An unmanned aerial vehicle cluster ad hoc network method based on internet of things driving

Abstract

The application discloses an unmanned aerial vehicle cluster ad hoc network method based on an Internet of Things driving, relates to the wireless network technical field and comprises the following steps: dividing an unmanned aerial vehicle cluster into multiple subclusters and selecting candidate computing nodes in the subclusters; collecting real-time state data by the unmanned aerial vehicle and selecting a temporary relay node; the temporary relay node determines a proportion of tasks allocated to the candidate computing nodes; an ad hoc network model is deployed in the candidate computing nodes; the candidate computing nodes perform distributed computing by using the ad hoc network model according to the proportion of tasks allocated, and obtain a local optimal solution; and the temporary relay node collects the local optimal solution and selects a global optimal solution. The application simultaneously considers the conditions of the unmanned aerial vehicle and the relay base station, determines the best relay node and the relay base station according to the environmental threat level of the unmanned aerial vehicle and data such as the blocking coefficient, distance, angle and signal strength between the unmanned aerial vehicle and the relay base station, and improves the stability of data transmission after the ad hoc network by the global consideration mode.

Description

Technical Field

[0001] This application relates to the field of wireless network technology, and in particular to a method for self-organizing a drone swarm using Internet of Things (IoT) technology. Background Technology

[0002] A swarm of drones consists of multiple drones. By installing payloads with different functions on different drones, the appropriate drones can perform tasks according to the requirements of the mission. Moreover, a swarm of multiple drones can also improve the success rate of mission execution because redundancy is usually set in the swarm. When one drone fails, other drones with the same payload can quickly take over, avoiding the interruption of the entire mission.

[0003] During missions, drone swarms need to communicate with ground equipment in real time to receive ground control or send real-time data back to the ground. Given the large number of drones in the swarm, one or a few drones typically act as relay nodes. Other drones send data to the relay node, which then relays it to the ground equipment, reducing network congestion caused by simultaneous data transmission from multiple devices. However, existing methods for selecting relay nodes are rather simplistic, only considering the drone's remaining battery power and computing power, neglecting the condition of the base stations communicating with the relay nodes. This leads to communication disruptions when drones transmit data back to the ground due to poor base station conditions. Summary of the Invention

[0004] This application provides an IoT-driven method for self-organizing drone swarms to address the problem of poor communication performance caused by ignoring ground base stations when selecting relay nodes in existing technologies.

[0005] This application provides an embodiment of an IoT-driven method for self-organizing unmanned aerial vehicle (UAV) swarms, including: The drone swarm is divided into multiple sub-clusters, and multiple drones are selected as candidate computing nodes in the sub-clusters. After the drone swarm is started, the drones in each sub-swarm collect real-time status data, and select a drone in each sub-swarm as a temporary relay node based on the real-time status data. The proportion of tasks to be allocated to each candidate computing node is determined by the temporary relay node based on the status of the candidate computing nodes in the same sub-cluster. A pruned ad hoc network model is deployed in all candidate computing nodes. The candidate computing nodes use the ad hoc network model to perform distributed computing according to the allocated proportion to obtain a local optimum. The local optimum includes the optimal fitness value and the corresponding node-base station combination. The optimal fitness value is the maximum fitness value between all UAVs in the sub-cluster and each relay base station on the ground. The node-base station combination is the combination of UAVs and relay base stations with the optimal fitness value. All temporary relay nodes summarize the local optimal solutions and select the global optimal solution from multiple local optimal solutions. The global optimal solution includes the best relay node. Other drones in the drone swarm send data to the best relay node. The best relay node forwards the data through the relay base stations and ground base stations in the global optimal solution, and finally sends it to the control terminal.

[0006] In one possible implementation, both the pre-pruning and post-pruning self-organizing network models include an input layer, a preprocessing layer, a core computation layer, and an output layer. Compared to the pre-pruning self-organizing network model, the post-pruning self-organizing network model removes the regularization module from the preprocessing layer, and also changes the objective function module in the core computation layer from adjusting weights based on gradient descent to fixed weights.

[0007] In one possible implementation, after the drones in each sub-cluster collect real-time status data, the real-time status data is preprocessed and normalized, and a temporary relay node is selected in each sub-cluster based on the preprocessed and normalized real-time status data.

