A perception-enhanced communication interference integrated network node deployment and resource optimization method

Abstract

The application provides a kind of perception enhanced communication interference integrated network node deployment and resource optimization method, method includes: step one, establish communication interference integrated network model;Step two, establish communication interference integrated network node deployment and resource optimization problem;Step three, spatial loss field establishment;Step four, channel gain sample acquisition and estimation given regional arbitrary link channel gain using Kriging interpolation method;Step five, communication interference integrated network node working state joint optimization.The application improves environmental perception ability, reduces the estimation error of link channel gain, improves network integration index, accelerates the convergence speed of algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, in particular to a perception-enhanced integrated communication and jamming network node deployment and resource optimization method. BACKGROUND

[0002] Communication and jamming are two key elements in the field of electromagnetic confrontation. Benefiting from the technical breakthroughs in software-defined radio and signal processing, integrated communication and jamming networks (ICAJN) are rapidly developing. An ICAJN is composed of multiple integrated communication and jamming nodes, which utilize shared hardware platforms, databases, and information processing units to enhance communication and jamming functions in one system. This integration not only reduces hardware costs but also reduces the need for information exchange between different systems. Despite the above advantages, ICAJN faces at least two challenges: On the one hand, the distributed structure increases the network coverage range, improves the uneven distribution of signal strength, but brings difficulties in network coordination and challenges in network connectivity maintenance. On the other hand, the communication between multiple nodes in ICAJN is affected by the propagation environment such as terrain and buildings, which requires environmental perception to achieve effective node deployment and resource optimization. Existing research on ICAJN node location deployment and resource allocation is mostly based on the circular coverage model caused by large-scale fading, which cannot simulate real signal coverage. This affects the effectiveness of network node location deployment and resource allocation, and reduces network performance. Therefore, enhancing the environmental perception capability of the network and improving the performance of ICAJN in complex geographical environments is a promising but unexplored direction.

[0003] In the field of electromagnetic environment perception, there are mainly two methods: deterministic channel state calculation and channel state estimation. Ray tracing is a typical deterministic channel state calculation method, which obtains the channel gain of a link by calculating the amplitude, phase, delay, and polarization of each ray connecting the end points according to the terrain data. This method is complex and requires the construction of a physical model in advance, and is suitable for scenarios with high estimation accuracy requirements and insensitive to real-time requirements. Channel state estimation methods mainly use spatial interpolation techniques such as inverse distance weighting (IDW) and Kriging, which estimate the channel state in a specific region by using the sampled points obtained by distributed nodes. A key limitation of channel state estimation methods is that they can only predict the channel state between a specified point (usually the location of the transmitter) and other points in the region, but not for links where the end points are not at the specified point.

[0004] Coverage and connectivity are the core metrics of ICAJN performance. From the perspective of network node deployment and resource allocation, coverage and connectivity show contradictory characteristics. Coverage metric requires nodes to be far enough from each other to reduce overlapping coverage areas, while connectivity requires nodes to be close enough to ensure that links between them are reachable. Some researches have been proposed to explore the design of simultaneously satisfying coverage and connectivity. The strip deployment pattern is proved to be close to optimal under the equal transmit power and free space path loss model; the optimal deployment scheme when the coverage distance and communication distance are not equal is also proposed. These works provide guidance for node deployment, but they mostly only consider the large-scale fading model, assuming that the signal attenuation in all directions is uniform. In fact, factors such as uneven terrain, buildings, etc. not only cause large-scale fading, but also cause shadow fading (mesoscale fading), small-scale fading and other effects on signal transmission. Considering only the circular domain coverage model caused by large-scale fading makes it challenging to simulate the real signal coverage situation. SUMMARY

[0005] The application provides a perception-enhanced communication and interference integrated network node deployment and resource optimization method, which can be used to solve the technical problem of low accuracy in simulating real signal coverage.

[0006] The application provides a perception-enhanced communication and interference integrated network node deployment and resource optimization method, which includes:

[0007] Step one, establish a communication and interference integrated network model.

[0008] The communication and interference integrated network model includes K integrated nodes numbered Peer; the network needs to complete double tasks in the adversarial scenario: ensure communication coverage in the target area , and the target area has a probability of authorized user existence; and implement interference coverage in the suspicious area , and the target area has a probability of illegal user existence;

[0009] In order to guarantee the connectivity of the network and make the coverage more fine, the communication and interference integrated network workflow includes an environment perception stage and an execution stage; all integrated nodes run in perception mode in the environment perception stage, sample channel gain in pairs, and further infer node coverage boundary through spatial interpolation; in the execution stage, the integrated nodes switch to communication or interference mode according to the perception and calculation results in the environment perception stage, to execute the communication and interference coverage tasks;

