Unmanned aerial vehicle cluster communication system based on weak signal connection

Abstract

The invention discloses an unmanned aerial vehicle cluster communication system based on weak signal connection. According to the invention, through the intelligent cooperation mechanism of the dynamic relay path optimization module, the communication reliability in a complex environment is significantly improved. The system analyzes cluster network topology and node states in real time, combines multi-dimensional parameters such as signal stability, energy consumption efficiency and task load, dynamically screens an optimal relay node and generates an efficient transmission path. For example, in a weak signal area, the system can automatically start an adjacent high-stability node as a relay, the problem of rigidity of a traditional fixed relay strategy is avoided, meanwhile, the signal attenuation risk is avoided in advance through predictive switching, and continuous return of key data is ensured. The capability of actively adapting to network fluctuation effectively reduces the probability of communication interruption, and can ensure the non-inductive connection of task instructions and key information especially in scenes with strict requirements on real-time performance, such as search and rescue, exploration and the like.

Description

Technical Field

[0001] The present invention belongs to the field of UAV communication technology, and specifically relates to a UAV cluster communication system based on weak signal connection. Background Art

[0002] The UAV communication system is the core technology system that ensures information exchange between UAVs and ground control stations or other platforms. It undertakes key tasks such as command transmission, status monitoring, real-time return of telemetry data and images / videos, and exchange of navigation and positioning information. The system usually includes an airborne communication payload, a ground communication terminal, and the corresponding network protocols and signal processing technologies. It supports multiple communication links, such as microwave and millimeter wave line-of-sight (LOS) and satellite communication beyond line of sight (BLOS). Depending on the mission requirements, broadband data links, narrowband data links, or dedicated links with strong anti-interference capabilities can be selected, and performance indicators such as transmission distance, bandwidth capacity, anti-interference, low detectability, and link reliability must be considered. It is the basic guarantee for UAVs to complete diverse tasks such as reconnaissance, mapping, logistics, inspection, and emergency communications.

[0003] However, traditional systems rely on fixed backup nodes or centralized host scheduling and are unable to dynamically adjust paths based on real-time network status, resulting in easy interruption of communication links and low recovery efficiency under weak signals. At the same time, resource allocation only focuses on a single indicator (such as signal strength or number of hops), ignoring the coordinated optimization of node energy consumption balance and task load, resulting in problems such as delays in high-priority tasks and premature exhaustion of nodes, which seriously limit the cluster's ability to continue operating in complex environments. Summary of the Invention

[0004] The purpose of the present invention is to provide a UAV cluster communication system based on weak signal connection in order to solve the above-mentioned problems.

[0005] The technical solution adopted by the present invention is as follows: a UAV cluster communication system based on weak signal connection, the system comprising: a signal strength monitoring and classification module, a task information priority management module, an adaptive communication scheduling module, a dynamic relay path optimization module and a host data processing and feedback module;

[0006] The dynamic relay path optimization module is internally configured with: a communication topology perception submodule, a relay node evaluation submodule, a reinforcement learning path decision submodule, a predictive relay switching submodule, and a relay link execution and monitoring submodule;

[0007] The output end of the signal strength monitoring and classification module is directly connected to the input end of the task information priority management module;

[0008] The output end of the task information priority management module transmits the classified data packets to the input end of the adaptive communication scheduling module;

[0009] The output end of the adaptive communication scheduling module transmits the compressed and encoded data to the input end of the dynamic relay path optimization module;

[0010] The output end of the dynamic relay path optimization module transmits the path instruction and data packet to the input end of the host data processing and feedback module, and the host completes the global data fusion and task logic processing;

[0011] The output end of the host data processing and feedback module transmits the decision results back to the control interface of the dynamic relay path optimization module through an independent feedback channel and a control instruction generator, forming a closed-loop parameter adjustment link, and at the same time directly sends bandwidth reallocation instructions to the adaptive communication scheduling module to adapt to real-time task requirements.

[0012] In a preferred embodiment, the signal strength monitoring and classification module is internally provided with:

[0013] Multi-band signal sensor and real-time analysis unit. The sensor continuously collects signal strength parameters between each slave and the host, including RSSI value, signal-to-noise ratio and packet transmission success rate. The analysis unit calculates the mean and fluctuation rate of signal strength through a sliding window algorithm, and dynamically divides the signal level based on the preset adaptive grading threshold.