[0008] In one possible implementation, the real-time status data includes environmental threat level, signal strength fitness, remaining battery power fitness, and remaining computing power fitness. The initial fitness value is calculated according to the following formula:

[0009] in, F temp ( i ) is the first i The initial fitness value of each drone, T i Environmental threat level, For the first i The mean of the signal strength compatibility between the drone and all candidate relay base stations. C i To adapt to the remaining computing power E i Adjust for remaining battery capacity; The drone with the highest initial fitness value in the sub-cluster is used as a temporary relay node.

[0010] In one possible implementation, the proportion of tasks assigned to each candidate computing node is calculated according to the following formula:

[0011] in, P task,i As a proportion, α , β and γ These are the remaining computing power weight, the inter-node signal strength weight, and the remaining power weight, respectively. C i To adapt to the remaining computing power S i,temp For the first i Signal strength compatibility between individual drones and temporary relay nodes E i To adapt to the remaining battery level, K This is the set of candidate computing nodes in the sub-cluster.

[0012] In one possible implementation, the objective function in the ad hoc network model is expressed as:

[0013] in, F node ( i , j ) is the first i The drone and the first j Fitness values ​​among relay base stations W j For the partition weight of the relay base station, T i Environmental threat level, O i,j The occlusion coefficient is... D i,j For distance adaptation, A i,j For angle adaptation, R i,j For signal strength adaptation; The highest fitness value is taken as the optimal fitness value.

[0014] In one possible implementation, an evolutionary algorithm is used to solve the objective function and obtain the optimal fitness value.

[0015] In one possible implementation, the operation of switching to the best relay node is initiated when any of the following conditions are met: The fitness value between the current best relay node and relay base station has decreased by more than 0.1 relative to the best fitness value in the global optimal solution; The current best relay node has a remaining power adaptability of less than or equal to 0.2, or a remaining computing power adaptability of less than or equal to 0.2. The signal strength compatibility between the current best relay node and the relay base station is less than or equal to 0.2, or the obstruction coefficient is greater than or equal to 0.8. The drone swarm moved beyond the communication range of the relay base station.

[0016] In one possible implementation, the operation of switching the best relay node includes: The ratio is recalculated by the temporary relay node, and then the candidate computing nodes recalculate according to the recalculated ratio to obtain the local optimal solution. The new best relay node is determined based on the local optimal solution. The new optimal relay node establishes a connection with the new relay base station and completes a link self-test; The data that was not forwarded by the original best relay node is synchronized to the new best relay node through the short-range 4G link within the cluster; Data forwarding is initiated by the new best relay node.

[0017] In one possible implementation, the temporary relay node monitors the average remaining computing power fit of the candidate computing nodes, adjusts the pruning rate of the ad hoc network model according to the magnitude of the average remaining computing power fit, and re-prunes the pruned ad hoc network model according to the adjusted pruning rate.

[0018] The IoT-driven drone swarm self-organizing network method in this application has the following advantages: Simultaneously considering the situation of drones and relay base stations, the fitness value is calculated based on the environmental threat level of drones and data such as the obstruction coefficient, distance, angle, and signal strength between drones and relay base stations. This determines the optimal relay node and relay base station, and the stability of data transmission after self-organizing the network is improved through this global consideration. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating an IoT-driven drone swarm self-organizing network method provided in this application embodiment. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] Figure 1 A flowchart illustrating an IoT-driven self-organizing drone swarm network method provided in this application embodiment. This application embodiment provides an IoT-driven self-organizing drone swarm network method, including: S100 divides the drone swarm into multiple sub-clusters and selects multiple drones as candidate computing nodes within each sub-cluster.

[0023] For example, after the drone swarm is started, it is automatically divided into several sub-clusters based on the initial geographical location and signal connectivity, with each sub-cluster having a size of N sub A scale of 5-10 aircraft is recommended; too large a scale can lead to communication congestion, while too small a scale cannot achieve distributed computing. Each sub-cluster exchanges its initial state, including remaining battery power and adaptability, via broadcast communication. E i Remaining computing power suitability C i And the initial position, filter out E i ≥0.5 and C i Drones with a value ≥0.4 are selected as candidate computing nodes, while the remaining drones are designated as data acquisition nodes, meaning they are only responsible for data acquisition and forwarding and do not participate in solving the objective function.

