A non-orthogonal channel modeling and distributed decision method for air-ground communication

Abstract

The application discloses a kind of non-orthogonal channel modeling and distributed decision method for air-ground communication, it is related to air-ground communication network communication technical field.For scene modeling, the application proposes a kind of probabilistic line-of-sight communication model, by considering the attenuation of path loss on communication distance, overcome the limitation of fixed communication distance in current non-orthogonal channel research.To solve the optimization problem of channel selection, a synchronous learning algorithm is proposed to optimize the distributed channel access strategy, reduce system interference by heterogeneous learning rate and appropriate action maintenance probability, improve network throughput, can more effectively adapt to communication requirements, so as to ensure that system performance is stable in user-intensive environment.

Description

Technical Field

[0001] This invention relates to the field of air-to-ground communication technology, and more specifically to a non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication. Background Technology

[0002] With the rapid development of drone technology, the large-scale deployment of communication nodes has significantly changed the communication landscape of modern cities and battlefields. However, the competition for spectrum resources and adjacent channel interference brought about by dense nodes are becoming increasingly severe problems. Traditional orthogonal channels are scarce, making it difficult to meet the communication needs of a large number of nodes, resulting in limited network capacity and degraded communication quality. In traditional wireless communication systems, sufficient spectrum spacing is required between two channels to avoid adjacent channel interference; such channels are called orthogonal channels. However, in dense networks, adjacent nodes often struggle to be allocated different orthogonal channels, leading to increased co-channel interference and severely impacting communication quality. In this situation, the rational use of partially overlapping channels can increase the number of parallel transmission links, improve spectrum resource reuse, and thus enhance network capacity.

[0003] However, current non-orthogonal channel transmission mechanisms have limitations. On the one hand, most models are based on idealized assumptions and fail to accurately characterize the actual impact of real-world environmental factors such as path loss on communication distance and interference, resulting in significant discrepancies between performance predictions and actual conditions. On the other hand, existing efficient channel selection algorithms mostly rely on centralized architectures, which are insufficiently adaptable to distributed cluster scenarios lacking information exchange; and existing distributed algorithms still lag behind in performance convergence speed, making it difficult to meet the real-time and efficient resource allocation requirements of high-density networks.

[0004] Therefore, how to model scenarios for complex real-world propagation environments and develop efficient distributed channel selection strategies suitable for large-scale clusters without information exchange is a key challenge for improving the communication capabilities of air-to-ground communication networks. Summary of the Invention

[0005] In view of this, this invention provides a non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication. This invention models path loss in real-world scenarios. Based on the flight characteristics of UAVs, the model effectively quantifies signal transmission attenuation caused by reflection and diffraction, and establishes a communication model based on probabilistic line-of-sight for non-orthogonal channels. Furthermore, this invention proposes a synchronous learning algorithm to optimize distributed channel access strategies. By using heterogeneous learning rates and appropriate action-hold probabilities, system interference is reduced, network throughput is improved, and the algorithm can more effectively adapt to communication requirements, thereby ensuring stable system performance in densely populated environments.

[0006] To achieve the above objectives, the non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication provided by embodiments of the present invention includes the following steps: Step 1: Construct a probabilistic line-of-sight communication model for the air-to-ground communication network, and calculate the line-of-sight link probability based on the environmental parameters of the scene and the elevation angle of each node. .

[0007] Step 2: Calculate the weighted interference distance for each node by calculating the average path loss under line-of-sight link probability, and construct different channel spacings. The relationship between down-line-of-sight link probability and interference distance is discussed, and interference user sets are defined for each user. The optimization objective is to find the optimal channel selection combination from the set of interfering users so as to minimize the cumulative interference in the network.

[0008] Step 3: Initialize the channel selection for each node in the network and set the number of iterations. Calculate the utility function in the current iteration. .

[0009] Step 4, in the (i+1)th iteration cycle, the network explores and updates the channel selection, and the nodes... According to probability Perform channel selection updates and calculate the utility function in this iteration. .

[0010] Step 5, compare nodes exist If the channel selection in the current iteration period is the same as that in the previous iteration period, return to step 4; otherwise, proceed to step 6.

[0011] Step 6, in the (i+2)th iteration cycle, the network learns and updates the channel selection, and the nodes... According to probability Perform channel selection updates and calculate the utility function in this iteration. .

