An unmanned aerial vehicle cluster distributed intelligent decision and cooperative control system

Abstract

This invention discloses a distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms, relating to the field of UAV technology. The invention includes a data acquisition module, a situational awareness construction module, a distributed strategy negotiation module, a motion planning module, and a consistency control module. UAVs interact with neighboring nodes only to exchange decision parameters, reducing communication load and bandwidth requirements, avoiding the risk of single-node failure, and improving the robustness and operational reliability of the swarm system in complex maritime communication environments. It filters environmental information strongly correlated with its own motion, suppresses noise interference, and provides complete and reliable situational awareness data for swarm decision-making, improving the accuracy and anti-interference capability of maritime environmental situational awareness. Control parameters are adapted in real-time based on flight speed, attitude changes, and remaining battery power, balancing high-speed response accuracy and low-speed control stability, reducing low-power energy consumption, and enhancing the stability, endurance, and adaptability of swarm collaborative control to maritime operations.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and more specifically to a distributed intelligent decision-making and collaborative control system for UAV swarms. Background Technology

[0002] With the rapid iteration of UAV autonomous control, swarm collaboration, and maritime operation technologies, UAV swarms have been widely applied in scenarios such as maritime reconnaissance, inspection, search and rescue, and material delivery. Existing UAV swarm control systems mostly adopt a centralized decision-making architecture, with a central control node uniformly handling situational awareness fusion, task allocation, path planning, and command issuance. This is insufficient to meet the actual needs of complex maritime weather, strong electromagnetic interference, limited communication bandwidth, and long-endurance operations. Therefore, it is necessary to analyze a distributed intelligent decision-making and collaborative control system for UAV swarms.

[0003] Existing technology, such as the invention application patent with announcement number CN119597013B, discloses a hybrid intelligent collaborative control system and method for operator / UAV swarms. The system includes an operator trust calculation module, an operator interaction command module, a swarm trust calculation module, a distributed collaborative control calculation module, and a visual display module. The invention method is as follows: Step 1: Given a given intelligent agent swarm model and a UAV swarm model and perform mapping; Step 2: Design a UAV swarm communication model; Step 3: Design a distributed collaborative controller for the UAV swarm; Step 4: Design a human-machine mutual trust model and determine trust levels. This invention, by considering the impact of external environmental interference on UAV swarms and introducing the advantages of intelligent decision-making by human operators and distributed autonomous control of UAVs, constructs an efficient, flexible, and scalable hybrid intelligent collaborative control framework, improving the task execution efficiency and safety of UAV swarms in dynamic environments.

[0004] While existing technologies can meet basic requirements for a distributed intelligent decision-making and collaborative control system for UAV swarms, they also present some potential defects and challenges, specifically in the following aspects: First, existing UAV swarm control systems mostly adopt a centralized decision-making architecture, relying on a central node to complete global calculations and command issuance. This results in high communication bandwidth consumption, single-node failures easily leading to swarm paralysis, and communication links being prone to interruption in complex maritime environments, making it difficult to meet the requirements for long-endurance and high-reliability maritime operations. Second, existing technologies lack deep integration of their own state and environmental information in the situational awareness stage, often using simple data splicing to generate situational information, leading to significant environmental noise interference, insufficient extraction of key features, and low accuracy in situational representation. Third, existing swarm collaborative control methods do not adaptively optimize based on the maritime operation scenario and UAV status. Control parameters are fixed and cannot be adjusted, resulting in lag in tracking response during high-speed maneuvers, easy control oscillation during low-speed flight, and excessive energy consumption for formation maintenance when battery is low. Furthermore, the lack of a dynamic communication weight mechanism in distributed negotiation reduces convergence speed, increases the frequency of parameter oscillations, and reduces the stability and endurance of collaborative control. Summary of the Invention

[0005] The purpose of this invention is to provide a distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms, which solves the problems existing in the background technology.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms, including a data acquisition module, a situational characterization construction module, a distributed strategy negotiation module, an action planning module, and a consistency control module.

[0007] The data acquisition module is used to acquire multidimensional self-state data and local environmental observation data of each UAV in the cluster.

[0008] The situation characterization construction module is used to extract features and interactively encode multidimensional self-state data and local environmental observation data to generate a joint situation characterization vector of UAV operating status, marine environment, and target situation.

[0009] The distributed policy negotiation module is used to conduct multiple rounds of policy negotiation among neighboring UAVs in a maritime operation cluster based on the joint situational representation vector, iteratively updating the collaborative decision parameters until convergence, and obtaining a globally consistent collaborative decision policy.