[0014] In a preferred embodiment, the task information priority management module is provided with a three-level data classification architecture and a dynamic screening engine. The first-level information is defined as identity identification, real-time coordinates and emergency control instructions, the second-level information covers motion state parameters such as speed, heading and ambient wind speed, and the third-level information includes sensor raw data such as IMU readings and attitude angles.

[0015] In a preferred embodiment, the adaptive communication scheduling module is provided with a bandwidth adaptive algorithm and a coding strategy library. The algorithm dynamically allocates transmission resources according to signal strength and data priority. The coding strategy library pre-stores multiple modulation schemes, including QPSK, 16-QAM and 64-QAM. Every time the signal strength increases by one level, it gradually switches to high-order modulation to improve throughput; the module has a built-in feedback mechanism to monitor the packet loss rate and delay in real time. If the packet loss in the weak signal scenario exceeds the threshold, the redundant transmission mode is triggered, and the first-level information is sent through dual-path backup.

[0016] In a preferred embodiment, the communication topology perception submodule is composed of a data fusion unit, a task logic engine and a control instruction generator. The data fusion unit receives discrete data packets uploaded by each sub-machine, reconstructs the global task view through timestamp alignment and spatial interpolation algorithms, and the task logic engine analyzes the task progress based on preset rules, automatically calculates the corrected path when each sub-machine deviates from the predetermined track, or initiates a cluster reorganization instruction after identifying a weak signal area; the control instruction generator encodes the decision result into a standardized instruction set, sends it to each sub-machine through an independent feedback channel, and generates a visual report and sends it back to the control background.

[0017] In a preferred embodiment, the relay node evaluation submodule includes real-time collection of signal strength, energy consumption status, and task load data of each slave in the cluster, evaluating signal stability by analyzing historical signal fluctuation rate and current RSSI average, measuring energy efficiency by combining remaining power and unit data transmission energy consumption, and determining the task load level of each slave based on whether it currently undertakes a critical task. Finally, based on the comprehensive score of signal stability, energy efficiency, and task load, a candidate node priority list is generated to screen out the optimal relay node.

[0018] The comprehensive scoring formula for relay nodes is:

[0019]

[0020] Where:

[0021] σ 2 Indicates the historical signal strength variance, reflecting signal volatility. The smaller the variance, the higher the stability.

[0022] μRSSI represents the current mean signal strength in dBm, which directly reflects the link quality.

[0023] α represents the signal stability weight coefficient, which is adjusted according to the real-time requirements of the task;

[0024] Q represents the percentage of remaining power of the node (0% to 100%);

[0025] E represents the energy consumption per unit of data transmission (e.g., the amount of electricity consumed per MB of data);

[0026] β represents the energy efficiency weight coefficient, which is increased when focusing on energy saving goals;

[0027] L represents the task load level, ranging from 1 to 3 (low / medium / high). When the load is high, the node will not be given priority to be selected as a relay; γ represents the load penalty coefficient, which prevents high-load nodes from excessively participating in relaying.

[0028] In a preferred embodiment, the reinforcement learning path decision submodule includes receiving dynamic network structure data generated by the communication topology perception submodule in real time and the candidate node priority table provided by the relay node evaluation submodule, constructing a state space through a deep reinforcement learning algorithm, where the state is defined as the position, signal strength, remaining power and task load level of the current cluster node, and the action space is to select the next hop relay node or directly connect to the host. The reward function integrates multi-objective optimization indicators, and the experience replay mechanism and target network separation technology are used in the training process to improve the convergence efficiency. Finally, the optimal relay path strategy based on the real-time environment is output to ensure the efficiency and robustness of data transmission under weak signals.

[0029] The calculation formula of the multi-objective fusion reward function is:

[0030]

[0031] Where:

[0032] ΔH: The change in the number of hops in a path, defined as the difference between the current path hop count and the historical optimal hop count. As the number of hops decreases, the exponential term increases, directly incentivizing the algorithm to select paths with fewer hops.

[0033] σ: Link signal strength variance, reflecting the signal volatility between nodes in the path. The smaller the variance, the more stable the link. The positive reward for a stable link is amplified by subtracting the variance from 1.

[0034] E_avg: The average specific energy consumption of all nodes in the path, in milliampere hours per megabyte (mAh / MB). The lower the energy consumption, the smaller the penalty value.

[0035] Q_max: The maximum remaining power percentage of the nodes in the path (range 0 to 100). The more power there is, the smaller the impact of the penalty term.

[0036] λ1, λ2, and λ3 are dynamic weight coefficients that are adjusted in real time based on the task type. In urgent tasks, λ1 is increased to prioritize reducing the number of hops, while in persistent tasks, λ2 and λ3 are increased to optimize energy consumption and stability.