[0012] Step 7: Assign i+2 to i, and repeat steps 4-6 until the entire network strategy converges.

[0013] Furthermore, in the air-to-ground communication model constructed in step 1, the probability that the communication link is a Loss of Space (LoS) link is... ;in: , These are environmental parameters; It is the elevation angle formed between the ground equipment and the drone.

[0014] The probability that the communication link is an NLoS link should be: .

[0015] Preferably, The calculation formula is as follows:

[0016] in: Indicates the altitude of the drone. This indicates the 3D distance between ground equipment and drones.

[0017] Furthermore, in step 2, the dynamic interference distance model calculates the weighted interference distance of each node by calculating the average path loss under the line-of-sight link probability.

[0018] The path loss for LoS and NLoS links is: The unit is dB. This represents the three-dimensional distance between ground equipment and the drone. Indicates the frequency of electromagnetic waves. At the speed of light, and These represent the additional path loss under LoS and NLoS links, respectively, which is determined by the environment.

[0019] Considering a statistical model based on link probability, the formula for calculating the average path loss of a wireless signal during propagation is as follows:

[0020] In an air-to-ground scenario, the signal for node communication Not lower than the threshold ; It is the signal-to-interference-plus-noise ratio in a communication link, calculated using the following formula: The path loss under non-interference conditions should satisfy the following condition: ,Right now The formula for calculating the interference distance of a wireless signal during propagation is as follows:

[0021] Let the channel numbers selected by the two users be respectively and ,So and The channel distance between them is

[0022] definition The channel distance is Throughput Normalization Factor The distance range is 0.5, when the distance between two users is less than When, the throughput normalization factor of both is... It is 0.5; when the distance between two users is greater than 0.5. At that time, throughput normalization factor Keep it at 1.

[0023] For a given transmission rate, let the effective transmission distance be denoted as . For a specific transmission rate, establish the interference distance corresponding to different channel intervals. .

[0024] The relationship between line-of-sight communication probability and interference distance at different channel distances.

[0025] The normalized throughput of a certain user is expressed as: ,in It's interference:

[0026] in It is the physical distance between the two users.

[0027] In the probabilistic line-of-sight communication model, the user At elevation angle The communication distance is ;user At elevation angle The communication distance is ;user and The physical distance between them is The channel distance is So, users and Interference between them :

[0028] in .

[0029] Similarly, users The normalized throughput that can be obtained is ,in User Accumulated MAC layer interference.

[0030] The user set is ,Right now For specific users The interference it experiences is determined by its own channel selection and the channel selection of other neighboring users.

[0031] Channel set is ,Right now Assuming the user choose As a communication channel, users are recorded. and The channel distance between them is ;remember For users Interference with users The number of users, that is:

[0032] in For indicator functions:

[0033] That is, if the user and The channel distance between them is no greater than ,but The value is 1 otherwise; when all users' channel selection combinations are... At that time, the user The achievable normalized throughput is:

[0034] in It is a user set China The number of users who caused the interference, i.e.

[0035] The optimization objective is to find the optimal combination of channel selections that minimizes the cumulative interference in the network, i.e. .

[0036] Furthermore, step 3 initializes the channel selection for each node in the network, that is, each node in the network selects a channel from the channel set. The system randomly selects a channel for communication and saves the selection of each node as the number of iterations. The initial state under the following conditions; each cognitive user From its available channel set A channel is randomly selected with moderate probability. .

[0037] Each iteration contains A series of consecutive time slots, in which Representing the During the next iteration, the user Number of successful channel accesses; User The achievable throughput is Furthermore, users The value affected by the interference is .

[0038] The utility function obtained by the user in each iteration is: .

[0039] Furthermore, step 4 performs an exploratory update, meaning that in the (i+1)th iteration cycle, each node in the network explores channel selection. According to probability Perform channel selection update:

[0040] in Representative set The cardinality is the number of elements in a set.

[0041] Set the strategy to maintain probability for:

[0042] in , , , This represents the number of iterations.

[0043] The convergence speed of the algorithm can be improved by setting different learning parameters for different users; The number of neighboring users is .

[0044] The utility function in the current iteration is: .

