Multi-functional unmanned ship cluster control method and system oriented to gridding deployment

Abstract

The invention discloses a gridding deployment-oriented multifunctional unmanned ship cluster control method and system, and relates to the technical field of multifunctional unmanned ship cluster intelligent control and communication network reconstruction, and the method comprises the steps: obtaining the communication state data of each unmanned ship, and carrying out the calculation and recognition of an interruption connection through a link index. And the ships with relay conversion conditions are screened, a relay candidate set is constructed, and a strategy control method is called to generate relay task behaviors. And updating the communication connection relation according to the behavior output result of the relay boat, and adjusting the adjacency relation to complete link reconstruction. According to the method, the three-section flow of interruption recognition, behavior generation and link repair is taken as a main line, and the unmanned ship cluster control method with intelligent sensing, autonomous decision making and self-organizing repair capabilities is constructed. The task execution continuity is kept, self-optimization of resource scheduling is achieved, and the method is suitable for a wide-area maritime cooperative task scene in a dynamic environment. The method has obvious advantages in intelligence, adaptability and communication robustness.

Description

Technical Field

[0001] The present invention relates to the technical field of multifunctional unmanned boat cluster intelligent control and communication network reconstruction, and specifically to a multifunctional unmanned boat cluster control method and system for grid deployment. Background Art

[0002] With the development of intelligent marine equipment, unmanned vessels (UAVs) have been widely used in military reconnaissance, ocean mapping, and environmental monitoring missions. Currently, the deployment of UAV swarms based on task collaboration has become a core research area for intelligent marine systems. To address the challenges of highly dynamic communications in wide-area, complex sea conditions, a grid-based deployment architecture has been proposed for building distributed, reconfigurable inter-vessel communication networks to support mission assignment, data synchronization, and emergency dispatch. Simultaneously, UAV systems with multifunctional collaborative capabilities are gradually emerging, and their control logic is evolving from static rules to data-driven intelligent scheduling. The introduction of intelligent algorithms such as reinforcement learning, graph neural networks, and edge decision-making is driving the development of UAV swarm control systems towards self-organization and self-repair.

[0003] Although various unmanned aerial vehicle (UAV) swarm scheduling methods have been proposed, current mainstream technologies still have significant limitations. First, in the event of communication disruption, existing systems mostly rely on centralized scheduling mechanisms or preset reconnection strategies. These systems fail to dynamically respond to the topological location and state of the disrupted UAV, leading to low reconnection efficiency and high latency. Second, existing schemes often use static thresholds such as residual energy and geometric distance to select relay vessels. These mechanisms lack dynamic task constraints and regional density adjustment capabilities, making them difficult to adapt to real-time network conditions and task pressures. Third, while some literature has attempted to use artificial intelligence methods to optimize relay strategies, they lack structured modeling of relay behavior outputs, making it difficult to achieve refined behavioral control and combined decision-making. Furthermore, for communication graph updates and adjacency structure adjustments, existing schemes often rely on manual rules, fail to establish closed-loop link self-repair processes, and fail to consider details such as relay node stability and routing table reissues. This results in the proneness of isolated paths or logical disconnections during network reconstruction. To address the above-mentioned issues, the present invention proposes a relay mechanism with the capabilities of adaptive behavior generation, reinforcement learning control, and closed-loop link reconstruction. The system is designed from the entire process of communication status collection, candidate boat screening, behavior combination output to connection relationship update, and has both intelligence and integrity that cannot be achieved simultaneously by existing technologies. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problems solved by the present invention are: the existing unmanned boat cluster control method has the problems of inaccurate identification of interrupted links, lack of intelligent decision-making mechanism for relay screening, and lack of closed-loop repair capability in the link reconstruction process, as well as how to realize a self-organizing relay reconstruction mechanism driven by a combination of reinforcement learning and behavior.