[0024] After determining the candidate computing nodes, it is also necessary to pre-store the parameters and algorithm parameters to provide the necessary data for the candidate computation. These parameters include local static parameters, evolutionary algorithm parameters, and threshold parameters, which will be explained below.

[0025] Local static parameters: initial location, zone affiliation, and communication radius of each relay base station. R max Signal transmission angle range; the drone's maximum computing power. C max Maximum battery capacity E max Sensor acquisition accuracy, etc.

[0026] Evolutionary algorithm parameters: population size P size=20-30, the specific value is adjusted according to the number of candidate computing nodes in the sub-cluster, each drone is assigned 3-5 individuals in the population, and the number of iterations. G =5~10, more iterations result in higher accuracy but also higher computational cost. This range is chosen after balancing. Crossover rate P c =0.6, variability P m =0.1, a fixed value to avoid the computational power consumption of adaptive adjustment.

[0027] Threshold parameter: Switching trigger threshold Δ F =0.1, triggering a reselection when the optimal fitness value in the global optimal solution drops below this threshold, parameter acquisition frequency. f =2Hz, local optimal solution interaction frequency f comm =0.5Hz.

[0028] S110: After the drone cluster is started, the drones in each sub-cluster collect real-time status data, and select a drone in each sub-cluster as a temporary relay node based on the real-time status data.

[0029] For example, real-time status data includes data from the drone itself, base station association parameters, and inter-node signal parameters. The data from the drone itself includes: the environmental threat level collected by each drone through its onboard sensors. T i Remaining battery capacity compatibility E i Remaining computing power suitability C i Acquisition frequency and parameter acquisition frequency f Consistent.

[0030] Base station association parameters include: candidate computing nodes scan all reachable relay base stations via 4G modules and collect data on each base station. j Occlusion coefficient O i,j Distance adaptability D i,j Angle adaptability A i,j Signal strength compatibility R i,j Only retain D i,j Relay base stations with a value of ≥0.3 are selected as candidate relay base stations, and those that are outside the communication range are eliminated to reduce the computational load.

[0031] Inter-node signal parameters include: candidate compute nodes and temporary relay nodes for each sub-cluster (the initial temporary relay node is within the sub-cluster). Ei and C i The optimal node is selected for signal interaction, and the signal strength compatibility between nodes is collected. S i,k This is used for subsequent task allocation adjustments.

[0032] After the drones in each sub-cluster collect real-time status data, the real-time status data is preprocessed and normalized. Based on the preprocessed and normalized real-time status data, a temporary relay node is selected in each sub-cluster.

[0033] Specifically, all collected raw real-time state data are filtered for outliers using the 3σ criterion to avoid solution bias caused by sensor noise, and then converted into standardized values ​​in the 0-1 interval using a linear normalization formula.

[0034] In the embodiments of this application, the real-time status data includes environmental threat level, signal strength adaptability, remaining power adaptability, and remaining computing power adaptability. The initial fitness value is calculated according to the following formula:

[0035] in, F temp ( i ) is the first i The initial fitness value of each drone, T i Environmental threat level, For the first i The mean of the signal strength compatibility between the drone and all candidate relay base stations. C i To adapt to the remaining computing power E i To determine the fitness level for remaining battery power, the drone with the highest initial fitness value in the sub-cluster is selected as a temporary relay node.

[0036] S120, the temporary relay node determines the proportion of tasks to be allocated to each candidate computing node based on the status of the candidate computing nodes in the same sub-cluster.

[0037] For example, the proportion of tasks assigned to each candidate computing node is calculated according to the following formula:

[0038] in, P task,i As a proportion, α , β and γ These are the remaining computing power weight, the inter-node signal strength weight, and the remaining power weight, respectively. Ci To adapt to the remaining computing power S i,temp For the first i Signal strength compatibility between individual drones and temporary relay nodes E i To adapt to the remaining battery level, K This is the set of candidate computing nodes in the sub-cluster.

[0039] The above formula also has the following constraints: P task,i ≤ C i ×0.8 indicates that the task proportion will not exceed 80% of the remaining computing power, so that computing power can be reserved for other necessary operations, and P task,i ≥0 indicates that tasks are assigned only to candidate computing nodes, representing the proportion of nodes that only collect data. P task,i =0.