[0010] Consider the wireless link from position x k to x l , where x k , x lrespectively, the channel gain of the link, denoted as is decomposed into three parts: path loss, shadowing fading and small-scale fading; by eliminating the effect of small-scale fading, the channel gain (in decibel) is modeled as

[0011]

[0012] d0is the reference distance, around which the path loss is characterized by the free space fading, ||·|| denotes the 2-norm, γ is the path loss exponent, which is determined by the signal propagation environment, is the shadowing fading part, which reflects the effect of obstacles on signal propagation and follows a Gaussian distribution;

[0013] For the connectivity of the network, the network’s adjacency matrix is denoted as where the element c k,l indicates whether there exists a reachable link between the kth and the lth node; when the received signal power exceeds the threshold p c,th under the maximum transmit power condition, there exists a reachable communication link; based on this, c k,l is denoted as

[0014]

[0015] where is an indicator function, whose value is 1 when the argument satisfies the constraint, otherwise 0, p max is the maximum transmit power of the node; substituting equation (1) into equation (2), the constraint condition is expressed in terms of the link shadowing fading, i.e.

[0016] For the communication coverage, the nodes are considered to work at different frequencies and do not interfere with each other; then, the node with the highest signal strength in the communication group is selected to provide service for the location x; based on this, the effective communication coverage area is denoted as which is written as

[0017]

[0018] where is the target area, is the set of nodes in the communication mode, is the index of the node serving the location x;

[0019] For the interference coverage, the effective interference coverage area, is denoted as

[0020]

[0021] where denotes the set of integrated nodes for interference.

[0022] Step two, establishing the communication and interference integrated network node deployment and resource optimization problem;

[0023] After defining the coverage area as a measure of communication and interference performance, an integrated index is designed to comprehensively consider communication, interference performance and network energy consumption, and to achieve energy-saving communication and interference coverage. The optimization variables of the problem include the deployment location of the nodes, the transmission power and the grouping strategy, and the optimization quantity is integrated into the working state set ; To this end, the optimization problem is described as follows:

[0024]

[0025] where denotes the communication coverage rate, denotes the interference coverage rate, and is the resource consumption caused by p k ; C1 represents the signal transmission power constraint of all nodes; C2 is the location constraint, which stipulates the distribution of nodes in the target area , and C3 represents the network connectivity constraint.

[0026] Problem (5) or formula (5) is a mixed integer nonlinear programming (MINLP) problem, which is very difficult to solve. On the one hand, in the objective function, and are determined, and the solution of the Area(·) function is a non-convex function about the optimization variable , and the two-layer non-convex functions are nested with each other, which increases the difficulty of solving. In the constraint condition, the matrix power calculation in the network connectivity constraint is also a non-convex process. On the other hand, from formula (3) and (4), it can be seen that in the case of node position to be optimized, the solution of and actually requires to give the channel quality g x,y of any link in the given area. The deterministic channel state calculation and channel state estimation method cannot meet the above requirements. Therefore, problem (5) cannot be solved directly.

[0027] Step three, spatial loss field establishment;

[0028] The channel quality g x,yis a necessary condition for solving problem (5), while the current pilot channel estimation method can only solve the channel gain of a certain link, and the channel state estimation method can only solve the link channel gain from the transmitting node to any position, and cannot estimate the link channel gain of which both ends are not at the transmitting node position.

[0029] According to formula (1), the channel gain includes path loss and shadow fading, wherein the path loss is obtained by the link distance, and therefore the challenge of estimating the channel gain lies in the solution of shadow fading s k,l . In order to solve this problem, the patent proposes the concept of spatial loss field, considering that the shadow loss experienced on the link is the result of the potential spatial loss field p(x);

[0030] In a non-specific scenario, that is, the distribution of obstacles causing fading obeys Poisson spatial randomness, the spatial loss field p(x) is a zero-mean and isotropic Gaussian random field with exponential fading spatial correlation; that is, The covariance of the spatial loss values of any two points x and y in the target area is written as:

[0031]

[0032] Wherein ||x-y|| represents the Euclidean distance between points x and y, σ s is the standard deviation of shadow fading, and δ is a spatial constant;

[0033] As described above, the shadow fading experienced on the link is the result of the potential spatial loss field; specifically, the shadow fading loss s x,y between any two points is the weighted line integral of the spatial loss field between the two points:

[0034]

[0035] Wherein As a weight coefficient, it is embodied that the shadow fading influence of the same obstacle on the short path is greater than that on the long path, because there are more scatterers and greater diffraction possibilities on the long path, while the short path is just the opposite;

[0036] Since all link shadow fading is a function of the spatial loss field, the correlation of the link shadow fading is derived from the correlation of the spatial loss field; the cross-correlation function of the shadow fading s x,y , s u,v on the link x→y and the link u→v is written as:

[0037]

[0038] After obtaining the correlation of link shadow fading, the distributed characteristics of ICAJN nodes are used to sample the shadow fading value of the link to obtain diverse samples, and the shadow fading value of any link is estimated by interpolation and the irregular coverage boundary is inferred.