[0045] Furthermore, step 6 learns based on the utility function obtained in the first two iterations, that is, it updates the channel selection using Boltzmann probing based on the results of the two explorations, avoiding getting trapped in local optima and increasing the probability of converging to the global optimum; user Channel updates are performed according to the following rules:

[0046] in These are learning parameters. and Representing users respectively In the and The function obtained in one estimation period .

[0047] The utility function obtained by each user in the iteration is also expressed by the formula. Make an estimate.

[0048] Beneficial effects: This invention proposes a non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication. Based on non-orthogonal channel research, it studies a probabilistic line-of-sight (LOS) communication model, comprehensively considering the mapping relationship between user channel distance and physical distance while addressing path attenuation caused by LOS. Based on scenario modeling, this invention also proposes an improved synchronous MAX-logit learning algorithm to optimize spectrum access strategies. By setting heterogeneous learning rates and appropriate strategy maintenance probabilities, it achieves an optimal balance between iteration speed and system performance. The proposed LOS communication model and improved MAX-logit learning algorithm enable efficient spectrum resource management, providing strong theoretical support for the design and optimization of air-to-ground communication networks. Attached Figure Description

[0049] Figure 1 This describes the relationship between Loss of Service (LoS) link probability and elevation angle under different environments. Figure 2 This relates the communication distance R to the probability of Loss of Station (LoS) links under different environments; Figure 3 This describes the relationship between communication distance R and elevation angle under different environments; Figure 4 This relates the probability of line-of-sight communication to the interference distance at different channel distances. Figure 5 This is a schematic diagram of five user sets divided according to channel distance; Figure 6 This is a schematic diagram of the non-orthogonal channel modeling and distributed decision-making method of the present invention; Figure 7 This is a graph showing the comparison results of model performance; Figure 8 This is a graph showing the algorithm convergence comparison results; Figure 9 This is a graph showing the comparison results of algorithm throughput performance. Detailed Implementation

[0050] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0051] This invention provides a non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication, such as... Figure 6 The diagram shown is a flowchart of the modeling and decision-making method of the present invention, and its technical solution includes the following steps: Step 1: Construct a probabilistic line-of-sight communication model for the air-to-ground communication network, and calculate the line-of-sight link probability based on the environmental parameters of the scene and the elevation angle of each node. .

[0052] In the air-to-ground communication model constructed in step 1 (which includes scenarios with air nodes and ground nodes), air-to-ground communication modeling is performed for cluster scenarios, and the probability of the Loss of Node (LoS) link in air-to-ground communication is as follows.

[0053]

[0054] in , These are environmental parameters; once the geographical environment is determined, they can be considered constants. It is the elevation angle formed between the transmitter and the receiver (in this embodiment of the invention, the transmitter and receiver refer to ground equipment and the UAV, respectively), and the calculation formula is as follows:

[0055] in Indicates the altitude of the drone. This indicates the 3D distance between ground equipment and drones. Figure 1 The probability of Loss links as a function of elevation angle under different environmental parameters is given. The change curve of environmental parameters. The values ​​are as follows: suburbs (4.88, 0.43), urban areas (9.61, 0.28), densely populated urban areas (12.08, 0.16), and high-rise building urban areas (27.23, 0.12).

[0056] Since only the Loss and Non-LoS cases are considered, the probability that the communication link is an NLoS link should be: .

[0057] Step 2: Calculate the weighted interference distance for each node by calculating the average path loss under line-of-sight link probability, and construct different channel spacings. The relationship between down-line-of-sight link probability and interference distance is discussed, and interference user sets are defined for each user. The optimization objective is to find the optimal channel selection combination from the set of interfering users so as to minimize the cumulative interference in the network.

[0058] In this embodiment of the invention, the weighted interference distance of each node is calculated by calculating the average path loss under the line-of-sight link probability: where the path loss (in dB) under LoS link and NLoS link is... , ,in This represents the three-dimensional distance between ground equipment and the drone; Indicates the frequency of electromagnetic waves; The speed of light; and These represent the additional path loss under LoS and NLoS links, respectively; the values ​​are determined by the environment. The values ​​are: suburbs (0.1,21), urban areas (1.0,20), densely populated urban areas (1.6,23), and high-rise building urban areas (2.3,34).