[0010] The motion planning module is used to generate a sequence of motion commands adapted to the marine environment for each UAV based on a collaborative decision-making strategy, and to plan the optimal operation path in combination with the UAV's endurance status.

[0011] The consistency control module is used to convert action command sequences into flight control parameters based on the consistency control protocol, thereby completing the coordinated control of the UAV swarm.

[0012] Furthermore, the data acquisition module specifically includes: acquiring multi-dimensional self-state data of each UAV through an airborne sensor system, wherein the multi-dimensional self-state data includes current position coordinates, flight speed vector, attitude angle data, and remaining battery power information.

[0013] Local environmental observation data of each UAV is collected by airborne sensing equipment. The local environmental observation data includes the location information of surrounding obstacles, the relative distance and relative azimuth information of neighboring UAVs.

[0014] Furthermore, the situation characterization construction module specifically includes: encoding and mapping the multidimensional self-state data to obtain state feature vectors, and encoding and mapping the local environmental observation data to obtain environmental feature vectors.

[0015] Establish an association mapping relationship between the state feature vector and the environment feature vector, and perform interactive filtering and enhancement on the environment feature vector based on the association mapping relationship to filter out environmental information related to its own state.

[0016] The enhanced environmental feature vector is combined with the state feature vector to generate a joint situational representation vector that includes information about its own motion state and the surrounding environment.

[0017] Furthermore, the distributed policy negotiation module specifically includes: inputting the joint situational awareness vector into the policy network to generate initial collaborative decision parameters.

[0018] In the preset communication topology, each drone exchanges its own collaborative decision-making parameters with its neighboring drones.

[0019] The collaborative decision-making parameters are updated in multiple rounds using a distributed consensus protocol. In each round of iteration, the parameters are adjusted based on the collaborative decision-making parameters of the neighboring drones.

[0020] Until the collaborative decision parameters of all UAVs converge to a consistent value, a globally consistent collaborative decision strategy is output. The collaborative decision strategy includes the overall formation configuration of the cluster, the target mission role of each UAV, and the direction of the cluster's movement trend.

[0021] Furthermore, the multi-round iterative update is based on the following formula: ,in, Indicates the first The drone in Collaborative decision-making parameters during rounds of negotiation Indicates the negotiation step size, , Indicates the first The set of neighboring drones in the communication topology of a drone. Indicates the first The drone and the first Communication weights between drones .

[0022] Furthermore, the communication weight is dynamically adjusted based on the communication link quality, specifically including: each UAV detects the communication link quality indicators with neighboring UAVs in each round of negotiation, the communication link quality indicators including signal strength, signal-to-noise ratio and packet loss rate.

[0023] The communication link quality indicators are mapped to link quality scores. When the signal strength is detected to be lower than a preset signal strength threshold, the signal-to-noise ratio is lower than a preset signal-to-noise ratio threshold, or the packet loss rate is higher than a preset packet loss rate threshold, the link quality score of the corresponding communication link is reduced.

[0024] The communication weights are normalized and updated based on the link quality score, so that the communication links with higher link quality scores correspond to larger communication weights, and the sum of the communication weights of all neighbors of each UAV remains 1.

[0025] Furthermore, the action planning module specifically includes: calculating the expected action of each UAV according to the globally consistent collaborative decision-making strategy, wherein the expected action includes the expected heading angle, the expected flight speed and the expected flight altitude.

[0026] The desired heading angle, desired flight speed, and desired flight altitude are arranged in chronological order to generate a sequence of action commands. The sequence of action commands includes the continuous action targets of each UAV within a future preset time window.

[0027] Furthermore, the consistency control module also includes an adaptive gain adjustment unit, which is used to dynamically adjust the control gain according to the current flight status of the UAV. Specifically, it includes: real-time monitoring of the current flight status of each UAV, which includes the current flight speed, the current attitude angle change rate, and the current remaining battery power.

[0028] Based on the comparison between the current flight speed and the preset speed threshold, the position error control gain is adjusted: when the current flight speed exceeds the preset high speed threshold, the position error control gain is increased; when the current flight speed is lower than the preset low speed threshold, the position error control gain is decreased.

[0029] Based on the comparison between the current attitude angle change rate and the preset attitude change rate threshold, the speed error control gain is adjusted: when the current attitude angle change rate exceeds the preset attitude change rate threshold, the speed error control gain is increased.

[0030] Based on the comparison between the current remaining power and the preset power threshold, the formation holding control gain is adjusted: when the current remaining power is lower than the preset low power threshold, the formation holding control gain is reduced.