[0037] In a preferred embodiment, the predictive relay switching submodule includes real-time collection of current relay link signal strength data and historical attenuation trends, analysis of signal strength change patterns through a time series model, establishment of a prediction curve based on signal attenuation rate, and dynamic calculation of the confidence interval of the signal quality within a certain time window in the future in combination with the location, remaining power, and stability parameters of adjacent nodes. If the predicted signal strength is lower than a preset threshold and the confidence level exceeds a set threshold, the relay migration process is triggered, and nodes with stable signals, balanced energy consumption, and low task load are selected from the pre-loaded backup node list as migration targets. At the same time, the old and new relay nodes are coordinated to complete the communication protocol switching and data cache synchronization to ensure that the link switching process is imperceptible and there is zero packet loss.

[0038] The migration triggering condition formula of the predictive relay switching submodule is:

[0039] MigrateFlag=μ·ΔS predicted +ν·(1-C node )+ξ·E ratio ;

[0040] Where:

[0041] ΔS_predicted: Predicted signal strength attenuation. The absolute value of the signal attenuation in the future time window is calculated using the time series model. The unit is dBm.

[0042] C_node: The comprehensive stability score of the candidate backup node, which is calculated by weighting the node's historical online rate (percentage) and the current signal volatility (variance), with a score range of 0 to 1;

[0043] E_ratio: The energy consumption difference ratio between the new and old relay nodes, defined as the percentage of the new node's unit data transmission energy consumption to the original node's energy consumption, used to quantify the energy consumption optimization space after migration;

[0044] μ, ν, ξ: dynamic weight coefficients that adjust the priority according to the task type; in urgent tasks, increase μ to quickly respond to signal degradation, and in persistent tasks, increase ν and ξ to prioritize node stability and energy consumption balance.

[0045] In a preferred embodiment, the relay link execution and monitoring submodule includes real-time reception of the optimal path instructions generated by the dynamic relay path optimization module, and coordination of the communication protocol switching operations between the sub-machines in the cluster. The module continuously monitors the real-time performance indicators of the current relay link, including packet loss rate, signal strength fluctuation, transmission delay and node remaining power, and performs real-time detection of abnormal conditions by setting dynamic thresholds.

[0046] In a preferred embodiment, the host data processing and feedback module is composed of a data fusion unit, a task logic engine and a control instruction generator. The data fusion unit receives discrete data packets uploaded by each sub-machine, and reconstructs the global task view through timestamp alignment and spatial interpolation algorithms; the task logic engine analyzes the task progress based on preset rules, automatically calculates the corrected path when each sub-machine deviates from the predetermined track, or initiates a cluster reorganization instruction after identifying a weak signal area; the control instruction generator encodes the decision result into a standardized instruction set, sends it to each sub-machine through an independent feedback channel, and generates a visual report and sends it back to the control background; the module has a built-in fault-tolerant mechanism. If the received data verification fails, it immediately requests each sub-machine to retransmit and marks the abnormal node to ensure the robustness of the system.

[0047] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0048] 1. In the present invention, the intelligent collaborative mechanism of the dynamic relay path optimization module significantly improves the communication reliability in complex environments. This module analyzes the cluster network topology and node status in real time, combines multi-dimensional parameters such as signal stability, energy efficiency and task load, dynamically screens the optimal relay node and generates an efficient transmission path. For example, in weak signal areas, the system can automatically enable adjacent high-stability nodes as relays to avoid the rigidity of traditional fixed relay strategies, while at the same time avoiding the risk of signal attenuation in advance through predictive switching to ensure continuous return of critical data. This ability to actively adapt to network fluctuations effectively reduces the probability of communication interruption, especially in scenarios with strict real-time requirements such as search and rescue and exploration, and can ensure the seamless connection between mission instructions and key information.