[0006] To address the aforementioned technical issues, the present invention provides the following technical solution: a multifunctional unmanned boat swarm control method for grid-based deployment, comprising obtaining communication status data for each unmanned boat and identifying disconnected connections through link metric calculation. Boats eligible for relay conversion are screened and a candidate relay set is constructed, followed by a policy control method to generate relay task behaviors. Based on the relay boat's behavior output, communication connection relationships are updated and adjacency relationships are adjusted to complete link reestablishment. Calling the policy control method to generate relay task behaviors involves extracting the current state vector of each boat in the constructed candidate set as input, and then invoking a decision control function to output a behavior identifier and action instructions indicating whether to execute the relay task. The decision control function is trained using historical successful relay behavior data. Inputs include energy status, position offset, communication demand intensity, and the remaining time in the current delay budget. The output is a specific behavior combination. The behavior combination includes whether to access the relay network, the target mobile location, whether to transmit the current payload, whether to request backup boat collaboration, and whether to enable low-power communication.

[0007] As a preferred solution of the multifunctional unmanned boat cluster control method for grid deployment described in the present invention, wherein: the acquisition of the communication status data of each unmanned boat includes periodically sampling the number of sent packets, the number of response packets, the average delay and the frame loss records between all unmanned boats in the cluster and the corresponding communication targets, and establishing a communication link status data table. The communication status data table is recorded in units of boat pairs, and the average round-trip delay value, the number of consecutive packet losses, and the maximum response time limit between the boat pairs are accumulated in each sampling period. If it is found that the average packet loss rate of three consecutive periods exceeds the set threshold during the sampling period, it is judged that there is a communication interruption, and the corresponding starting boat number, end boat number, judgment timestamp and the current geographic coordinates of the boat body are recorded to form a link break judgment data set.

[0008] As a preferred solution of the multifunctional unmanned boat cluster control method for grid deployment described in the present invention, the method includes: identifying the interrupted connection through link index calculation, constructing a global communication graph after each round of communication status calculation, and the communication graph is an adjacency matrix structure, and each element value is filled according to the current effective communication signal strength. If an unconnected path appears between any two nodes in the communication graph, a broken link check is automatically performed. The broken link check traverses all connection paths, identifies the non-closed area of the link based on the adjacency matrix, and marks the boat pair corresponding to the path that failed to close as a broken link object. After the marking is completed, the broken link distribution map is recorded based on the real-time position of the boat body and the upstream and downstream connection conditions.

[0009] As a preferred embodiment of the multifunctional unmanned boat swarm control method for grid-based deployment described in the present invention, the method of screening boats that meet the conditions for becoming relays and constructing a candidate relay set includes traversing boats currently in an idle state, obtaining four state attributes: geographic location, remaining energy ratio, density of surrounding connected nodes, and Euclidean distance from the disconnection point, and combining these attributes to form a boat state vector. The candidate boats are sorted according to the state vector, prioritizing boats with remaining energy greater than 60%, distance from the disconnection area less than 50% of the maximum communication range, not currently in a critical mission state, and surrounding density not exceeding a system-set density threshold. This sets the candidate boats that can be scheduled as signal relays, and the set is continuously adjusted as the disconnection area is dynamically updated.

[0010] As a preferred embodiment of the multifunctional unmanned vehicle swarm control method for grid-based deployment described in the present invention, the invocation of a policy control method to generate relay task behaviors includes extracting the current state vector of each vehicle from a constructed candidate set as input. The state vector includes the vehicle's current energy ratio, its positional offset from the disconnection point, the communication density index of the waters it is in, and the system's configured response time budget. The state vector is then fed into a trained decision control function, a behavior-strategy mapping structure constructed using historical relay behavior data as training samples. During training, the ratio of relay effectiveness to energy consumption serves as the core metric of the reward function, and the policy parameters are iteratively updated using a reinforcement learning algorithm. Upon receiving the state vector, the control function outputs a behavior combination for the unmanned vehicle. The behavior combination includes whether to immediately connect to the relay network, move to the target location, allow the current task to be interrupted, enable low-power communication, synchronously request intervention from cooperative vehicles, and retain current load data. The output behavior combination is timestamped and prioritized, and broadcast to all vehicles within the local sensing range via an internal communication channel.