[0031] Furthermore, the consistency control module specifically includes: determining the target flight state of each UAV according to the action command sequence, wherein the target flight state includes target position coordinates, target velocity vector, and target attitude angle.

[0032] A consistency control protocol is used to calculate the deviation between the target flight state and the current flight state.

[0033] Flight control parameters are generated based on the deviation, including control surface deflection angle control and throttle opening control.

[0034] The actual flight status of each drone tracks the flight status of the target, enabling coordinated control of the drone swarm.

[0035] Furthermore, the flight control parameters are generated based on the following formula: ,in, Indicates the first A drone in Flight control parameters at any given time, They represent the first A drone in The target's position coordinates and velocity vector at any given time. Let represent the current position coordinates and current velocity vector of the i-th UAV at time t, respectively. Indicates the first A drone in The communication connection status between the drones These represent the position error control gain, velocity error control gain, and formation holding control gain, respectively.

[0036] The beneficial effects of this invention are as follows: 1. This invention adopts a distributed strategy negotiation mechanism, where the UAV only interacts with neighboring nodes to exchange decision parameters, eliminating the need for unified scheduling by a central node. This reduces communication load and bandwidth requirements, avoids the risk of single-node failure, and improves the robustness and operational reliability of the cluster system in complex maritime communication environments. 2. Through enhanced state-environment feature interaction via an attention mechanism, environmental information strongly correlated with its own motion is filtered, noise interference is suppressed, and a high-precision joint situational awareness vector is generated. This provides a complete and reliable situational basis for cluster decision-making, improving the accuracy and anti-interference capability of maritime environmental situational awareness. 3. Adaptive gain adjustment and dynamic communication weight collaborative optimization are achieved. Control parameters are adapted in real time based on flight speed, attitude changes, and remaining battery power, balancing high-speed response accuracy and low-speed control stability, reducing low-power energy consumption, and simultaneously improving the distributed negotiation convergence speed. This enhances the stability of cluster collaborative control, endurance, and adaptability to maritime operations. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Reference Figure 1 As shown, the present invention provides a distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms, comprising: a data acquisition module, a situational characterization construction module, a distributed strategy negotiation module, an action planning module, and a consistency control module.

[0041] The data acquisition module is used to acquire multidimensional self-state data and local environmental observation data of each UAV in the cluster.

[0042] The situation characterization construction module is used to extract features and interactively encode multidimensional self-state data and local environmental observation data to generate a joint situation characterization vector of UAV operating status, marine environment, and target situation.

[0043] The distributed policy negotiation module is used to conduct multiple rounds of policy negotiation among neighboring UAVs in a maritime operation cluster based on the joint situational representation vector, iteratively updating the collaborative decision parameters until convergence, and obtaining a globally consistent collaborative decision policy.

[0044] The motion planning module is used to generate a sequence of motion commands adapted to the marine environment for each UAV based on a collaborative decision-making strategy, and to plan the optimal operation path in combination with the UAV's endurance status.

[0045] The consistency control module is used to convert action command sequences into flight control parameters based on the consistency control protocol, thereby completing the coordinated control of the UAV swarm.

[0046] In the above embodiments, the data acquisition module specifically includes: acquiring multi-dimensional self-state data of each UAV through an airborne sensor system, wherein the multi-dimensional self-state data includes current position coordinates, flight speed vector, attitude angle data and remaining battery power information.

[0047] Local environmental observation data of each UAV is collected by airborne sensing equipment. The local environmental observation data includes the location information of surrounding obstacles, the relative distance and relative azimuth information of neighboring UAVs.

[0048] It should be noted that the multi-dimensional self-state data collected by the airborne sensor system is used to describe the motion state and energy status of each UAV at the current moment. The current position coordinates are obtained through the Global Positioning System, the flight speed vector is obtained through the inertial measurement unit combined with the airspeed tube, the attitude angle data, including pitch angle, roll angle and yaw angle, are obtained through the inertial navigation system, and the remaining power information is collected through the battery management system.

[0049] It should be noted that the local environmental observation data collected by the airborne sensing equipment is used to describe the surrounding environmental information perceived by each UAV; the location information of surrounding obstacles is obtained by scanning with airborne lidar or millimeter-wave radar; the relative distance and relative azimuth information of neighboring UAVs are calculated by the signal arrival time difference and signal arrival angle of the airborne communication module, or can be obtained by identifying and ranging neighboring UAVs through airborne visual sensors.

[0050] In the above embodiments, the situation characterization construction module specifically includes: encoding and mapping the multidimensional self-state data to obtain state feature vectors, and encoding and mapping the local environmental observation data to obtain environmental feature vectors.