[0049] 2. In the present invention, the overall efficiency of the cluster is greatly improved through the design of resource optimization and task adaptation. The dynamic relay path optimization module takes into account the node energy consumption balance and task load distribution while ensuring the communication quality, so as to avoid excessive consumption of a single node and shortening the life of the cluster. For example, in persistent monitoring tasks, the system gives priority to nodes with sufficient remaining power and low load as relays to extend the task cycle; while in emergency tasks, it focuses on low-hop and highly stable links to shorten the response time. This flexible resource scheduling strategy not only enhances the task fault tolerance capability in weak signal environments, but also provides accurate decision-making support for the control background through global data fusion and closed-loop feedback mechanism, ultimately realizing efficient collaboration and autonomous operation of drone clusters in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A block diagram of the subject system of the present invention;

[0051] Figure 2 This is a system block diagram of the dynamic relay path optimization module in the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0053] Example:

[0054] Reference Figure 1-2 , a UAV cluster communication system based on weak signal connection, the system includes: signal strength monitoring and classification module, task information priority management module, adaptive communication scheduling module, dynamic relay path optimization module and host data processing and feedback module;

[0055] The internal settings of the dynamic relay path optimization module include: communication topology perception submodule, relay node evaluation submodule, reinforcement learning path decision submodule, predictive relay switching submodule and relay link execution and monitoring submodule;

[0056] The output of the signal strength monitoring and classification module is directly connected to the input of the task information priority management module;

[0057] The output of the task information priority management module transmits the classified data packets to the input of the adaptive communication scheduling module;

[0058] The output end of the adaptive communication scheduling module transmits the compressed and encoded data to the input end of the dynamic relay path optimization module;

[0059] The output of the dynamic relay path optimization module transmits the path instructions and data packets to the input of the host data processing and feedback module, and the host completes the global data fusion and task logic processing;

[0060] The output end of the host data processing and feedback module transmits the decision results back to the control interface of the dynamic relay path optimization module through an independent feedback channel and a control instruction generator, forming a closed-loop parameter adjustment link. At the same time, it directly sends bandwidth reallocation instructions to the adaptive communication scheduling module to adapt to real-time task requirements.

[0061] The internal settings of the signal strength monitoring and classification module are:

[0062] The multi-band signal sensor and real-time analysis unit continuously collect the signal strength parameters between each sub-machine and the main machine, including RSSI value, signal-to-noise ratio and packet transmission success rate. The analysis unit calculates the mean and fluctuation rate of the signal strength through a sliding window algorithm, and dynamically divides the signal level based on the preset adaptive grading threshold. For example, a weak signal is set as an RSSI lower than -80dBm and a fluctuation rate of more than 15%, a medium signal is -80dBm to -60dBm and a fluctuation rate of less than 10%, and a strong signal is an RSSI higher than -60dBm and a fluctuation rate of less than 5%. The grading result triggers the corresponding communication strategy. In weak signal conditions, only core data transmission is allowed. In medium and strong signals, the expanded data stream is gradually opened. At the same time, it supports customized thresholds for mission scenarios. For example, in search and rescue missions, weak signal judgment is relaxed to maintain basic connection.

[0063] The mission information priority management module is equipped with a three-level data classification architecture and a dynamic filtering engine. Level 1 information is defined as identity, real-time coordinates, and emergency control instructions. Level 2 information covers motion state parameters such as speed, heading, and ambient wind speed. Level 3 information contains raw sensor data such as IMU readings and attitude angles. The engine automatically activates the corresponding data filter based on the current signal level. For example, only level 1 information is retained in weak signals, and level 2 and level 3 data are appended in sequence for medium and strong signals. The filtered data packets are compressed using a lightweight encapsulation protocol, and a priority tag is embedded in the header to ensure that the host can quickly identify key content. In addition, the module supports dynamic adjustment of classification rules during tasks, such as raising the priority of heading data to level 1 in tracking tasks.

[0064] The adaptive communication scheduling module is equipped with a bandwidth adaptation algorithm and a coding strategy library. The algorithm dynamically allocates transmission resources based on signal strength and data priority. For example, under weak signals, the transmission frequency is reduced from 10Hz to 2Hz, and low-complexity coding such as Turbo code is enabled to reduce the bit error rate, while compressing non-critical data to less than 30% of the original volume. The coding strategy library pre-stores a variety of modulation schemes, including QPSK, 16-QAM and 64-QAM. With each level increase in signal strength, it gradually switches to higher-order modulation to increase throughput. The module has a built-in feedback mechanism to monitor packet loss rate and delay in real time. If the packet loss in a weak signal scenario exceeds the threshold, the redundant transmission mode is triggered, and the first-level information is sent through dual-path backup.

[0065] The communication topology perception submodule consists of a data fusion unit, a task logic engine, and a control instruction generator. The data fusion unit receives discrete data packets uploaded by each slave and reconstructs the global task view through timestamp alignment and spatial interpolation algorithms. For example, it integrates the location data of multiple slaves to generate a cluster distribution heat map. The task logic engine analyzes the task progress based on preset rules. For example, it automatically calculates the corrected path when detecting that each slave deviates from the planned track, or initiates cluster reorganization instructions after identifying weak signal areas. The control instruction generator encodes the decision results into a standardized instruction set, which is sent to each slave through an independent feedback channel. At the same time, it generates a visual report and sends it back to the control background. The module has a built-in fault-tolerant mechanism. If the received data verification fails, it immediately requests each slave to retransmit and marks the abnormal node to ensure the robustness of the system.