[0011] As a preferred embodiment of the multifunctional unmanned boat cluster control method for grid deployment described in the present invention, the method includes updating the communication connection relationship based on the behavior output of the relay boat, which includes initiating an adjacency detection process after the boat whose behavior output includes the instruction to access the relay network reaches the designated target area, collecting the communication response data of all boats within the coverage area, and constructing a list of the actual adjacent communication range at the current location by sending handshake requests one by one and monitoring the return delay. The adjacency list is compared with the original communication graph. If it is found that the disconnected boat pair has reestablished communication due to the introduction of the unmanned boat, the unmanned boat is inserted into the broken link path, the communication graph is updated, and the original link interruption path is opened. At the same time, the number of connection links, link delays, and node routing tables of the link-repairing path are recorded, and a state backup is performed.

[0012] As a preferred solution of the multifunctional unmanned boat cluster control method for grid deployment described in the present invention, wherein: the adjustment of the adjacency relationship to complete the link reconstruction includes, after the communication graph is updated, broadcasting a synchronous notification to all boats currently connected to the new relay node, and the notification content includes the role status of the relay node, the communication time slot number, the data length allowed for transmission, and the next allowed connection change time. After the new adjacency relationship is established, all boats connected to the relay node complete the adjacency path registration within the specified period. If the line is frequently disconnected, the node will be marked as unstable based on the results of the last two rounds of communication, triggering the substitute candidate signal boat to execute the secondary link filling process. After all adjacency relationship updates are completed, all boats in the broken link path will re-allocate communication time slots and re-issue IDs to complete the final link closed loop.

[0013] Another object of the present invention is to provide a multifunctional unmanned boat cluster control system for grid deployment, which can screen boats that meet the conditions for becoming relays and build a relay candidate set, and call the policy control method to generate relay task behavior, thereby solving the problems of existing unmanned boat cluster control methods such as inaccurate identification of interrupted links, lack of intelligent decision-making mechanism for relay screening, lack of closed-loop repair capability in the link reconstruction process, and how to realize a self-organizing relay reconstruction mechanism driven by a combination of reinforcement learning and behavior.

[0014] As an optimal solution for the grid-deployed multifunctional unmanned boat cluster control system described in the present invention, it includes: a real-time communication status perception and broken link location identification module, a signal boat relay candidate screening and reinforcement learning decision module, and a routing reconstruction and communication chain closed-loop module.

[0015] The communication status real-time perception and link break position identification module is used to obtain the communication status data of each unmanned boat and identify the broken connection through link indicator calculation.

[0016] The signal boat relay candidate screening and reinforcement learning decision module is used to screen boats that meet the conditions for becoming relays and build a relay candidate set, and call the strategy control method to generate relay task behavior.

[0017] The routing reconstruction and communication chain closed-loop module is used to update the communication connection relationship and adjust the adjacency relationship according to the behavior output results of the relay boat to complete the link reconstruction.

[0018] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a multifunctional unmanned boat cluster control method for grid deployment.

[0019] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a multifunctional unmanned boat cluster control method for grid deployment.

[0020] Beneficial effects of the present invention: The multifunctional unmanned boat cluster control method for grid deployment provided by the present invention obtains the communication status data of each unmanned boat, identifies the interrupted connection through link indicator calculation, and enables the system to have low-latency and high-precision link break identification capability, which is a prerequisite for supporting the self-organizing repair mechanism of the unmanned boat network, and improves the perception agility and response reliability of the cluster control system to sudden communication failure events.

[0021] By screening boats that meet the conditions for becoming relays and building a set of relay candidates, the policy control method is called to generate relay task behavior, which greatly improves the adaptability of the unmanned boat relay node configuration and the precision of behavior control, and builds an adjustable, learnable and controllable relay access framework within the cluster, which significantly optimizes the communication repair efficiency and system energy consumption management capabilities.