[0039] Step 4: Obtain channel gain samples and estimate the channel gain of any link within a given region using Kriging interpolation.

[0040] The process of obtaining channel gain samples is as follows:

[0041] During the sensing phase, nodes sequentially transmit pilot signals in a time-division manner, dividing one sensing period into K time slots. Within time slot k, the k-th node transmits a signal with a specific power, while other nodes receive the signal. The channel gain between the two nodes is obtained based on the power difference between the received and transmitted signals. Given that the relative positions of the receiver and transmitter are known, measurements of shadow fading between the two points are obtained. The relationship between measured and actual shadow fading values ​​is modeled as follows: Where ε k,l Including measurement errors and uncertainties from small-scale fading in the channel, it follows a zero-mean Gaussian distribution; from the perspective of the k-th node, it receives training signals from the other K-1 nodes in one sensing cycle, and these K-1 nodes are represented as a set. definition It is the signal received by the k-th node within one sensing period. A set; definition This is the collection of sampled data from all sampling points; It can also be expressed as:

[0042]

[0043] Where s and ε are respectively derived from {s k,l} and {ε k,l}, Reconstructed to obtain;

[0044] By utilizing the spatial correlation of link shadow fading s, we estimate the shadow fading s of any link x→y using Kriging interpolation. x,y Value:

[0045]

[0046] in Let C represent the covariance vector between the sampled points and the predicted points. s =cov{s},C ε =cov{ε}, Q=C s +C ε; p(x) is a random field with zero mean. According to the relationship in formula (7), s x,y It is also a variable with zero mean, therefore, In shadow fading parameters Given that δ is fixed, and C s It is obtained through equation (8);

[0047] The link channel gain is estimated using Kriging interpolation, as follows:

[0048] The inputs are the link endpoint values ​​to be estimated, x = (x1, x2), y = (y1, y2), and the estimated sampling error covariance C. ε ;

[0049] The output is the link channel gain estimate.

[0050] Step 4-1: A sensing period T has K time slots, and K users operate in a time-division multiple access manner; where the k-th user has a transmit power of p in the k-th time slot. k Pilot signal;

[0051] Step 4-2: The l-th user in the time slot Internally, the signal received from the k-th user is expressed as: Further calculations

[0052] Step 4-3: Calculate according to formula (1)

[0053] Step 4-4: Summarize the sampled signal of the l-th user within one period. and sampling data from all sampling points

[0054] Steps 4-5: Based on the sampling data Get the current environment for The estimated value;

[0055] Steps 4-6: Use formula (10) to calculate s x,y Perform interpolation prediction to obtain the estimated value.

[0056] Steps 4-7: Calculation

[0057] The entire process of estimating the link channel gain within the target area is summarized in the algorithm described above. Steps 4-1 to 4-4 complete the acquisition of link shadow fading samples within one sensing period; step 4-5 is the parameter estimation stage; and steps 4-6 to 4-7 provide estimates of the link shadow fading portion and the overall gain.

[0058] Complexity analysis: The complexity of this algorithm is mainly caused by the execution of steps 4-6. In the worst case, the complexity of each link channel gain estimation can be estimated as where p denotes the number of partitions when calculating the link shadow fading covariance in MATLAB using the function integral2 according to (8), and the second term is the complexity of matrix inversion.

[0059] Step five: joint optimization of communication and interference integrated network node working states;

[0060] According to the estimation information of the link channel gain in step four, the joint optimization of the working states of the communication and interference integrated network is carried out, including the working states of node deployment position, power allocation and grouping strategy, to realize the communication and interference coverage under the network connectivity constraint. In order to cope with the challenge of solving the problem caused by the non-convexity of the objective function and the constraint, the grey wolf optimization (GWO) algorithm is used to give the approximate optimal solution of the problem.

[0061] As a new swarm intelligence (SI) algorithm, the grey wolf optimization algorithm simulates the hunting process of wolf packs in nature, models the hunting techniques and social hierarchy of wolf packs, and regards the optimal solution of the problem as prey. In the process of gradual iteration, the prey (optimal solution) is approached. Compared with other swarm intelligence algorithms, the grey wolf optimization algorithm has the advantages of no need to calculate the gradient of the objective function, fewer defined parameters, convenient adjustment and clear physical meaning, and has been proved to be a convenient and reliable algorithm in many fields.