[0066] The relay node evaluation submodule collects real-time data on the signal strength, energy consumption, and task load of each slave within the cluster. It evaluates signal stability by analyzing historical signal fluctuations and the current RSSI average. It also measures energy efficiency by combining remaining battery life with unit data transmission energy consumption. It also determines the task load level of each slave based on whether it is currently undertaking a critical task. Finally, based on the combined scores of signal stability, energy efficiency, and task load, it generates a priority list of candidate nodes to select the optimal relay node.

[0067] The comprehensive scoring formula for relay nodes is:

[0068]

[0069] Where:

[0070] σ 2 It represents the historical signal strength variance, reflecting the signal volatility. The smaller the variance, the higher the stability.

[0071] μRSSI indicates the current mean signal strength in dBm, which directly reflects the link quality.

[0072] α represents the signal stability weight coefficient, which is adjusted according to the real-time requirements of the task (for example, \alphaα increases in urgent tasks).

[0073] Q represents the remaining power percentage of the node (0% to 100%).

[0074] E represents the energy consumption per unit data transmission (such as the amount of electricity consumed per MB of data).

[0075] β represents the energy efficiency weight coefficient, which is increased when focusing on energy saving goals.

[0076] L represents the task load level, ranging from 1 to 3 (low / medium / high). Nodes with high loads are not preferred for relay selection. γ represents the load penalty coefficient, which prevents high-load nodes from excessively participating in relays.

[0077] The reinforcement learning path decision submodule includes the real-time reception of dynamic network structure data generated by the communication topology perception submodule and the candidate node priority table provided by the relay node evaluation submodule. The state space is constructed through the deep reinforcement learning algorithm. The state is defined as the current cluster node's position, signal strength, remaining power and task load level. The action space is to select the next-hop relay node or directly connect to the host. The reward function integrates multi-objective optimization indicators, such as the minimum hop number reward, link stability reward and energy consumption balance penalty. The experience replay mechanism and target network separation technology are used in the training process to improve the convergence efficiency. Finally, the optimal relay path strategy based on the real-time environment is output to ensure the efficiency and robustness of data transmission under weak signals.

[0078] The calculation formula of the multi-objective fusion reward function is:

[0079]

[0080] Where:

[0081] ΔH: The change in path hop count, defined as the difference between the current path hop count and the historical optimal hop count. As the hop count decreases, the exponential term increases, directly incentivizing the algorithm to select the path with fewer hops.

[0082] σ: Link signal strength variance, reflecting the signal volatility between nodes in the path. The smaller the variance, the more stable the link. Subtracting the variance from 1 amplifies the positive reward for a stable link.

[0083] E_avg: The average specific energy consumption of all nodes in the path, in milliampere hours per megabyte (mAh / MB). The lower the energy consumption, the smaller the penalty value.

[0084] Q_max: The maximum remaining power percentage of the nodes in the path (range 0 to 100). The more sufficient the power, the smaller the impact of the penalty term.

[0085] λ1, λ2, and λ3 are dynamic weight coefficients that are adjusted in real time based on the task type. For example, in urgent tasks, λ1 is increased to prioritize reducing the number of hops, while in persistent tasks, λ2 and λ3 are increased to optimize energy consumption and stability.

[0086] The predictive relay switching submodule includes real-time collection of current relay link signal strength data and historical attenuation trends, analysis of signal strength change patterns through time series models, establishment of a prediction curve based on signal attenuation rate, and dynamic calculation of the confidence interval of signal quality within a certain time window in the future, combined with the location, remaining power and stability parameters of adjacent nodes. If the predicted signal strength is lower than the preset threshold and the confidence level exceeds the set threshold, the relay migration process is triggered, and nodes with stable signals, balanced energy consumption and low task load are selected from the pre-loaded backup node list as migration targets. At the same time, the new and old relay nodes are coordinated to complete the communication protocol switching and data cache synchronization to ensure that the link switching process is imperceptible and there is zero packet loss.

[0087] The migration triggering condition formula of the predictive relay switching submodule is:

[0088] MigrateFlag=μ·ΔS predicted +ν·(1-C node )+ξ·E ratio ;

[0089] Where:

[0090] ΔS_predicted: Predicts the signal strength attenuation. The absolute value of the signal attenuation in the future time window is calculated using a time series model. The unit is dBm.