[0022] According to the behavioral output results of the relay boat, the communication connection relationship is updated and the adjacency relationship is adjusted to complete the link reconstruction, realizing link self-repair and real-time closed-loop adjustment of the communication topology of the unmanned boat cluster in the communication interruption scenario, significantly improving the system operation stability, the flexibility of path management and the consistency of cluster communication scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 The first embodiment of the present invention provides an overall flow chart of a multifunctional unmanned boat cluster control method for grid deployment.

[0025] Figure 2 A system flow chart of a grid-deployed multifunctional unmanned boat cluster control method provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0026] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0027] Example 1, with reference to Figure 1 , as one embodiment of the present invention, provides a multifunctional unmanned boat swarm control method for grid deployment, comprising:

[0028] S1: Obtain the communication status data of each unmanned boat and identify the disconnection through link indicator calculation.

[0029] The number of sent packets, response packets, average delay, and frame loss between all UAVs in the swarm and their corresponding communication targets are periodically sampled to create a communication link status data table. The communication status data table records the average round-trip delay, number of consecutive packet losses, and maximum response time between each UAV pair within each sampling period.

[0030] If the average packet loss rate for three consecutive cycles exceeds the set threshold during the sampling period, it is determined that there is a communication interruption. The corresponding starting boat number, ending boat number, judgment timestamp, and current geographic coordinates of the boat are recorded to form a link break judgment dataset.

[0031] If the average packet loss rate for three consecutive cycles exceeds a set threshold within the sampling period, a preferred solution specifically includes: in the present invention, the set threshold is a packet loss rate greater than 40%. If the average packet loss rate for three consecutive cycles exceeds 40%, it is determined that communication is interrupted. Once a boat pair is identified as having a communication interruption, the event is recorded as a link break event, and a record is generated including the starting boat number and target boat number, the timestamp of the link break determination, the current position coordinates of the starting boat and target boat, the subnet number (if divided into regions), the time of the last normal communication, and its response parameters as reference data.

[0032] After each round of communication state calculation, a global communication graph is constructed. This graph is an adjacency matrix, with each element value filled with 1 or 0 depending on whether the current effective communication signal strength meets a set threshold. If an unconnected path appears between any two nodes in the communication graph, a link break check is automatically performed. This check traverses all connected paths, identifying areas of non-closed links based on the adjacency matrix. Boat pairs corresponding to paths that failed to close are marked as broken links. After marking, a broken link distribution map is recorded based on the boat's real-time position and upstream and downstream connections.

[0033] A preferred approach to constructing a global communication graph involves constructing a global communication graph for the current cycle using all communication status results after each sampling cycle. This graph is represented by an adjacency matrix, where each element in the matrix represents whether a valid communication connection exists between a pair of boats. If the communication quality between the two boats meets a communication quality threshold, the corresponding element in the adjacency matrix is assigned a value of 1; otherwise, it is assigned a value of 0. The communication quality thresholds are shown in Table 1.

[0034] Table 1 Communication quality threshold table

[0035]

[0036]

[0037] The constructed adjacency matrix is used to calculate graph connectivity, identifying all node pairs that cannot communicate with each other through a depth-first traversal. If the paths between certain nodes are broken and no relay paths are available, this is considered a non-closed connection area. For this area, all involved node pairs are extracted, their status recorded, and included in a list of broken link objects. Based on the boat numbers, geographic locations, and original upstream and downstream connection information involved in the broken link objects, a broken link distribution map is generated. This map represents the distribution of failed paths and the scope of the broken link in the current grid communication network, providing basic support data for subsequent relay node scheduling and link replenishment.

[0038] It should be noted that S1 constructs a communication graph through periodic sampling and multi-dimensional communication quality indicators, accurately identifies the broken link relationship between unmanned boats, and combines threshold judgment and graph traversal algorithms to achieve efficient detection and positioning of broken link events, providing quantifiable and traceable data support for subsequent relay chain filling, and improving the system's self-healing capabilities.