[0040] Temporary relay nodes will allocate task proportions via short-range 4G communication. P task,i The corresponding population individuals are sent to each candidate computing node. After receiving the task, each candidate computing node sends back confirmation information. If no confirmation information is received, the temporary relay node resends the task after 0.1 seconds to ensure that the task is properly assigned.

[0041] S130, a pruned ad hoc network model is deployed in all candidate computing nodes. The candidate computing nodes use the ad hoc network model to perform distributed computing according to the allocated proportion to obtain a local optimal solution. The local optimal solution includes the optimal fitness value and the corresponding node-base station combination. The optimal fitness value is the maximum fitness value between all UAVs in the sub-cluster and each relay base station on the ground. The node-base station combination is the combination of UAVs and relay base stations with the optimal fitness value.

[0042] For example, the self-organizing network model completes pre-training at the control terminal, and can train the execution logic of the objective function and evolutionary algorithm based on historical environmental data and UAV state data, according to the pruning rate. Pr Initial pruning is performed at a value of 0.5, generating lightweight model files. The pruned model files are then distributed to candidate computing nodes via ground base stations, with the model file size controlled to within 100KB to adapt to the embedded storage and loading speed of UAVs.

[0043] After each candidate computing node loads the model, it automatically reads the pre-stored parameters to complete the initialization and establishes a local execution link for parameter acquisition, calculation, and result output.

[0044] In the embodiments of this application, both the pre-pruning and post-pruning self-organizing network models include an input layer, a preprocessing layer, a core computation layer, and an output layer. Compared to the pre-pruning self-organizing network model, the preprocessing layer of the post-pruning self-organizing network model removes the regularization module, and the objective function module in the core computation layer is changed from adjusting weights based on gradient descent to fixed weights.

[0045] Specifically, the objective function in the ad hoc network model is expressed as:

[0046] in, F node ( i , j ) is the first i The drone and the first j Fitness values ​​among relay base stations W j For the partition weight of the relay base station, T i Environmental threat level, O i,j The occlusion coefficient is... D i,j For distance adaptation, A i,j For angle adaptation, R i,j For signal strength fitness, the maximum fitness value is taken as the optimal fitness value.

[0047] In the embodiments of this application, an evolutionary algorithm is used to solve the objective function to obtain the optimal fitness value. The solution process is described in detail below.

[0048] (1) Population initialization.

[0049] Encoding method: Binary encoding is used, and the length of each chromosome is log2. N +log2 M log2 N The bit represents the number of the relay node, for example... N =8 requires 3 bits of binary encoding, followed by log2 M The bit represents the number of the relay base station, for example M When the value is 4, 2 bits of binary encoding are required; Distributed initialization: Each candidate computing node generates a corresponding local population according to the assigned task ratio, ensuring that the local population covers different node-base station combinations and avoiding homogenization of the initial population.

[0050] (2) Evolutionary iteration process.

[0051] The selection process uses a roulette wheel selection method, and the specific process is as follows: Each candidate computing node calculates the fitness value of each individual in the local population. F node ( i , j ); Calculate the total fitness of the local population. F sum ; The probability of each individual's choice P select ( p )= F node ( p ) / F sum The higher the probability, the greater the probability of being selected; Based on roulette rules P local Each individual acts as the parent population, and the crossover operation is performed. The selection operation only requires summation and division operations, resulting in low computational power consumption.

[0052] The crossover operation uses a single-point crossover, and the specific process is as follows: Randomly select an intersection point; For two randomly paired individuals in the parent population, the gene segments after the crossover point are exchanged to generate offspring individuals; The probability of crossover operation execution is P c =0.6, if crossover is not triggered, the parent individual is directly retained to simplify the calculation logic.

[0053] The mutation operation uses bit-flip mutation, and the specific process is as follows: For the offspring after crossover, each gene locus is... P m A bit flip (0→1 or 1→0) is performed with a probability of 0.1. Mutation operations target only a single gene locus, avoiding a sharp drop in fitness values ​​caused by large-scale gene changes, while maintaining population diversity.

[0054] After each iteration, each candidate computing node records the optimal fitness value in its local population. F local−best The corresponding node-base station combination is taken as the local optimal solution for this iteration.