[0062] The standard grey wolf optimization algorithm operation process starts from the initialization of the feasible solution of the agent. The number of agents, also known as the number of grey wolf populations, is defined as N, and the position vector of each agent is the feasible solution of the problem (5), y n is represented as:

[0063]

[0064] wherein, represents the value of the z-th dimension of the position vector of the n-th agent, and the total dimension is Z, which includes the positions, power allocation and communication and interference node grouping of all integrated nodes, and the set of dimension definitions is

[0065] After the initialization is completed, the grey wolf optimization algorithm enters the exploration stage; in this stage, the top three agents with the best fitness are found as alpha wolf, beta wolf and delta wolf, and the others are defined as omega wolf. Alpha, beta and delta wolves are considered to be the wolves that can best approach the prey position under the current conditions;

[0066] The parameter updating stage, that is, the attack on the prey; at this time, the agent is positioned as an omega wolf except for alpha, beta and delta wolves, and the position of the first three wolves is iteratively optimized, and then the node position of the alpha, beta and delta wolves with better fitness is selected to update;

[0067] In order to better model the exploration stage and the parameter updating stage, the number of iterations is added to formula (11), and the position vector of the first three agents in the lth iteration is defined as:

[0068]

[0069] Wherein, the number of alpha, beta and delta satisfies the following conditions:

[0070]

[0071] Wherein, f(·) represents the calculation function of the objective function in formula (5);

[0072] After determining the positions of alpha, beta and delta wolves, the position of omega wolf is further updated, and first define two state update vectors

[0073]

[0074] Wherein, a z (l) is a parameter that decreases linearly with the number of iterations, ranging from 2 to 0, and is modeled as:

[0075]

[0076] On this basis, the relative distance between each omega wolf and alpha, beta and delta wolves is respectively represented as And The specific calculation method is as follows:

[0077]

[0078] The position updating formula of omega wolf is:

[0079]

[0080] The alpha, beta and delta solutions of the nth omega wolf in the lth iteration are defined as:

[0081]

[0082] The updated solution of the nth omega wolf is:

[0083]

[0084] After completing the optimization iteration of all agents, we get

[0085] The working state joint optimization process is as follows:

[0086] The input quantity is the initial position of the N agents defined by satisfying the constraint conditions 1 and 2 of problem (5) The maximum number of iterations l max ;

[0087] The output quantity is the optimal working state y of the network node obtained by the algorithm α (l max );

[0088] Step 5-1: input the positions of the N agents into step four, calculate the link channel gain, and further obtain their fitness function

[0089] Step 5-2: verify whether the solution corresponding to the agent satisfies the connectivity constraint, if not, impose a penalty by multiplying the fitness function by the discount factor λ;

[0090] Step 5-3: use formula (13) to calculate the indices of the top three solutions so far, i.e. α(l), β(l) and δ(l);

[0091] Step 5-4: use formula (14) and formula (15) to calculate the search coefficient vector;

[0092] Step 5-5: update the new position of the agent y n (l+1) according to formula (16)-(19);

[0093] Step 5-6: adjust the values in y n (l+1) that exceed the search space to correspond to the space boundaries;

[0094] Step 5-7: update l = l + 1 until l > l max , end; otherwise, return to step 5-1.

[0095] Algorithm convergence and complexity analysis: since the working state set By using the GWO algorithm to optimize in an iterative manner, the objective function f(y n (0)) ≤ f(y n (1)) ≤ … ≤ f(y n (l max )). In addition, the working state set is constrained by C1 and C2, so f(y n (l max)) will converge to a maximum limit point. The main idea of the algorithm is to calculate the fitness function of the agent in each iteration. The worst-case complexity can be estimated as where p represents the number of partitions when calculating the link shadow fading covariance in MATLAB using the function integral2 according to formula (10), and the complexity of calculating the fitness function of each agent is

[0096] The perception-enhanced communication interference integrated network node deployment and resource optimization method provided by the application has the following technical effects:

[0097] Improve environmental perception capability: through the implementation of the spatial loss field, the system can more accurately quantify the link shadow fading, and by combining path loss and shadow fading, it can generate more realistic irregular coverage boundaries, thereby improving the accuracy of signal coverage. Thus better adapt to the actual signal transmission environment, improve the performance of communication and interference.

[0098] Reduce the estimation error of link channel gain: the proposed Kriging estimation algorithm is superior to the traditional IDW estimation and path loss model estimation algorithm in terms of perception performance, especially when the number of antennas is increased, the performance gain can reach about 4dB.

[0099] Improve network integration index: under different prior information, the system can achieve higher integration index by adjusting the transmission power and optimization algorithm, that is, while maintaining network connectivity, it realizes energy-saving communication and interference coverage.

[0100] Accelerate the convergence speed of the algorithm: compared with PSO and ABC algorithms, GWO algorithm shows faster convergence speed and better performance, which indicates that in practical application, the system can find the optimal solution more quickly.

[0101] Dynamic adjustment of node deployment: ICAJN can dynamically adjust the node position according to the distribution of the spatial loss field, avoiding the link passing through the high loss area, so as to realize better communication and interference coverage effect under the constraint of network connectivity.