[0091] C_node: The comprehensive stability score of the candidate backup node, which is calculated by weighting the node's historical online rate (percentage) and the current signal volatility (variance), with a score range of 0 to 1.

[0092] E_ratio: The energy consumption difference ratio between the new and old relay nodes, defined as the percentage of the new node's unit data transmission energy consumption to the original node's energy consumption, used to quantify the energy consumption optimization space after migration.

[0093] μ, ν, ξ: Dynamic weight coefficients that adjust the priority based on the task type. For example, in urgent tasks, μ is increased to quickly respond to signal degradation, while in persistent tasks, ν and ξ are increased to prioritize node stability and energy consumption balance.

[0094] The relay link execution and monitoring submodule receives optimal path instructions generated by the dynamic relay path optimization module in real time and coordinates communication protocol switching operations between slaves within the cluster, such as adjusting transmission frequency bands, enabling encrypted channels, or switching to redundant links, ensuring smooth and uninterrupted relay link switching. This module continuously monitors the real-time performance indicators of the current relay link, including packet loss rate, signal strength fluctuations, transmission delay, and node remaining battery life. It detects abnormal conditions in real time by setting dynamic thresholds. For example, if the packet loss rate exceeds 5% or the signal strength drops by 20dBm, an alarm mechanism is triggered, automatically initiating an emergency rerouting process, selecting a suboptimal link from a predefined backup path library and performing a rapid switchover. The module also records link operation logs, including switch time, node status, abnormal events, and handling results, providing data support for subsequent path optimization. During execution, the module collaborates with the predictive relay switching submodule to preload backup node resources and synchronizes global task status with the host data processing module. This creates a closed-loop control system from command issuance, link execution, and abnormality recovery, ensuring the continuity and reliability of UAV cluster communication in weak signal environments.

[0095] The host data processing and feedback module consists of a data fusion unit, a task logic engine, and a control instruction generator. The data fusion unit receives discrete data packets uploaded by each sub-machine, and reconstructs the global task view through timestamp alignment and spatial interpolation algorithms. For example, it integrates the position data of multiple sub-machines to generate a cluster distribution heat map. The task logic engine analyzes the task progress based on preset rules. For example, it automatically calculates the corrected path when detecting that each sub-machine deviates from the predetermined track, or initiates cluster reorganization instructions after identifying weak signal areas. The control instruction generator encodes the decision results into a standardized instruction set, sends it to the sub-machine through an independent feedback channel, and generates a visual report back to the control background. The module has a built-in fault-tolerant mechanism. If the received data verification fails, it immediately requests each sub-machine to retransmit and marks the abnormal node to ensure the robustness of the system.

[0096] From the above we can know:

[0097] In the present invention, the intelligent collaborative mechanism of the dynamic relay path optimization module significantly improves the communication reliability in complex environments. The module analyzes the cluster network topology and node status in real time, combines multi-dimensional parameters such as signal stability, energy efficiency and task load, dynamically screens the optimal relay node and generates an efficient transmission path. For example, in weak signal areas, the system can automatically enable adjacent high-stability nodes as relays to avoid the rigidity of traditional fixed relay strategies, and at the same time, through predictive switching, avoid the risk of signal attenuation in advance to ensure continuous return of critical data. This ability to actively adapt to network fluctuations effectively reduces the probability of communication interruption, especially in scenarios with strict real-time requirements such as search and rescue and exploration, and can ensure the seamless connection between mission instructions and key information.

[0098] In the present invention, the overall efficiency of the cluster is greatly improved through the design of resource optimization and task adaptation. The dynamic relay path optimization module takes into account the node energy consumption balance and task load distribution while ensuring the communication quality, so as to avoid excessive consumption of a single node and shortening the life of the cluster. For example, in persistent monitoring tasks, the system gives priority to nodes with sufficient remaining power and low load as relays to extend the task cycle; while in emergency tasks, it focuses on low-hop and highly stable links to shorten the response time. This flexible resource scheduling strategy not only enhances the task fault tolerance capability in weak signal environments, but also provides accurate decision-making support for the control background through global data fusion and closed-loop feedback mechanism, ultimately achieving efficient collaboration and autonomous operation of drone clusters in complex scenarios.