[0051] Establish an association mapping relationship between the state feature vector and the environment feature vector, and perform interactive filtering and enhancement on the environment feature vector based on the association mapping relationship to filter out environmental information related to its own state.

[0052] The enhanced environmental feature vector is combined with the state feature vector to generate a joint situational representation vector that includes information about its own motion state and the surrounding environment.

[0053] It should be noted that encoding and mapping multidimensional self-state data to obtain state feature vectors means normalizing the parameters in the multidimensional self-state data and then mapping them into fixed-dimensional feature vectors through a fully connected network to characterize the UAV's own motion state characteristics. Encoding and mapping local environmental observation data to obtain environmental feature vectors means structuring the parameters in the local environmental observation data and then mapping them into fixed-dimensional feature vectors through a convolutional network or graph neural network to characterize the environmental scene characteristics perceived by the UAV.

[0054] It should be noted that establishing the association mapping relationship between the state feature vector and the environment feature vector means calculating the attention weight of the state feature vector to each dimension of the environment feature vector through an attention mechanism, filtering out environmental information that is strongly correlated with its own motion state, while suppressing feature components corresponding to environmental noise or irrelevant information, thereby optimizing and enhancing the environment feature vector.

[0055] It should be noted that combining the enhanced environmental feature vector with the state feature vector means concatenating the two feature vectors along the feature dimension to form a dimension-enlarged joint feature representation, namely a joint situational characterization vector. This vector simultaneously contains the UAV's own motion state information and the filtered and enhanced surrounding environment information, providing a complete situational understanding foundation for subsequent distributed decision-making.

[0056] In the above embodiments, the distributed policy negotiation module specifically includes: inputting the joint situational awareness vector into the policy network to generate initial collaborative decision parameters.

[0057] In the preset communication topology, each drone exchanges its own collaborative decision-making parameters with its neighboring drones.

[0058] The collaborative decision-making parameters are updated in multiple rounds using a distributed consensus protocol. In each round of iteration, the parameters are adjusted based on the collaborative decision-making parameters of the neighboring drones.

[0059] Until the collaborative decision parameters of all UAVs converge to a consistent value, a globally consistent collaborative decision strategy is output. The collaborative decision strategy includes the overall formation configuration of the cluster, the target mission role of each UAV, and the direction of the cluster's movement trend.

[0060] It should be noted that the joint situational awareness vector is input into the policy network to generate initial cooperative decision parameters. The policy network is a pre-trained deep neural network, whose input is the joint situational awareness vector and whose output is the initial cooperative policy parameters that the UAV should adopt in the swarm, including the expected formation position offset, target velocity preference value and task priority coefficient, etc.

[0061] It should be noted that in the preset communication topology, each UAV exchanges its own collaborative decision parameters with its neighboring UAVs. The communication topology is a communication connection diagram that is preset before the UAV cluster executes the task. Each UAV only exchanges data with its neighboring UAVs in the communication topology, without having to establish direct communication with all UAVs in the cluster, thereby reducing communication load and bandwidth requirements.

[0062] It should be noted that using a distributed consensus protocol to iterate and update the collaborative decision parameters in multiple rounds means that each UAV receives the collaborative decision parameters from all neighboring UAVs in each round of negotiation, calculates the adjustment amount of its own parameters according to the consensus protocol rules, updates its own parameters, sends the new parameters to the neighboring UAVs, and enters the next round of negotiation iteration.

[0063] It should be noted that the consensus is reached when the maximum difference between the collaborative decision parameters of all UAVs is less than the preset convergence threshold. At this point, the collaborative decision parameters of each UAV tend to be consistent, and a globally consistent collaborative decision strategy is obtained.

[0064] The globally consistent collaborative decision-making strategy includes the following components to guide the movement behavior of the entire UAV swarm: the overall formation configuration of the swarm defines the relative spatial positions of the UAVs; the target mission role of each UAV defines the specific mission of each UAV in the formation, whether it is to perform reconnaissance or to perform strike; and the swarm's movement trend direction defines the overall movement direction of the entire formation.

[0065] In the above embodiments, the multi-round iterative update is based on the following formula: ,in, Indicates the first The drone in Collaborative decision-making parameters during rounds of negotiation Indicates the negotiation step size, , Indicates the first The set of neighboring drones in the communication topology of a drone. Indicates the first The drone and the first Communication weights between drones .