[0102] Improve the connectivity rate: through the optimization algorithm and improved working protocol, the system improves the proportion of test times that meet the connectivity constraint in multiple experiments, that is, improves the connectivity rate of the network. BRIEF DESCRIPTION OF DRAWINGS

[0103] Figure 1 The system model diagram and protocol diagram of the communication interference integrated network provided by the application are provided;

[0104] Figure 2 The spatial loss field implementation and the corresponding channel gain contour map provided by the application are provided;

[0105] Figure 3 The integration index and node quantity relationship graph corresponding to the five algorithms provided by the present application;

[0106] Figure 4 The two most advanced node deployment schemes provided by the present application simultaneously consider coverage and connectivity;

[0107] Figure 5 The estimation error and node number relationship graph under different estimation algorithms provided by the present application;

[0108] Figure 6 The target function and maximum transmission power relationship graph under different prior information provided by the present application;

[0109] Figure 7 The target function and iteration number relationship graph under different swarm intelligence algorithms provided by the present application;

[0110] Figure 8 The spatial loss field and corresponding position optimization result and signal strength distribution graph provided by the present application. DETAILED DESCRIPTION

[0111] In order to make the purpose, technical scheme and advantages of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the drawings.

[0112] The embodiments of the present application will be introduced below with reference to the drawings.

[0113] In order to evaluate the performance of the proposed method, the ICAJN containing K=6 nodes in a 200m×200m area, representing a city microcell scenario, is analyzed, in which each multi-functional node acts as a mobile base station to jointly provide communication and interference coverage. The maximum transmission power of the node p max =40dBm, the network operating frequency f=2570-2620MHz, the path loss exponent γ=3.2, the communication reception power threshold p c,th =-60dBm and the effective interference power threshold p j,th =-60dBm. The shadow fading parameter δ=50. The discount factor in the improved GWO algorithm is set to λ=0.5. The suspicious area is defined as a square with a coordinate range of [0m:50m; 0m:50m].

[0114] Figure 2 (a) gives a specific implementation of the spatial loss field in a 200m×200m area, which is obtained using the spatial correlation relationship in equation (8). In order to prove that the spatial loss field is effective in quantifying link shadow fading, Figure 2(b) and 2(c) show the channel gain contour map emitted from point [70.5m, 12.5m]. Figure 2 (b) shows the circular contour map considering only path loss, while Figure 2 (c) shows the irregular contour map obtained by combining path loss and shadowing fading. Compared with Figure 2 (b), Figure 2 (c) indicates that when the link goes through the area of positive p(x), such as [15m:35m; 0m:25m], [10m:25m; 95m:110m], the spacing between two adjacent contours becomes larger, which indicates that the signal attenuation slows down. On the contrary, when the link goes through the area of negative p(x), such as [120m:160m; 115m:145m], the spacing between two adjacent contours becomes smaller, which indicates that the signal attenuation speeds up, and the shadowing fading plays a reverse role on signal propagation.

[0115] Figure 3 The relationship between the integration index and the number of nodes in five different cases is shown. The red curve represents the upper limit, which is the theoretical unattainable value obtained by using the proposed algorithm with the exact coverage boundary. The blue and yellow curves are obtained using the estimated coverage boundary and the circular coverage boundary, respectively. As a baseline scheme, two node deployment schemes are provided that simultaneously satisfy the network coverage and connectivity requirements based on the circular domain coverage model. These schemes are represented by the black and green lines, respectively, corresponding to the strip and grid node deployment schemes, respectively, whose schematic diagrams are shown in Figure 4 It is clear that the proposed method is significantly better than the baseline method, and the improvement is mainly due to two aspects: perception gain and execution gain. The execution gain is represented by the yellow shadow, which is reflected in the difference between the proposed algorithm using the circular coverage boundary and the baseline scheme, both of which are based on the circular coverage model and have no environmental perception ability. The perception gain comes from the reduction of the link channel gain and the estimation error of the coverage boundary, which is represented by the blue shadow. In addition, as the number of nodes increases, the gap between the blue curve and the red curve decreases. This indicates that increasing the number of perception nodes can reduce the gap between the estimated coverage boundary and the exact coverage boundary.

[0116] Figure 5The proposed Kriging estimation algorithm is compared with two other estimation algorithms: the inverse distance weighted (IDW) estimation and the path loss model estimation in terms of the sensing performance. The IDW estimation algorithm takes the link distance as the weight of the weighted average, where the link distance is represented as the sum of the distances of two endpoints. The path loss model estimation only considers the path loss part and gives the link channel gain according to the link length, which is given as a benchmark. The normalized root mean square error (NRMSE) is used as the performance indicator.