[0099] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further limitations, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0100] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A UAV cluster communication system based on weak signal connection, characterized by: The system includes: a signal strength monitoring and classification module, a task information priority management module, an adaptive communication scheduling module, a dynamic relay path optimization module and a host data processing and feedback module; The dynamic relay path optimization module is internally configured with: a communication topology perception submodule, a relay node evaluation submodule, a reinforcement learning path decision submodule, a predictive relay switching submodule, and a relay link execution and monitoring submodule; The output end of the signal strength monitoring and classification module is directly connected to the input end of the task information priority management module; The output end of the task information priority management module transmits the classified data packets to the input end of the adaptive communication scheduling module; The output end of the adaptive communication scheduling module transmits the compressed and encoded data to the input end of the dynamic relay path optimization module; The output end of the dynamic relay path optimization module transmits the path instruction and data packet to the input end of the host data processing and feedback module, and the host completes the global data fusion and task logic processing; The output end of the host data processing and feedback module transmits the decision results back to the control interface of the dynamic relay path optimization module through an independent feedback channel and a control instruction generator, forming a closed-loop parameter adjustment link, and at the same time directly sends bandwidth reallocation instructions to the adaptive communication scheduling module to adapt to real-time task requirements.

2. The UAV cluster communication system based on weak signal connection according to claim 1, characterized in that: The internal configuration of the signal strength monitoring and classification module includes: Multi-band signal sensor and real-time analysis unit. The sensor continuously collects signal strength parameters between each slave and the host, including RSSI value, signal-to-noise ratio and packet transmission success rate. The analysis unit calculates the mean and fluctuation rate of signal strength through a sliding window algorithm, and dynamically divides the signal level based on the preset adaptive grading threshold.

3. The UAV cluster communication system based on weak signal connection according to claim 1, characterized in that: The mission information priority management module is equipped with a three-level data classification architecture and a dynamic screening engine. The first-level information is defined as identity identification, real-time coordinates and emergency control instructions. The second-level information covers motion state parameters such as speed, heading and ambient wind speed. The third-level information includes sensor raw data such as IMU readings and attitude angles.

4. The UAV cluster communication system based on weak signal connection according to claim 1, characterized in that: The adaptive communication scheduling module is equipped with a bandwidth adaptation algorithm and a coding strategy library. The algorithm dynamically allocates transmission resources based on signal strength and data priority. The coding strategy library pre-stores multiple modulation schemes, including QPSK, 16-QAM and 64-QAM. With each level increase in signal strength, it gradually switches to higher-order modulation to improve throughput. The module has a built-in feedback mechanism to monitor packet loss rate and delay in real time. If the packet loss in a weak signal scenario exceeds the threshold, the redundant transmission mode is triggered and the first-level information is sent through dual-path backup.

5. The UAV cluster communication system based on weak signal connection according to claim 1, characterized in that: The communication topology perception submodule consists of a data fusion unit, a task logic engine, and a control instruction generator. The data fusion unit receives discrete data packets uploaded by each slave and reconstructs the global task view through timestamp alignment and spatial interpolation algorithms. The task logic engine analyzes the task progress based on preset rules, automatically calculates the corrected path when each slave deviates from the predetermined track, or initiates cluster reorganization instructions after identifying weak signal areas. The control instruction generator encodes the decision results into a standardized instruction set, sends it to each slave through an independent feedback channel, and generates a visual report that is sent back to the control background.

6. The UAV cluster communication system based on weak signal connection according to claim 1, characterized in that: The relay node evaluation submodule collects the signal strength, energy consumption status and task load data of each slave in the cluster in real time, evaluates signal stability by analyzing historical signal fluctuation rate and current RSSI average, measures energy efficiency by combining remaining power and unit data transmission energy consumption, and determines the task load level of each slave based on whether it is currently undertaking a critical task. Finally, based on the comprehensive score of signal stability, energy efficiency and task load, a candidate node priority list is generated to screen out the optimal relay node. The comprehensive scoring formula for relay nodes is: Where: σ 2 Indicates the historical signal strength variance, reflecting signal volatility. The smaller the variance, the higher the stability. μRSSI represents the current mean signal strength in dBm, which directly reflects the link quality. α represents the signal stability weight coefficient, which is adjusted according to the real-time requirements of the task; Q represents the percentage of remaining power in the node; E represents the energy consumption per unit data transmission; β represents the energy efficiency weight coefficient, which is increased when focusing on energy saving goals; L represents the task load level, ranging from 1 to 3. When the load is high, the node will not be selected as a relay; γ represents the load penalty coefficient, which prevents high-load nodes from excessively participating in relaying.