[0039] S2: Screen boats that meet the conditions for becoming relays and build a relay candidate set, and call the policy control method to generate relay task behavior.

[0040] Based on the recorded broken link distribution map, the center of the broken link is extracted and used as a reference for the target area for relay screening. The spatial position of all vessels relative to this center is used as the basis for calculations, and the next stage of candidate vessel screening is entered. The process then iterates through the vessels currently in an idle state, obtaining four state attributes: their geographic location, remaining energy ratio, density of connected nodes around them, and Euclidean distance from the broken link point. These attributes are combined to form a vessel state vector. The currently available candidate vessels are sorted based on the state vector, prioritizing vessels with remaining energy greater than 60%, distance from the broken link area less than 50% of the maximum communication range, not currently in a critical mission state, and surrounding density not exceeding the system-set density threshold. This creates a candidate set of vessels currently available for scheduling as signal relays, which is continuously adjusted as the broken link area is dynamically updated.

[0041] For the constructed candidate set, the current state vector of each boat is extracted as input. This state vector contains the current energy ratio of the unmanned boat, the position offset distance from the link break point, the communication demand density index of the water area in which it is located, and the response time budget of the system configuration. The state vector is input into the trained decision-making control function, which is a behavior-strategy mapping structure constructed using historical relay behavior data as training samples. During the training process, the ratio of relay performance to energy consumption is used as the core indicator of the reward function, and the policy parameters are iteratively updated through a reinforcement learning algorithm.

[0042] A preferred method for inputting the state vector into a trained decision-making control function specifically involves constructing a state vector containing multiple dynamic indicators to describe the current state of the vessel and performing normalization. This normalized state vector is then input into the control function, which is trained using reinforcement learning. The training objective is to maximize the expected return between communication recovery capability and energy cost. After training, the function generates a set of actions for each input state, which are used to determine whether to switch to a relay vessel.

[0043] Construct a state vector containing multiple dynamic indicators to describe the current state of the boat, and normalize each state vector:

[0044] V k =[ξ k ,Δ k ,Ω k ,Φ k ]

[0045] Among them, V k Represents the final normalized state input vector. ξ k Indicates the normalized current energy ratio of the unmanned boat. Ω k Represents the normalized communication density of the waters where the unmanned boat is located. Δ k Indicates the distance offset from the center point of the broken chain. Φ krepresents the normalized control response time budget ratio of the unmanned boat. k represents the kth unmanned boat.

[0046] The training goal is to maximize the expected benefit between communication recovery capability and energy cost. The discounted reward function of relay benefit-energy consumption ratio is expressed as:

[0047]

[0048] Among them, n is the long-term expected reward of the relay behavior of the n-th boat. n (t) represents the rate of successful bandwidth recovery of relay at time t. n (t) is the energy consumption per unit time. μ n (t) represents the rate of change of the relay offset path length. α is the energy consumption adjustment factor (set to 0.8 in the present invention). β is the discount factor (set to 0.05 in the present invention). T is the task scheduling window time, which is set to 1 minute in the present invention.

[0049] The state vector is input to the control function, which is trained by reinforcement learning as follows:

[0050]

[0051] in, Represents the PPO loss function. ρ n Represents the ratio of the current strategy to the old strategy. ∈ represents the strategy disturbance tolerance factor (set to 0.1 in this invention).

[0052] The final trained policy control function is expressed as:

[0053]

[0054] Among them, π * (V k ) represents the 6-dimensional behavior combination output. W represents the behavior weight matrix. b represents the bias vector. V k Input vector representing the normalized state. Indicates that the Softmax function is used to normalize the output behavior value. The specific form is:

[0055]

[0056] Among them, each z i Indicates the activation of the i-th corresponding behavior item, and the maximum value corresponds to the behavior being executed. i represents the index of the currently executed behavior item, corresponding to a specific action. j represents the summary of all behavior items, which is another index, indicating the activation compared with other behavior items. jRepresents the activation of the j-th behavior, which is used for comparison and calculation with other behaviors.