[0117] where Q is the total number of experiments, K val is the number of links in one experiment, is the estimated value of the validation link in the qth experiment, g val (q) are their true values. The locations of the validation nodes are indicated by black squares in Figure 2 (c) with K val = 36 and Q = 100. As K increases, the NRMSE of the link channel gain of the two spatial interpolation estimation algorithms decreases, and the NRMSE is significantly reduced compared with the path loss model estimation. The proposed Kriging estimation algorithm is superior to the IDW estimation algorithm, and this gain becomes greater as the number of antennas increases, and the performance gain reaches about 4 dB when K = 8.

[0118] Figure 6 The impact of the transmit power on the comprehensive indicator under different prior information is demonstrated. The proposed scheme refers to ICAJN working according to the proposed protocol. Both the upper bound and the lower bound 1 are obtained by step six, the difference lies in that the upper bound is calculated using the exact coverage boundary, and the lower bound 1 is calculated using the circular domain coverage boundary. On the basis of the lower bound 1, the lower bound 2 improves the communication threshold p c,th = -55 dBm in the optimization process so as to more easily meet the network connectivity constraint, and its comprehensive indicator is superior to that of the lower bound 1. Figure 6 It is further shown that the sensing function in ICAJN better adapts to the actual signal transmission environment by estimating the link shadow fading, thereby enhancing the ability of communication and interference. In addition, in all four schemes, the comprehensive indicator increases with the increase of the maximum transmit power.

[0119] In order to verify the effectiveness of the GWO algorithm used in the patent, Figure 7The relationship between the comprehensive index and the number of iterations of three swarm intelligence algorithms, GWO algorithm, particle swarm optimization (PSO) algorithm and artificial bee colony (ABC) algorithm, is shown. It can be seen that the GWO algorithm has faster convergence speed and better algorithm performance than the PSO and ABC algorithms, and converges after 100 iterations when K=6.

[0120] Figure 8 Two implementations of the spatial loss field are shown, as well as the node position optimization results and signal strength distribution under the corresponding spatial loss field. Figure 8 (a) simulates a scenario where the distribution of attenuation obstacles follows a Poisson spatial random process; Figure 8 (b) is based on Figure 8 (a), the spatial loss field in the area [60m:100m; 100m:140m] is set to p(x)=-5 to simulate the effect of large obstacles on signal propagation in the precise area. Figure 8 (c) and Figure 8 (d) respectively show the node position optimization results and signal strength distribution under the spatial loss field in Figure 8 (a) and Figure 8 (b), where the blue cross indicates the deployment position of the node in the communication mode, the red cross indicates the deployment position of the node in the interference mode, and the black straight line indicates the reachable link between the two nodes. It can be found that in the anisotropic signal propagation environment, ICAJN can dynamically adjust the node position to avoid the link passing through the position with high spatial loss field value, thereby achieving better communication and interference coverage effect under the network connectivity constraint.

[0121] In order to solve the problem of high computational complexity in the field of electromagnetic environment perception and the problem of being unable to predict the channel gain of any link in a given area, the invention introduces the concept of spatial loss field and deduces the relationship between link channel gain and spatial loss field. On this basis, the channel gain of any link is inferred by sampling and spatial interpolation, and the coverage area of the node is further inferred, thereby enhancing the environmental perception ability.

[0122] In order to solve the problem of network coverage and connectivity optimization difficulty in a specific propagation environment, the invention formulates the problem of simultaneously meeting the coverage and connectivity. A comprehensive index is developed to optimize the deployment position, power allocation and grouping strategy of the node. The grey wolf optimization algorithm is used to iteratively solve the problem to determine the best working state in the current environment.

[0123] An extended simulation is provided in the urban microcell scenario. The simulation results show that the coverage boundary estimation and working state joint optimization algorithm has an advantage in single field capacity compared with the baseline algorithm, and the designed protocol can coordinate multiple functions, showing adaptability to dynamic electromagnetic environment.

[0124] The above-described embodiments of the present application are not intended to limit the scope of the present application.