7. The UAV cluster communication system based on weak signal connection according to claim 1, characterized in that: The reinforcement learning path decision submodule includes receiving dynamic network structure data generated by the communication topology perception submodule in real time and the candidate node priority table provided by the relay node evaluation submodule. The state space is constructed through the deep reinforcement learning algorithm. The state is defined as the location, signal strength, remaining power and task load level of the current cluster node. The action space is to select the next hop relay node or directly connect to the host. The reward function integrates multi-objective optimization indicators. The experience replay mechanism and target network separation technology are used in the training process to improve the convergence efficiency. Finally, the optimal relay path strategy based on the real-time environment is output to ensure the efficiency and robustness of data transmission under weak signals. The calculation formula of the multi-objective fusion reward function is: Where: ΔH: The change in the number of hops in a path, defined as the difference between the current path hop count and the historical optimal hop count. As the number of hops decreases, the exponential term increases, directly incentivizing the algorithm to select paths with fewer hops. σ: Link signal strength variance, reflecting the signal volatility between nodes in the path. The smaller the variance, the more stable the link. The positive reward for a stable link is amplified by subtracting the variance from 1. E_avg: the average specific energy consumption of all nodes in the path, in milliampere-hours per megabyte; the lower the energy consumption, the smaller the penalty value; Q_max: The maximum remaining power percentage of the node in the path. The more sufficient the power, the smaller the impact of the penalty term; λ1, λ2, and λ3 are dynamic weight coefficients that are adjusted in real time based on the task type. In urgent tasks, λ1 is increased to prioritize reducing the number of hops, while in persistent tasks, λ2 and λ3 are increased to optimize energy consumption and stability.

8. The UAV cluster communication system based on weak signal connection according to claim 1, characterized in that: The predictive relay switching submodule includes real-time collection of current relay link signal strength data and historical attenuation trends, analysis of signal strength change patterns through a time series model, establishment of a prediction curve based on signal attenuation rate, and dynamic calculation of the confidence interval of signal quality within a certain time window in the future, in combination with the location, remaining power, and stability parameters of neighboring nodes. If the predicted signal strength is lower than a preset threshold and the confidence level exceeds a set threshold, the relay migration process is triggered, and nodes with stable signals, balanced energy consumption, and low task load are selected from the pre-loaded list of standby nodes as migration targets. At the same time, the old and new relay nodes are coordinated to complete communication protocol switching and data cache synchronization, ensuring that the link switching process is imperceptible and there is zero packet loss. The migration triggering condition formula of the predictive relay switching submodule is: MigrateFlag=μ·ΔS predicted +ν·(1-C node )+ξ·E ratio ; Where: ΔS_predicted: Predicted signal strength attenuation. The absolute value of the signal attenuation in the future time window is calculated using the time series model. The unit is dBm. C_node: The comprehensive stability score of the candidate backup node, which is calculated by weighting the node's historical online rate and the current signal volatility, with a score range of 0 to 1; E_ratio: The energy consumption difference ratio between the new and old relay nodes, defined as the percentage of the new node's unit data transmission energy consumption to the original node's energy consumption, used to quantify the energy consumption optimization space after migration; μ, ν, ξ: dynamic weight coefficients, which adjust the priority according to the task type; In emergency tasks, μ is increased to quickly respond to signal degradation, and in persistent tasks, ν and ξ are increased to prioritize node stability and energy consumption balance.

9. The UAV cluster communication system based on weak signal connection according to claim 1, characterized in that: The relay link execution and monitoring submodule includes real-time reception of the optimal path instructions generated by the dynamic relay path optimization module, coordinating the communication protocol switching operations between the sub-machines in the cluster, and continuously monitoring the real-time performance indicators of the current relay link, including packet loss rate, signal strength fluctuation, transmission delay and node remaining power, and performing real-time detection of abnormal conditions by setting dynamic thresholds.

10. The UAV cluster communication system based on weak signal connection according to claim 1, characterized in that: The host data processing and feedback module consists of a data fusion unit, a task logic engine, and a control instruction generator. The data fusion unit receives discrete data packets uploaded by each slave and reconstructs the global task view through timestamp alignment and spatial interpolation algorithms. The mission logic engine analyzes the mission progress based on preset rules, automatically calculates the corrected path when detecting that each sub-machine deviates from the planned track, or initiates cluster reorganization instructions after identifying weak signal areas; the control instruction generator encodes the decision results into a standardized instruction set, sends it to each sub-machine through an independent feedback channel, and generates a visual report and sends it back to the control background.

Images