[0059] Based on the above analysis, and considering a statistical model based on link probability, the formula for calculating the average path loss of a wireless signal during propagation is as follows:

[0060] In an air-to-ground scenario, the signal for node communication Not lower than the threshold This allows communication connections to be maintained even under interference. . It is the signal-to-interference-plus-noise ratio in a communication link, calculated using the following formula: The path loss under non-interference conditions should satisfy the following condition: .Right now The formula for calculating the interference distance of a wireless signal during propagation is as follows:

[0061]

[0062] Figure 2 The communication distance under different environmental parameters is given. The curve of the probability of the Loss of Source (LoS) link. Figure 3 The communication distance under different environmental parameters is given. The probability of a Loss link varies with elevation angle. The curve showing the change.

[0063] The simulation parameters are shown in Table 1, and the environmental parameters are shown in Table 2.

[0064] Table 1 Simulation parameter settings

[0065] Table 2 Environmental parameters under four scenarios

[0066] Studies have shown that throughput normalization factor Physical distance exhibits a binary relationship: for a given specific physical distance, when the physical distance increases slightly, The value changes from 0.5 (at which point there is mutual interference between the two users) to 1 (at which point there is no interference between the two users).

[0067] Let the channel numbers selected by the two users be respectively and Then the channel distance between them can be expressed as

[0068] For example, if the two users' channels are channel 1 and channel 4, then To describe the binary interference model, we define... The channel distance is Throughput Normalization Factor The distance range is 0.5. Specifically, when the distance between two users is less than... At that time, their throughput normalization factor It is 0.5; when the distance between them is greater than At that time, throughput normalization factor Keep it at 1.

[0069] For a given transmission rate, let the effective transmission distance be denoted as . For a specific transmission rate, the interference distance corresponding to different channel spacings As shown in Table 3. Taking a transmission rate of 2 Mb / s as an example, (i) when When two users select the same channel, the interference distance is twice the transmission distance. (ii) when When, that is, when two users choose orthogonal channels, the interference distance is 0, meaning that no matter how far apart they are, they will not interfere with each other. (iii) When As the value increases from 0 to 5, the corresponding interference distance decreases monotonically. For other transmission rates, such as 5.5 Mb / s and 11 Mb / s, the interference distance exhibits the same trend, although the specific values ​​differ slightly.

[0070] Based on Table 3, the relationship between line-of-sight communication probability and interference distance at different channel distances in urban scenarios can be obtained as follows: Figure 4 As shown.

[0071] Table 3 Relationship between channel distance and interference distance

[0072] The normalized throughput of a user can be represented as ,in It's interference:

[0073] in This refers to the physical distance between the two users. The physical meaning of this formula is as follows: (i) When the distance between the two users is less than the interference distance, they can hear each other's transmitted signals. According to the distributed coordination mechanism of 802.11DCF, only one user can transmit at a time. Therefore, on average, each user only gets half the throughput. (ii) When the distance between the two users is greater than the interference distance, they can transmit data simultaneously.

[0074] In the probabilistic line-of-sight communication model, the user and At elevation angle and The communication distance is and .

[0075] User records and The physical distance between them is The channel distance is Therefore, the interference between them is :

[0076] in

[0077] Similarly, users The normalized throughput that can be obtained is ,in User Accumulated MAC layer interference.

[0078] Let the user set be ,Right now For specific users The interference it experiences is determined by its own channel selection and the channel selection of other neighboring users. For ease of explanation, we define channel distance. The user set in the corresponding interference area is ,Right now ,in The values ​​are shown in Table 3; in addition, the definition is in the interference region. But in the interference area users outside of this group ,Right now Similarly, it can be defined , and It can be noted that... This represents a user who might interfere with the user's experience. The channel set allocation described above is determined by the network topology and is independent of the user's channel selection. Figure 5A schematic diagram of the five user sets mentioned above is provided.

[0079] Let the channel set be ,Right now Assuming the user choose As a communication channel, users are recorded. and The channel distance between them is .remember For users Interference with users The number of users, that is:

[0080] in The following are indicator functions:

[0081] In other words, if the user and The channel distance between them is no greater than ,but The value is 1 otherwise, 0. Therefore, when the channel selection combination for all users is... At that time, the user The normalized throughput that can be obtained is

[0082] in It is a user set China The number of users who caused the interference, i.e.

[0083] The optimization objective of this invention is to find the optimal combination of channel selections that minimizes cumulative interference in the network, i.e.