[0055] Furthermore, after every two iterations, each candidate computing node sends its local optimal solution (including the optimal fitness value and the corresponding node-base station combination code) to the temporary relay node. After receiving all local optimal solutions, the temporary relay node selects the local optimal solution for the current sub-cluster. F sub-bestThis is then broadcast to all candidate computing nodes as a reference for the next iteration. Each candidate computing node can then... F sub-best The corresponding individuals are added to the local population to improve the convergence speed.

[0056] S140, all temporary relay nodes summarize the local optimal solutions and select the global optimal solution from multiple local optimal solutions. The global optimal solution includes the best relay node.

[0057] For example, the number of iterations reaches a preset value. G After reaching fitness values ​​of 5 to 10, each temporary relay node broadcasts its stored local optimum to all other temporary relay nodes. The temporary relay nodes then aggregate all local optima and calculate the global optimum fitness value. F global =max{ F local-best,1 , F local-best,2 ,..., F local-best,K’}, K’ This represents the number of temporary relay nodes, i.e., the number of sub-clusters. The corresponding node-base station combination is the optimal relay node and relay base station. If multiple node-base station combinations have a fitness value equal to... F global If the current partition belongs to a relay base station, the combination corresponding to that station will be selected first; if there are still multiple combinations, the combination with the higher remaining power of the nodes will be selected.

[0058] The temporary relay node broadcasts the global optimal solution to all drones in the cluster. After receiving confirmation, the best relay node initiates data forwarding preparation, which involves activating the 4G connection with the target relay base station and clearing the cache space. After confirmation by other drones, the collected data is sent to the best relay node, forming an ad hoc network link with the best relay node as the core.

[0059] S150: Other drones in the drone swarm send data to the best relay node. The best relay node forwards the data through the relay base stations and ground base stations in the global optimal solution, and finally sends it to the control terminal.

[0060] For example, the operation of switching the best relay node is initiated when any of the following conditions are met: The fitness value between the current best relay node and relay base station has decreased by more than 0.1 relative to the best fitness value in the global optimal solution; The current best relay node has a remaining power adaptability of less than or equal to 0.2, or a remaining computing power adaptability of less than or equal to 0.2. The signal strength compatibility between the current best relay node and the relay base station is less than or equal to 0.2, or the obstruction coefficient is greater than or equal to 0.8. The drone swarm moved beyond the communication range of the relay base station.

[0061] The steps to switch to the best relay node include: The ratio is recalculated by the temporary relay node, and then the candidate computing nodes recalculate according to the recalculated ratio to obtain the local optimal solution. The new best relay node is determined based on the local optimal solution. The new optimal relay node establishes a connection with the new relay base station and completes a link self-test; The data that was not forwarded by the original best relay node is synchronized to the new best relay node through the short-range 4G link within the cluster; Data forwarding is initiated by the new best relay node.

[0062] Temporary relay nodes monitor the average remaining computing power fit of candidate computing nodes and adjust the pruning rate of the ad hoc network model according to the magnitude of the average remaining computing power fit. The pruned ad hoc network model is then pruned again according to the adjusted pruning rate.

[0063] Specifically, the average remaining computing power of the candidate computing nodes is expressed as:

[0064] Adjust the pruning rate according to the following rules. Pr : when ≥0.6 (sufficient computing power): Pr =0.3, retaining more calculation modules and improving solution accuracy; When 0.3 < <0.6 (medium computing power): Pr =0.5, balancing accuracy and computing power consumption; when ≤0.3 (Computing power is limited): Pr =0.7, cut more redundant modules, and prioritize real-time performance.