[0066] It should be noted that the multi-round iterative update formula is used to achieve global consistency convergence of decision parameters for multiple drones in a distributed scenario; the calculation process of this formula is as follows: First, calculate the... In each round of negotiation, the difference between the collaborative decision parameters of each neighboring drone and its own parameters is calculated. Then, each difference is multiplied by the corresponding communication weight and summed using a weighted average. Finally, the weighted sum is multiplied by the negotiation step size to obtain the adjustment amount for this iteration. This adjustment amount is then compared with the first iteration. The parameters from each round of negotiation are added together to obtain the first round. Updated parameters during rounds of negotiation.

[0067] It should be noted that the negotiation step size is used to control the adjustment range of each iteration, and its value is between 0 and 1. An excessively large step size may cause parameter oscillations and prevent convergence, while an excessively small step size will result in slow convergence. Typically, the value is between 0.1 and 0.3. The communication weight is used to measure the degree of influence of neighbor information on its own decision-making. The better the communication link signal quality or the closer the distance between two drones, the larger the communication weight value. The weight value ranges from 0 to 1, and the sum of the communication weights of all neighbors for each drone is 1.

[0068] It should be noted that the multi-round iterative update process continues until the maximum difference between the collaborative decision parameters of all UAVs is less than a preset convergence threshold. The convergence threshold is usually between 0.001 and 0.01, and the specific value is determined according to the mission accuracy requirements.

[0069] In the above embodiments, the communication weight is dynamically adjusted based on the communication link quality, specifically including: each UAV detects the communication link quality indicators with neighboring UAVs in each round of negotiation, the communication link quality indicators including signal strength, signal-to-noise ratio and packet loss rate.

[0070] The communication link quality indicators are mapped to link quality scores. When the signal strength is detected to be lower than a preset signal strength threshold, the signal-to-noise ratio is lower than a preset signal-to-noise ratio threshold, or the packet loss rate is higher than a preset packet loss rate threshold, the link quality score of the corresponding communication link is reduced.

[0071] The communication weights are normalized and updated based on the link quality score, so that the communication links with higher link quality scores correspond to larger communication weights, and the sum of the communication weights of all neighbors of each UAV remains 1.

[0072] It should be noted that the communication link quality indicators detected by each drone in each round of negotiation with neighboring drones refer to the following: while receiving the cooperative decision parameter data frames from neighboring drones, each drone measures the signal strength indicator of the received signal through the radio frequency front-end circuit, measures the signal-to-noise ratio through the demodulator, and measures the packet loss rate by detecting the continuity statistics of the data frame sequence numbers. These three indicators together characterize the actual transmission quality of the communication link.

[0073] It should be noted that mapping the communication link quality indicators to a link quality score means that the signal strength, signal-to-noise ratio (SNR), and packet loss rate are normalized separately, and then weighted and summed according to preset weighting coefficients to obtain the link quality score. When the signal strength is lower than a preset signal strength threshold, typically -90dBm, it indicates that the communication link is becoming unstable, and a lower component score is assigned accordingly. When the SNR is lower than a preset SNR threshold, typically 10dB, it indicates that the communication is severely affected by noise interference, and a lower component score is assigned accordingly. When the packet loss rate is higher than a preset packet loss rate threshold, it indicates poor data transmission reliability, and a lower component score is assigned accordingly.

[0074] It should be noted that the link quality score used to normalize and update the communication weights involves first using the link quality score between each UAV and its neighboring UAVs as the initial value for the corresponding communication weight. Then, all weight values ​​are divided by the sum of all weights, normalizing the sum of the communication weights of all neighbors for each UAV to 1. Through this dynamic adjustment mechanism, neighbor information with better communication quality receives greater influence during the negotiation process, while the influence of neighbor information with poorer communication quality is automatically suppressed.

[0075] In the above embodiments, the action planning module specifically includes: calculating the expected action of each UAV according to the globally consistent collaborative decision-making strategy, wherein the expected action includes the expected heading angle, the expected flight speed and the expected flight altitude.

[0076] The desired heading angle, desired flight speed, and desired flight altitude are arranged in chronological order to generate a sequence of action commands. The sequence of action commands includes the continuous action targets of each UAV within a future preset time window.

[0077] It should be noted that calculating the expected actions of each UAV according to the globally consistent collaborative decision-making strategy means mapping the formation configuration parameters, mission role parameters, and motion trend parameters included in the collaborative decision-making strategy to the specific motion control targets of each UAV through a decoding network. Specifically, the expected heading angle is calculated based on the motion trend direction of the cluster in the collaborative decision-making strategy combined with the relative position offset of the UAV in the formation configuration; the expected flight speed is determined based on the target speed preference value in the collaborative decision-making strategy; and the expected flight altitude is determined based on the altitude hierarchy allocation in the formation configuration.