[0057] After receiving the state vector, the control function outputs a combination of behaviors for the UAV. These include whether to immediately connect to the relay network, move to the target location, allow the current mission to be interrupted, enable low-power communication, synchronously request intervention from a cooperating vehicle, and retain the current payload data. The output of this behavior is timestamped and prioritized, and broadcasted via an internal communication channel to all UAVs within the local sensing range.

[0058] A preferred solution for whether to immediately access the relay network specifically includes an instruction to immediately enter the relay state if the boat ranks in the top 10% in priority, the energy ratio in the state vector is greater than 0.7, and the expected navigation distance does not exceed 50% of its maximum cruising range.

[0059] A preferred solution for moving the target position specifically includes: including a navigation target point in the behavior function, which is a communication coverage boundary formed by the broken link center point offset outward by a certain radius, and selecting the navigation point using the minimum angle criterion according to the current position of the hull.

[0060] A preferred solution for determining whether to allow the current task to be interrupted specifically includes: if the current task conflict level of the boat is less than 2 and the relay priority is higher than the task priority, the task is allowed to be interrupted and the role is switched. Otherwise, the original task is retained and does not participate in the current round of scheduling.

[0061] Furthermore, the current task conflict level includes levels 1 to 3, and the judgment criteria are as follows:

[0062] Level 1 is low conflict. The mission requires minimal communication resources, and the communication density in the area where the drone is located is low. The mission can continue in its current state without significantly interfering with the relay mission. This level is suitable for routine missions not involving mission-critical operations.

[0063] Level 2 is a moderate conflict. The current task has moderate requirements for communication resources and is located in an area with high communication density. The current task may be affected by relay tasks, but execution can be maintained under certain conditions. This level is suitable for tasks with certain requirements for latency and data transmission.

[0064] Level 3 is high conflict. The mission requires high communication resources, and the UAV's communication bandwidth is severely limited. The mission requires high timeliness and is not suitable for interruption or suspension. This level is suitable for urgent missions, such as real-time data transmission and emergency response.

[0065] A preferred solution for enabling the low-power communication mode specifically includes automatically switching to the low-power communication mode when the hull energy ratio is lower than 0.5, turning off redundant broadcasts, and only maintaining heartbeat frames and local link responses.

[0066] A preferred solution for requesting the intervention of a cooperative boat specifically includes: if the coverage range of a single boat cannot simultaneously access the nodes at both ends of the broken link, a cooperative relay request is broadcast to the nearest boat, and the system retrieves the boat with the second highest priority in the candidate pool to respond to the cooperation.

[0067] A preferred approach to retaining current payload data involves, if the boat is engaged in an important data forwarding task, marking the action instruction as retaining the data, keeping the original cached data intact, and not executing the forwarding action to avoid data loss. If the action combination output by the policy control function includes an instruction to immediately access the relay network, the boat is marked as the initiator of the link replenishment action and enters S3 for access detection and adjacency update.

[0068] It should be noted that S2 constructs a dynamic state vector, integrating multiple factors such as energy, location, communication demand density indicators, and task response constraints to form a quantifiable candidate boat screening mechanism. It also introduces a reinforcement learning training control function to achieve adaptive strategy output for relay behavior. Compared to existing relay scheduling methods that rely on static rules or preset weights, this step has the advantages of being data-driven, behavior-learnable, and strategy-generalizable. It can optimize behavior combinations in real time based on the actual link break location and boat state. Ultimately, a specific relay behavior combination is output through the strategy function, ensuring the system has dynamic response capabilities, energy-saving control capabilities, and multi-node collaborative link replenishment capabilities, significantly improving the self-organizing communication recovery performance and intelligent scheduling level of unmanned boat clusters in complex waters.

[0069] S3: Update the communication connection relationship and adjust the adjacency relationship according to the behavior output of the relay boat to complete the link reconstruction.