[0065] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0066] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for self-organizing drone swarms based on the Internet of Things, characterized in that, include: The drone swarm is divided into multiple sub-clusters, and multiple drones are selected as candidate computing nodes in the sub-clusters. After the drone swarm is started, the drones in each sub-swarm collect real-time status data, and select a drone in each sub-swarm as a temporary relay node based on the real-time status data. The temporary relay node determines the proportion of tasks to be allocated to each of the candidate computing nodes based on the status of the candidate computing nodes in the same sub-cluster. A pruned ad hoc network model is deployed in all the candidate computing nodes. The candidate computing nodes perform distributed computing using the ad hoc network model according to the allocated proportion to obtain a local optimum. The local optimum includes an optimal fitness value and a corresponding node-base station combination. The optimal fitness value is the maximum fitness value between all UAVs in the sub-cluster and each relay base station on the ground. The node-base station combination is the combination of UAVs with the optimal fitness value and the relay base station. All the temporary relay nodes summarize the local optimal solutions and select the global optimal solution from the multiple local optimal solutions, wherein the global optimal solution includes the best relay node; Other drones in the drone swarm send data to the optimal relay node, which then forwards the data through the relay base station and ground base station in the global optimal solution before finally sending it to the control terminal. The real-time status data includes environmental threat level, signal strength adaptability, remaining power adaptability, and remaining computing power adaptability. The initial fitness value is calculated according to the following formula: in, For the first i The initial fitness value of each drone, The environmental threat level is defined as follows. For the first i The mean of the signal strength fit between the drone and all candidate relay base stations. For the remaining computing power adaptation, For the remaining battery power adaptation; The drone with the highest initial fitness value in the sub-cluster is designated as the temporary relay node. The proportion of tasks assigned to each candidate computing node is calculated according to the following formula: in, For the stated ratio, α , β and γ These are the remaining computing power weight, the inter-node signal strength weight, and the remaining power weight, respectively. S i,temp For the first i Signal strength compatibility between the drone and the temporary relay node K The set of candidate computing nodes in the sub-cluster.

2. The method for self-organizing unmanned aerial vehicle (UAV) swarms based on Internet of Things (IoT) as described in claim 1, characterized in that, The objective function module in the core computation layer of the self-organizing network model has been changed from adjusting weights based on gradient descent to using fixed weights.

3. The method for self-organizing unmanned aerial vehicle (UAV) swarms based on Internet of Things (IoT) as described in claim 1, characterized in that, After the drones in each sub-cluster collect the real-time status data, the real-time status data is preprocessed and normalized. Based on the preprocessed and normalized real-time status data, the temporary relay node is selected in each sub-cluster.

4. The method for self-organizing unmanned aerial vehicle (UAV) swarms based on Internet of Things (IoT) as described in claim 1, characterized in that, The objective function in the self-organizing network model is expressed as: in, F node ( i , j ) is the first i The drone and the first j The fitness values ​​among the relay base stations are as follows. W j The partition weight of the relay base station, T i Environmental threat level, O i,j The occlusion coefficient is... D i,j For distance adaptation, A i,j For angle adaptation, R i,j For signal strength adaptation; The largest fitness value is taken as the optimal fitness value.

5. The method for self-organizing unmanned aerial vehicle (UAV) swarms based on Internet of Things (IoT) according to claim 4, characterized in that, The objective function is solved using an evolutionary algorithm to obtain the optimal fitness value.

6. The method for self-organizing unmanned aerial vehicle (UAV) swarms based on Internet of Things (IoT) according to claim 1, characterized in that, The operation of switching the optimal relay node will be initiated when any of the following conditions are met: The fitness value between the current optimal relay node and the relay base station has decreased by more than 0.1 relative to the optimal fitness value in the global optimal solution; The current optimal relay node has a remaining power adaptability of less than or equal to 0.2, or a remaining computing power adaptability of less than or equal to 0.

2. The signal strength compatibility between the current optimal relay node and the relay base station is less than or equal to 0.2, or the obstruction coefficient is greater than or equal to 0.

8. The drone swarm moved beyond the communication range of the relay base station.

7. A method for self-organizing unmanned aerial vehicle (UAV) swarms based on Internet of Things (IoT) according to claim 6, characterized in that, The operation of switching the optimal relay node includes: The ratio is recalculated by the temporary relay node, and then the candidate computing node recalculates the local optimal solution according to the recalculated ratio. A new optimal relay node is determined based on the local optimal solution. The new optimal relay node establishes a connection with the new relay base station and completes a link self-test; The original best relay node will synchronize any unforwarded data to the new best relay node via a short-range 4G link within the cluster; Data forwarding is initiated by the new optimal relay node.

8. The method for self-organizing unmanned aerial vehicle (UAV) swarms based on Internet of Things (IoT) according to claim 1, characterized in that, The temporary relay node monitors the average remaining computing power suitability of the candidate computing nodes, adjusts the pruning rate of the self-organizing network model according to the magnitude of the average remaining computing power suitability, and re-prunes the pruned self-organizing network model according to the adjusted pruning rate.

Images