[0084] The model involves communication between nodes within an air-to-ground area, and inter-channel interference.

[0085] Step 3: Initialize the channel selection for each node in the network and set the number of iterations. Calculate the utility function in the current iteration. ; In this embodiment of the invention, the channel selection of each node in the network is initialized, that is, each node in the network selects a channel from the channel set. The system randomly selects a channel for communication and saves the selection of each node as the number of iterations. The initial state under which each cognitive user... From its available channel set A channel is randomly selected with moderate probability. Assume each iteration contains A series of consecutive time slots, in which Representing the During the next iteration, the user The number of times the user successfully accessed the channel. So, the user... The achievable throughput can be expressed as Furthermore, users The value affected by the interference is Therefore, the utility function obtained by the user in each iteration can be estimated using the following formula:

[0086] Step 4, in the (i+1)th iteration cycle, the network explores and updates the channel selection, and the nodes... According to probability Perform channel selection updates and calculate the utility function in this iteration. ; In this embodiment of the invention, exploratory updates are performed, that is, in the (i+1)th iteration cycle, each node in the network explores channel selection. According to probability Perform channel selection update.

[0087]

[0088] in Representative set The cardinality is the number of elements in a set.

[0089] To achieve algorithm convergence, the probability of each user exploring new strategies should gradually decrease with each iteration until all users stop probing and the system reaches a stable state. To improve the convergence speed during the probing phase, we designed a dynamically adjusted probing probability: in the initial stage, the probability of maintaining a strategy is set low to ensure a more comprehensive exploration of possible strategies; as iterations progress, many users have found relatively good strategies, so the exploration rate gradually decreases, and users are more inclined to maintain their current strategy. In the mid-to-late stages, the probability of maintaining a strategy increases rapidly, eventually approaching 1.

[0090] To better meet the iterative requirements of the model in this invention, this invention proposes a good strategy to maintain the probability. The expression is

[0091] in , , , This represents the number of iterations.

[0092] The convergence speed of the algorithm can be improved by setting different learning parameters for different users. The number of neighboring users is So, for those with smaller Users with high learning values ​​should set larger learning parameters, while those with higher values... For users who value this feature, smaller learning parameters should be set.

[0093] It is clear from the expression for the policy retention probability that in the initial stage of the algorithm (number of iterations)... (Smaller) probability of retention Mainly affected by parameters The impact is low, and the probability of maintaining the connection is low, meaning users are more willing to explore new strategies, which in this model means accessing different channels; while the probability of maintaining the connection is low at the end of the algorithm. Mainly affected by parameters The influence is high, and the probability of maintaining the existing strategy is high, meaning there's a greater willingness to stick to the current strategy, which in this model means accessing a fixed channel. Based on this, the utility function in this iteration is calculated. .

[0094] Step 5, compare nodes exist If the channel selection in the current iteration period is the same as that in the previous iteration period, return to step 4; otherwise, proceed to step 6. Step 6, in the (i+2)th iteration cycle, the network learns and updates the channel selection, and the nodes... According to probability Perform channel selection updates and calculate the utility function in this iteration. ; In this embodiment of the invention, learning is performed based on the utility function obtained in the first two iterations. That is, the channel selection is updated using Boltzmann probing based on the results of the two explorations, which can avoid getting trapped in local optima and increase the probability of converging to the global optimum. Channel update is performed according to probability P2:

[0095] in These are learning parameters. and Representing users respectively In the and The function obtained in one estimation period .

[0096] The utility function obtained by each user in the iteration is also expressed by the formula. Make an estimate.

[0097] Step 7: Assign i+2 to i, and repeat steps 4-6 until the entire network strategy converges.

[0098] Example 1 A specific embodiment of the present invention is described below. The simulation was performed using Matlab software, and the parameter settings do not affect the generality. The simulation is based on IEEE 802.11b, which has 11 partially overlapping channels available for transmission. Each channel has a transmission rate of 2 Mb / s, and the users are distributed in a 1000m × 1000m area at an altitude of 50m. The parameters of the probabilistic line-of-sight communication model are consistent with those in Table 1, and the scenario is a densely populated urban area. =12.08, =0.16, =1.6, =23, so the effective communication distance when the line-of-sight communication probability is 1 is approximately 100m. Based on this, referring to the conclusions in Table 3, the interference distance corresponding to different channel distances can be obtained.