Claims

1. A method for perception-augmented communication-interference integrated network node deployment and resource optimization, the method comprising: receiving a plurality of network node deployment and resource optimization requests; determining a plurality of network node deployment and resource optimization solutions; and providing a plurality of network node deployment and resource optimization solutions to a plurality of network nodes. The method comprises: Step one, establishing a communication interference integrated network model; Step two, establishing a communication interference integrated network node deployment and resource optimization problem; Step three, spatial loss field establishment; Step four, channel gain sample acquisition and estimation of any link channel gain in a given region using Kriging interpolation method; Step five, communication interference integrated network node working state joint optimization; The communication-interference integrated network model comprises K integrated nodes numbered as The peer-to-peer integrated nodes; the network needs to complete double tasks in the confrontation scene: ensure the communication coverage on the target area There is a probability of authorized users in the target area; and in the suspicious area Implement interference coverage, there is a probability of illegal users in the target area; The communication interference integrated network workflow includes an environment perception stage and an execution stage; in the environment perception stage, all integrated nodes operate in a perception mode, sample channel gains in pairs, and further infer node coverage boundaries through spatial interpolation; in the execution stage, the integrated nodes switch to communication or interference mode according to the perception and calculation results in the environment perception stage to perform communication and interference coverage tasks; Consider a wireless link from location x k to x l , where x k , x l denote the locations of the kth and lth nodes, respectively; the channel gain of the link, denoted as , is decomposed into three parts: path loss, shadowing and small-scale fading; by eliminating the effect of small-scale fading, the statistical model of is given as d0is a reference distance, around which the path loss is characterized by free space fading, || · || denotes the 2-norm, γ is a path loss exponent, which is determined by the signal propagation environment, is the shadow fading part, which reflects the impact of obstacles on signal propagation and follows a Gaussian distribution; Step three, spatial loss field establishment; including: Using the perception ability of the integrated nodes in ICAJN, the channel gain in the target area is estimated to guide the solution of the optimization problem (5); According to formula (1), the channel gain includes path loss and shadow fading, where the path loss is obtained from the link distance; the concept of spatial loss field is proposed, considering that the shadow loss experienced on the link is the result of the potential spatial loss field p(x); In a non-specific scenario, i.e. the distribution of obstacles causing fading is subject to Poisson space randomness, the spatial loss field p(x) is an isotropic, generalized, stationary Gaussian random field with zero mean and exponential fading spatial correlation; i.e. The covariance of the spatial loss values of any two points x and y in the target region is written as: where ||x - y|| denotes the Euclidean distance between points x and y, σ s is the standard deviation of shadow fading, and δ is a spatial constant. The shadow fading experienced on a link is the result of the underlying spatial loss field; specifically, the shadow fading loss s between any two points x,y is the weighted line integral of the spatial loss field between the two points: wherein As a weight coefficient, it embodies that the same obstacle has greater shadow fading influence on the short path than on the long path. Since all the link shadow fading is a function of the spatial loss field, the correlation of the link shadow fading is derived from the correlation of the spatial loss field; the cross-correlation function of the shadow fading s x,y , s u,v on link x→y and link u→v is written as: After obtaining the correlation of link shadow fading, the distributed characteristics of ICAJN nodes are used to sample the diversity of shadow fading values of the link, and interpolation is used to estimate the shadow fading values of any link and infer irregular coverage boundaries.

2. The method of claim 1, wherein, Step one, establishing a communication interference integrated network model, including: For connectivity of the network, the network's adjacency matrix is denoted as C where the element c k,l indicates whether there is a reachable link between the kth and the lth node; there is a reachable communication link when the received signal power exceeds a threshold p c,th under the maximum transmit power condition; based on this, c k,l is denoted as: wherein is an indicator function that takes the value 1 when the argument satisfies the constraint and 0 otherwise, p max is the maximum transmit power of the node; substituting equation (1) into equation (2), the constraint is expressed in terms of the link shadow fading, i.e. For communication coverage, consider that the nodes work at different frequencies and do not interfere with each other; then, select the node with the highest signal strength in the communication group to provide service for position x; based on this, the effective communication coverage area is represented as is written as: wherein is a target region, is a set of nodes in a communication mode, is a node index of a service location x; For the interference coverage, the effective interference coverage area, is represented as: wherein denotes a set of integrated nodes for interference.

3. The method of claim 2, wherein, Step two, establishing a communication interference integrated network node deployment and resource optimization problem; including: The optimization variables of the problem include the deployment locations of the nodes, the transmission power and the packet policy, the optimization quantity is integrated into the working state set For this purpose, the optimization problem is formulated as follows: wherein denotes the communication coverage, denotes the interference coverage, and is the resource consumption caused by p k ; C1 represents the signal transmission power constraint of all nodes; C2 is a location constraint, which stipulates the distribution of nodes in the target area ; C3 represents the network connectivity constraint; Problem (5) is a mixed integer nonlinear programming problem.