[0078] It should be noted that arranging the desired heading angle, desired flight speed, and desired flight altitude in chronological order to generate a sequence of action commands means dividing a future preset time window into multiple time segments, setting corresponding target values ​​for the desired heading angle, desired flight speed, and desired flight altitude within each time segment, thus forming a continuous and smooth action command trajectory; the preset time window is typically 5 to 10 seconds, and the resolution of the time segment is 0.1 to 0.5 seconds.

[0079] In the above embodiments, the consistency control module further includes an adaptive gain adjustment unit, which is used to dynamically adjust the control gain according to the current flight status of the UAV. Specifically, it includes: real-time monitoring of the current flight status of each UAV, the current flight status including the current flight speed, the current attitude angle change rate and the current remaining battery power.

[0080] Based on the comparison between the current flight speed and the preset speed threshold, the position error control gain is adjusted: when the current flight speed exceeds the preset high speed threshold, the position error control gain is increased; when the current flight speed is lower than the preset low speed threshold, the position error control gain is decreased.

[0081] Based on the comparison between the current attitude angle change rate and the preset attitude change rate threshold, the speed error control gain is adjusted: when the current attitude angle change rate exceeds the preset attitude change rate threshold, the speed error control gain is increased.

[0082] Based on the comparison between the current remaining power and the preset power threshold, the formation holding control gain is adjusted: when the current remaining power is lower than the preset low power threshold, the formation holding control gain is reduced.

[0083] It should be noted that increasing the position error control gain improves the response speed of high-speed tracking; decreasing the position error control gain avoids over-adjustment at low speeds; increasing the velocity error control gain enhances the velocity holding capability under attitude disturbances; and decreasing the formation holding control gain reduces the energy consumption required for formation holding.

[0084] It should be noted that real-time monitoring of the current flight status of each drone provides input for subsequent adaptive gain adjustments. The current flight speed is obtained through an airspeed tube or a GPS speed measurement module; the current attitude angle change rate is obtained through the gyroscope output of the inertial measurement unit; and the current remaining battery power is obtained through real-time monitoring by the battery management system.

[0085] Furthermore, adjusting the position error control gain based on the current flight speed is to address the adaptation issue of tracking response characteristics at different flight speeds. When the UAV flies at high speed, the delay effect of position tracking is more significant. Increasing the position error control gain can accelerate the correction speed of position deviation, ensuring accurate tracking of the target position even at high speeds. When the UAV flies at low speed, excessive gain may cause oscillations or overshoot in position control. Decreasing the position error control gain can make the control response more stable.

[0086] It should be noted that adjusting the speed error control gain based on the current rate of change of attitude angle is to address the stability issue of maintaining speed during violent maneuvers of the UAV. When the UAV performs large maneuvers such as sharp turns or rapid climbs, the rate of change of attitude angle increases significantly. At this time, the drastic change in the fuselage attitude can interfere with the stability of flight speed. Increasing the speed error control gain can enhance the anti-disturbance capability of the speed closed loop, enabling the UAV to maintain the target flight speed under attitude disturbances.

[0087] Furthermore, adjusting the formation-keeping control gain based on the current remaining battery power is to optimize energy consumption when battery power is low. Formation-keeping control requires the drone to constantly adjust its position to maintain its relative position with its neighbors, a process that consumes additional energy. When the remaining battery power is below a preset low-battery threshold, typically 20% to 30% of the total battery power, appropriately reducing the formation-keeping control gain can relax the requirements for formation position accuracy, reduce the drone's maneuver frequency and amplitude, thereby extending endurance and prioritizing the drone's safe return.

[0088] In the above embodiments, the consistency control module specifically includes: determining the target flight state of each UAV according to the action command sequence, wherein the target flight state includes target position coordinates, target velocity vector and target attitude angle.

[0089] A consistency control protocol is used to calculate the deviation between the target flight state and the current flight state.

[0090] Flight control parameters are generated based on the deviation, including control surface deflection angle control and throttle opening control.

[0091] The actual flight status of each drone tracks the flight status of the target, enabling coordinated control of the drone swarm.

[0092] It should be noted that determining the target flight state of each UAV based on the action command sequence means converting the expected heading angle, expected flight speed, and expected flight altitude corresponding to the current moment in the action command sequence into target position coordinates, target velocity vector, and target attitude angle in a unified coordinate system, which are then used as the tracking targets of the consistency control module.