[0099] This embodiment uses a random network as the simulation scenario. To verify the performance of the proposed algorithm, Spatial Adaptive Learning (SAP algorithm), Synchronous Log-Linear Learning, and Synchronous MAX-logit algorithm were selected as comparison algorithms. The SAP algorithm requires knowledge of other user information and only one user can update the policy at a time; although it has strong performance, it has significant limitations in real-world air-to-ground communication scenarios.

[0100] First, the model performance was compared using a random network of 80 users, and the simulation algorithm employed the learning algorithm proposed in this invention. The performance comparison results between the model proposed in this invention and the model that does not consider probabilistic line-of-sight communication are as follows: Figure 7 As shown in the figure, the model proposed in this invention exhibits lower cumulative interference throughout the entire iteration process, and the cumulative interference at convergence is also significantly lower. Therefore, it can be seen that the model proposed in this invention can better utilize channel resources.

[0101] This embodiment verifies the convergence performance of the proposed synchronous logarithmic linear learning algorithm. Through multiple parameter adjustments, a suitable parameter setting was finally obtained: , , The convergence performance results of the simulation experiment are as follows: Figure 8 As shown in the figure, the simulation results are obtained by averaging 20 independent experiments. In this case, the algorithm can simultaneously achieve faster convergence and a lower convergence value. Figure 8As can be seen, the algorithm proposed in this invention, as a synchronous uncoupled algorithm, has convergence performance very close to that of the SAP algorithm, achieving convergence in approximately 200 iterations, while the other two synchronous algorithms require approximately 400 iterations. This is because the heterogeneous learning parameters fully utilize the spatial neighbor relationships of users, and the selected probability preservation function is more suitable for model convergence.

[0102] The throughput performance of the algorithm under different network densities was compared by simulating a random network consisting of 40-130 users. Figure 9 The simulation results of the four algorithms clearly show that when the network density is low (less than 70 nodes), the performance difference between the four algorithms is not significant. However, as the network density increases, the performance of the proposed algorithm is slightly inferior to the SAP algorithm. Compared with the other two synchronization algorithms, the improved synchronous linear learning proposed in this invention has certain performance advantages.

[0103] Based on the study of non-orthogonal channels, this invention investigates a probabilistic line-of-sight (LOS) communication model. It comprehensively considers the mapping relationship between user channel distance and physical distance, addressing the path attenuation inherent in LOS communication. Based on scenario modeling, this invention also proposes an improved synchronous MAX-logit learning algorithm to optimize spectrum access strategies. By setting heterogeneous learning rates and appropriate strategy maintenance probabilities, it achieves an optimal balance between iteration speed and system performance. Simulation results demonstrate that the proposed model and algorithm can maximize the spectrum utilization of the communication system, reduce interference, and improve the average network throughput.

[0104] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication, characterized in that, Includes the following steps: Step 1: Construct a probabilistic line-of-sight communication model for the air-to-ground communication network, and calculate the line-of-sight link probability based on the environmental parameters of the scene and the elevation angle of each node. ; Step 2: Calculate the weighted interference distance for each node by calculating the average path loss under line-of-sight link probability, and construct different channel spacings. The relationship between down-line-of-sight link probability and interference distance is discussed, and interference user sets are defined for each user. The optimization objective is to find the optimal combination of channel selections from the set of interfering users to minimize the cumulative interference in the network. Step 3: Initialize the channel selection for each node in the network and set the number of iterations. Calculate the utility function in the current iteration. ; Step 4, in the (i+1)th iteration cycle, the network explores and updates the channel selection, and the nodes... According to probability Perform channel selection updates and calculate the utility function in this iteration. ; Step 5, compare nodes exist If the channel selection in the current iteration period is the same as that in the previous iteration period, return to step 4; otherwise, proceed to step 6. Step 6, in the (i+2)th iteration cycle, the network learns and updates the channel selection, and the nodes... According to probability Perform channel selection updates and calculate the utility function in this iteration. ; Step 7: Assign i+2 to i, and repeat steps 4-6 until the entire network strategy converges.

