wherein in step 2, the SINR formula of the center node and the edge nodes is:

in formula (2), P _i Indicating that data node i is from the target cellControlling the effective power, P, obtained by the node _j Representing the interference power, A, obtained by the data node i from the interfering cell control node _i (l,f)，A _j (l, f) transmission attenuations of the control node and the interference node at node I, respectively, [ delta ] f represents the frequency bandwidth of each carrier in the underwater sound channel, and I _j For received interference from neighboring cells at the same frequency, α is the ratio of the transmission power of the cell edge node to the central node,/ _i 、l _j Respectively expressed as the distances between the target cell control node and the interfering cell control node and the data node i, k is an expansion factor, and k is 1 when the cylinder surface is transmitted, k is 2 when the sphere surface is transmitted, h _i (l _i ,f)、h _j (l _j F) are the channel gains obtained by the data node i from the target cell control node and the interfering cell control node, respectively,

the transmission attenuation of the target cell control node and the interference cell control node at the data node i, the power spectral density N (f) of the ocean noise are calculated according to a Wenz model, and the seawater absorption loss coefficient a (f) is calculated according to a Thorp formula:

4. the method for interference mitigation and resource allocation of an underwater acoustic soft frequency reuse network according to claim 1 or 2, wherein the specific method of step 5 is:

firstly, according to the predicted CSI obtained in the step (4), a proportional fair algorithm is adopted to consider the instantaneous transmission rate and the average transmission rate of the data nodes to determine the priority of using the subcarriers by the data nodes at the data scheduling moment, and subcarrier resources are distributed to each data node according to the priority;

secondly, optimizing the number of bits loaded on each data node subcarrier by adopting an adaptive modulation Chow algorithm, and maximizing the channel capacity on each subcarrier under the constraints of a system target bit error rate BER and total power;

finally, according to bit distribution result bit _n Calculating load power

Where n is the number of subcarriers, and Γ ═ ln (5 × BER)/1.6.

Images