4. The method of claim 3, wherein, Step four, channel gain sample acquisition and estimation of any link channel gain in a given region using Kriging interpolation method; including: The channel gain sample acquisition process is as follows: In the sensing phase, the nodes transmit pilot signals in turn in time division, dividing a sensing period into K time slots, in time slot k, the kth node transmits a signal with a specific power, and other nodes receive the signal, and the channel gain between the two nodes is obtained according to the power difference between the received signal and the transmitted signal, and in the case that the relative position between the receiver and the transmitter is known, the measurement value of shadow fading between the two points is obtained The relationship between the measurement value of shadow fading and the actual value is modeled as Where ε k,l Including measurement error and uncertainty from channel small-scale fading, subject to zero-mean Gaussian distribution; from the perspective of the kth node, in a sensing period, the training signals from other K-1 nodes are received, and the K-1 nodes are represented as a set Definition is the set of signals received by the kth node in a sensing period Definition is the set of sampling data of all sampling points is represented as: where s and ε are given by {s k,l} and {ε k,l}, reconstructed as; The value of the shadow fading s for any link x→y is estimated by means of Kriging interpolation, exploiting the spatial correlation of the shadow fading s x,y . where denotes the covariance vector between the sample point and the prediction point, and C s = cov{s}, C ε = cov{ε}, Q = C s + C ε ; p(x) is a zero-mean random field, and s x,y is also a zero-mean variable, therefore, with the shadow fading parameters and δ determined, and C s are obtained through equation (8); The process of using Kriging interpolation method for link channel gain estimation is as follows: The input quantities are the link endpoint values to be estimated x = (x1, x2), y = (y1, y2) and the estimated sampling error covariance C ε ; The output quantity is a link channel gain estimate Step 4-1: One sensing period T has K time slots, K users work in time division multiple access mode; the kth user transmits a pilot signal with power pk in the kth time slot k ; Step 4-2: The first user receives the signal from the kth user in time slot k, The received power is denoted as Further computation Step 4-3: Calculate according to formula (1) Step 4-4: aggregate the sampled signals of the lth user in a cycle and the sampled data of all sampling points Step 4-5: Based on the sampled data obtain an estimate of the current environment under the current environment; Step 4-6: Interpolation prediction is performed using equation (10) on s x,y to obtain the estimate Step 4-7: Calculation Steps 4-1 to 4-4 complete the acquisition of link shadow fading samples in a perception cycle; step 4-5 is the parameter estimation stage, and steps 4-6 to 4-7 give the estimation of the link shadow fading part and the overall gain.

5. The method of claim 4, wherein, Step five, communication interference integrated network node working state joint optimization; according to the estimation information of the link channel gain in step four, the communication interference integrated network working state joint optimization is carried out, including the working state of node deployment position, power allocation and grouping strategy, to realize communication and interference coverage under network connectivity constraints, including: The grey wolf optimization (GWO) algorithm is used to give an approximate optimal solution to the problem. The number of agents, also referred to as the grey wolf population, is defined as N, and the position vector of each agent is a feasible solution to the problem (5), y n is represented as: wherein, represents the value of the z-th dimension of the position vector of the n-th agent, the total dimension is Z, which contains the positions of all integrated nodes, power allocation and communication interference node grouping, defining the set of dimensions After initialization, the grey wolf optimization algorithm enters the exploration stage; in this stage, the top three fitness intelligent agents are found as alpha wolf, beta wolf and delta wolf, and the others are defined as omega wolf, and alpha, beta and delta wolf are considered as the wolves closest to the prey position under the current conditions; The parameter updating phase is entered, that is, the attack on the prey is launched; at this time, the intelligent agents other than the alpha, beta and delta wolves are positioned as omega wolves, and the position of the alpha, beta and delta wolves is iteratively optimized according to the positions of the first three wolves, and then the node positions of the alpha, beta and delta wolves with better fitness are selected to update the node positions of the alpha, beta and delta wolves; On the basis of formula (11), the number of iterations is added, and the position vectors of the first three intelligent agents in the lth iteration are defined as: Wherein, the numbers of the alpha, beta and delta wolves satisfy the following conditions: Wherein, f(·) represents the calculation function of the objective function in formula (5); After the positions of the alpha, beta and delta wolves are determined, the position of the omega wolf is further updated, and first, two state update vectors are defined wherein, a z (l) is a parameter that linearly decreases with the iteration number, ranging from 2 to 0, modeled as: On this basis, the relative distance of each ω wolf to α, β and δ wolves is represented as and The specific calculation is as follows: The position updating formula of the omega wolf is: In the lth iteration, the alpha, beta and delta solutions of the nth omega wolf are defined as: The updated solution of the nth omega wolf is: After completing the optimization iteration for all agents, we obtain The working state joint optimization process is as follows: The input quantity is the initial position of N agents defined by satisfying the constraint conditions 1 and 2 of problem (5) Maximum number of iterations l max ; The output quantity is the optimal working state of the network node y obtained by the algorithm α (l max ); Step 5-1: The positions of N agents are inputted into Step Four, the link channel gains are calculated, and further their fitness functions are obtained Step 5-2: verify whether the problem solution corresponding to the intelligent agent satisfies the connectivity constraint, if not, apply a penalty by multiplying the fitness function by a discount factor λ; Step 5-3: calculate the indexes of the first three best solutions so far, that is, alpha(l), beta(l) and delta(l), using formula (13); Step 5-4: calculate the search coefficient vector using formula (14) and formula (15); Step 5-5: Update the new position y of the agent according to formula (16)-(19) n (l+1); Step 5-6: Adjust y n (l+1) to correspond to the space boundary. Step 5-7: Update l = l + 1 until l > l max End; otherwise go back to step 5-1.

Images