2. The non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication as described in claim 1, characterized in that, In the air-to-ground communication model constructed in step 1, the probability that the communication link is a Loss of Space (LoS) link is... ;in: , These are environmental parameters; It is the elevation angle formed between the ground equipment and the drone; The probability that the communication link is an NLoS link should be: .

3. The non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication as described in claim 2, characterized in that, The calculation formula is as follows: in: Indicates the altitude of the drone. This indicates the 3D distance between ground equipment and drones.

4. The non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication as described in claim 1, characterized in that, The dynamic interference distance model in step 2 calculates the weighted interference distance of each node by calculating the average path loss under the line-of-sight link probability. The path loss for LoS and NLoS links is: The unit is dB. This represents the three-dimensional distance between ground equipment and the drone. Indicates the frequency of electromagnetic waves. At the speed of light, and These represent the additional path loss under LoS and NLoS links, respectively, which is determined by the environment. Considering a statistical model based on link probability, the formula for calculating the average path loss of a wireless signal during propagation is as follows: In an air-to-ground scenario, the signal for node communication Not lower than the threshold ; It is the signal-to-interference-plus-noise ratio in a communication link, calculated using the following formula: The path loss under non-interference conditions should satisfy the following condition: ,Right now The formula for calculating the interference distance of a wireless signal during propagation is as follows: Let the channel numbers selected by the two users be respectively and ,So and The channel distance between them is ; definition The channel distance is Throughput Normalization Factor The distance range is 0.5, when the distance between two users is less than When, the throughput normalization factor of both is... It is 0.5; when the distance between two users is greater than 0.

5. At that time, throughput normalization factor Keep it at 1; For a given transmission rate, let the effective transmission distance be denoted as . For a specific transmission rate, establish the interference distance corresponding to different channel intervals. ; The relationship between line-of-sight communication probability and interference distance at different channel distances; The normalized throughput of a certain user is expressed as: ,in It's interference: in It is the physical distance between the two users; In the probabilistic line-of-sight communication model, the user At elevation angle The communication distance is ;user At elevation angle The communication distance is ;user and The physical distance between them is The channel distance is So, users and Interference between them : in ; Similarly, users The normalized throughput that can be obtained is ,in User Accumulated MAC layer interference; The user set is ,Right now For specific users The interference it experiences is determined by its own channel selection and the channel selection of other neighboring users. Channel set is ,Right now Assuming the user choose As a communication channel, users are recorded. and The channel distance between them is ;remember For users Interference with users The number of users, that is: in For indicator functions: That is, if the user and The channel distance between them is no greater than ,but The value is 1 otherwise; when all users' channel selection combinations are... At that time, the user The achievable normalized throughput is: in It is a user set China The number of users who caused the interference, i.e. The optimization objective is to find the optimal combination of channel selections that minimizes the cumulative interference in the network, i.e. 。 5. The non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication as described in claim 1, characterized in that, Step 3 initializes the channel selection for each node in the network, that is, each node in the network selects a channel from the channel set. The system randomly selects a channel for communication and saves the selection of each node as the number of iterations. The initial state under the following conditions; each cognitive user From its available channel set A channel is randomly selected with moderate probability. ; Each iteration contains A series of consecutive time slots, in which Representing the During the next iteration, the user Number of successful channel accesses; User The achievable throughput is Furthermore, users The value affected by the interference is ; The utility function obtained by the user in each iteration is: 。 6. The non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication as described in claim 1, characterized in that, Step 4 involves exploratory updates, meaning that in the (i+1)th iteration cycle, each node in the network explores channel selection. According to probability Perform channel selection update: in Representative set The cardinality is the number of elements in a set. Set the strategy to maintain probability for: in , , , This represents the number of iterations. The convergence speed of the algorithm can be improved by setting different learning parameters for different users; The number of neighboring users is ; The utility function in the current iteration is: 。 7. The non-orthogonal channel modeling and distributed decision-making method for air-to-ground communication as described in claim 1, characterized in that, Step 6 involves learning based on the utility function obtained in the first two iterations. Specifically, it updates the channel selection using Boltzmann probing based on the results of the two explorations, avoiding local optima and increasing the probability of convergence to the global optimum. Channel update is performed according to probability P2: in These are learning parameters. and Representing users respectively In the and The function obtained in one estimation period ; The utility function obtained by each user in the iteration is also expressed by the formula. Make an estimate.

Images