[0093] It should be noted that calculating the deviation between the target flight state and the current flight state using the consistency control protocol refers to calculating the position deviation vector between the target position coordinates and the current position coordinates, the velocity deviation vector between the target velocity vector and the current velocity vector, and simultaneously calculating the attitude deviation angle between the target attitude angle and the current attitude angle. These deviations are used to drive the generation of flight control parameters.

[0094] It should be noted that generating flight control parameters based on the aforementioned deviations means linearly combining the position deviation, speed deviation, and formation holding deviation according to a preset control gain to generate three control surface deflection angles for controlling the pitch, yaw, and roll of the UAV, as well as a throttle opening control for controlling flight speed.

[0095] It should be noted that enabling the actual flight state of each UAV to track the target flight state means that after receiving the flight control parameters, the flight control system of each UAV drives the servo motor and throttle actuator to gradually approach and maintain the position, speed and attitude of the UAV as the target flight state. At the same time, through the negative feedback effect of the formation to maintain the deviation, the relative positional relationship between the UAVs is maintained, thereby achieving the coordinated control of the entire swarm.

[0096] In the above embodiments, the flight control parameters are generated based on the following formula: ,in, Indicates the first A drone in Flight control parameters at any given time, They represent the first A drone in The target's position coordinates and velocity vector at any given time. Let represent the current position coordinates and current velocity vector of the i-th UAV at time t, respectively. Indicates the first A drone in The communication connection status between the drones These represent the position error control gain, velocity error control gain, and formation holding control gain, respectively.

[0097] It should be noted that the formula for generating flight control parameters is used to map high-level action commands to low-level flight control signals, while simultaneously achieving formation keeping functionality. The calculation process is as follows: First, the difference between the target's position coordinates and the current position coordinates is calculated and multiplied by the position error control gain to obtain the position tracking control quantity, used to guide the UAV towards the target position. Simultaneously, the difference between the target's velocity vector and the current velocity vector is calculated and multiplied by the velocity error control gain to obtain the velocity tracking control quantity, used to control the UAV's flight speed to match the target's speed. Then, the difference between the position coordinates of each neighboring UAV and its own position coordinates is calculated, multiplied by the communication connection status for filtering, summed, and multiplied by the formation keeping control gain to obtain the formation keeping control quantity, used to maintain the relative position configuration between UAVs. Finally, the above three control quantities are added together to obtain the final flight control parameters.

[0098] It should be noted that the communication connection status is used to characterize whether two UAVs can communicate effectively. When the i-th UAV and the j-th UAV can communicate directly, i.e., the communication link is connected, the communication connection status value is 1, indicating that the position information of the neighbor is considered in the formation holding control. When the communication link is disconnected, the communication connection status value is 0, indicating that the position information of the neighbor is ignored. The position error control gain, velocity error control gain, and formation holding control gain are all determined through actual machine debugging or adaptive optimization, wherein the position error control gain is 0.5 to 2.0, the velocity error control gain is 0.1 to 0.5, and the formation holding control gain is 0.2 to 1.0.

[0099] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms, characterized in that, include: The data acquisition module is used to acquire multidimensional self-state data and local environmental observation data of each UAV in the cluster; The situation characterization construction module is used to extract features and interactively encode multidimensional self-state data and local environmental observation data to generate a joint situation characterization vector of UAV operating status, marine environment and target situation. The distributed policy negotiation module is used to conduct multiple rounds of policy negotiation among neighboring UAVs in the maritime operation cluster based on the joint situational representation vector, iteratively updating the collaborative decision parameters until convergence, and obtaining a globally consistent collaborative decision strategy. The motion planning module is used to generate a sequence of motion commands adapted to the marine environment for each UAV based on a collaborative decision-making strategy, and to plan the optimal operation path in combination with the UAV's endurance status. The consistency control module is used to convert action command sequences into flight control parameters based on the consistency control protocol, thereby completing the coordinated control of the UAV swarm.

2. The distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that, The data acquisition module specifically includes: The airborne sensor system collects multi-dimensional self-state data for each UAV, including current position coordinates, flight speed vector, attitude angle data, and remaining battery power information. Local environmental observation data of each UAV is collected by airborne sensing equipment. The local environmental observation data includes the location information of surrounding obstacles, the relative distance and relative azimuth information of neighboring UAVs.

3. The distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms according to claim 2, characterized in that, The situation characterization construction module specifically includes: The multidimensional self-state data is encoded and mapped to obtain state feature vectors, and the local environmental observation data is encoded and mapped to obtain environmental feature vectors. Establish an association mapping relationship between the state feature vector and the environment feature vector, and perform interactive filtering and enhancement on the environment feature vector based on the association mapping relationship to filter out environmental information related to its own state; The enhanced environmental feature vector is combined with the state feature vector to generate a joint situational representation vector that includes information about its own motion state and the surrounding environment.

4. The distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms according to claim 3, characterized in that, The distributed strategy negotiation module specifically includes: The joint situational awareness vector is input into the policy network to generate initial collaborative decision parameters. In the preset communication topology, each drone exchanges its own collaborative decision-making parameters with its neighboring drones; The collaborative decision-making parameters are updated iteratively in multiple rounds using a distributed consensus protocol. In each round of iteration, the parameters of the drone are adjusted according to the collaborative decision-making parameters of its neighbors. Until the collaborative decision parameters of all UAVs converge to a consistent value, a globally consistent collaborative decision strategy is output. The collaborative decision strategy includes the overall formation configuration of the cluster, the target mission role of each UAV, and the direction of the cluster's movement trend.

5. The distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms according to claim 4, characterized in that, The multiple rounds of iterative updates are based on the following formula: ,in, Indicates the first The drone in Collaborative decision-making parameters during rounds of negotiation Indicates the negotiation step size, , Indicates the first The set of neighboring drones in the communication topology of a drone. Indicates the first The drone and the first Communication weights between drones .

6. The distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms according to claim 5, characterized in that, The communication weights are dynamically adjusted based on the quality of the communication link, specifically including: Each drone detects communication link quality metrics with neighboring drones in each round of negotiation. These communication link quality metrics include signal strength, signal-to-noise ratio, and packet loss rate. The communication link quality indicators are mapped to link quality scores. When the signal strength is detected to be lower than a preset signal strength threshold, the signal-to-noise ratio is lower than a preset signal-to-noise ratio threshold, or the packet loss rate is higher than a preset packet loss rate threshold, the link quality score of the corresponding communication link is reduced. The communication weights are normalized and updated based on the link quality score, so that the communication links with higher link quality scores correspond to larger communication weights, and the sum of the communication weights of all neighbors of each UAV remains 1.

7. The distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms according to claim 6, characterized in that, The action planning module specifically includes: The expected actions of each UAV are calculated based on the globally consistent collaborative decision-making strategy. The expected actions include the expected heading angle, the expected flight speed, and the expected flight altitude. The desired heading angle, desired flight speed, and desired flight altitude are arranged in chronological order to generate a sequence of action commands. The sequence of action commands includes the continuous action targets of each UAV within a future preset time window.

8. The distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms according to claim 7, characterized in that, The consistency control module also includes an adaptive gain adjustment unit, used to dynamically adjust the control gain according to the current flight state of the UAV, specifically including: Real-time monitoring of the current flight status of each drone, including current flight speed, current rate of change of attitude angle, and current remaining battery power; Based on the comparison between the current flight speed and the preset speed threshold, the position error control gain is adjusted: when the current flight speed exceeds the preset high speed threshold, the position error control gain is increased; when the current flight speed is lower than the preset low speed threshold, the position error control gain is decreased. Based on the comparison between the current attitude angle change rate and the preset attitude change rate threshold, the speed error control gain is adjusted: when the current attitude angle change rate exceeds the preset attitude change rate threshold, the speed error control gain is increased; Based on the comparison between the current remaining power and the preset power threshold, the formation holding control gain is adjusted: when the current remaining power is lower than the preset low power threshold, the formation holding control gain is reduced.

9. A distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms according to claim 8, characterized in that, The consistency control module specifically includes: The target flight state of each UAV is determined based on the sequence of action commands. The target flight state includes the target position coordinates, the target velocity vector, and the target attitude angle. A consistency control protocol is used to calculate the deviation between the target flight state and the current flight state; Flight control parameters are generated based on the deviation, including control surface deflection angle control and throttle opening control. The actual flight status of each drone tracks the flight status of the target, enabling coordinated control of the drone swarm.

10. A distributed intelligent decision-making and collaborative control system for unmanned aerial vehicle (UAV) swarms according to claim 9, characterized in that, The flight control parameters are generated based on the following formula: ,in, Indicates the first A drone in Flight control parameters at any given time, They represent the first A drone in The target's position coordinates and velocity vector at any given time. Let represent the current position coordinates and current velocity vector of the i-th UAV at time t, respectively. Indicates the first A drone in The communication connection status between the drones These represent the position error control gain, velocity error control gain, and formation holding control gain, respectively.

Images