[0070] After a boat whose behavior output includes the instruction to access the relay network arrives at the designated target area, an adjacency detection process is initiated to collect communication response data from all boats within the coverage area. By sending handshake requests one by one and monitoring the return delay, a list of actual adjacent communication ranges at the current location is constructed.

[0071] Furthermore, upon access, the relay boat immediately initiates the adjacency detection process. This process broadcasts handshake request frames to each boat within its physical communication range, collecting the confirmation information and response time returned by each responding node. The relay boat records the ID, signal strength (RSSI), and round-trip response delay of each accessible boat to form a list of actual adjacent nodes, and then timestamps this list as a valid window record.

[0072] The adjacency list is compared with the original communication graph. If the introduction of an unmanned boat allows a disconnected boat to reestablish communication, the unmanned boat is inserted into the broken path, the communication graph is updated, and the original interrupted path is reconnected. The number of connected links, link delays, and node routing tables of the reconnected path are recorded, and a state backup is performed. After the communication graph is updated, a broadcast synchronization notification is sent to all boats currently connected to the new relay node. The notification content includes the relay node's role status, communication slot number, allowed data length, and the next allowed connection change time. After the new adjacency relationship is established, all boats connected to the relay node complete the adjacency path registration within the specified period. If the connection is frequently disconnected, the node is marked as unstable based on the results of the last two rounds of communication, triggering the replacement candidate signal boat to execute the secondary link repair process. After all adjacency relationships are updated, all boats in the broken path reassign communication slots and reissue IDs, completing the final link closure.

[0073] A preferred solution for updating the communication graph specifically includes: the communication graph structure is represented by an adjacency matrix. Each time a relay node is connected, the system compares its adjacency list with the original matrix. If the element in the original matrix is 0, and it is found to be connected through the relay node, the element is updated to 1 and synchronized to the routing table structure.

[0074] After all adjacencies are registered, all nodes involved in the broken path will uniformly reissue communication path IDs, reallocate time slots, and adjust broadcast priorities to ensure consistent communication scheduling and fair link scheduling. This completes the link restoration cycle, marking the broken link event as "Closed-loop Processing Completed" and transitioning to the next round of status monitoring.

[0075] It should be noted that S3 achieves automatic repair and closed-loop management of communication paths through adjacency detection after relay boat access, dynamic comparison of communication graphs, link reconstruction, and unstable node replacement mechanisms. Compared to static recovery methods, this method provides real-time link structure perception, adaptive connection updates, and abnormal switching capabilities, significantly improving the self-organizing recovery efficiency and network robustness of the unmanned boat swarm in the event of communication interruptions.

[0076] Example 2 is an embodiment of the present invention, which provides a multifunctional unmanned boat cluster control method for grid deployment. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0077] In order to verify the effectiveness of the adaptive relay behavior control mechanism of the present invention in the communication link disconnection scenario, a set of simulation experimental scenarios was constructed, and the response capabilities and relay behavior rationality of the unmanned boat cluster equipped with the policy control mechanism of the present invention and the traditional unmanned boat based on fixed rule scheduling in the dynamic communication reconstruction task were compared and analyzed. In this experiment, a total of 6 unmanned boats were selected and named as unmanned boats A to F, and a communication path recovery task simulation was carried out in a certain mission area on the sea. The experiment was initially set to have a broken link area, and the system scheduling mechanism needed to complete the relay repair after identifying the broken link point. Before the start of the mission, each unmanned boat obtains and records its current energy ratio, the distance from the geometric center of the broken link area, the communication density in the area, and the current task conflict level, and constructs a candidate boat pool based on the state vector structure proposed by the present invention.

[0078] During the experimental implementation, traditional rules (such as distance priority and energy priority) and the policy control function of the present invention were used to evaluate the status of the unmanned boat and select its behavior. Under traditional rules, the boat with the closest distance and the highest remaining energy is preferentially selected to participate in relaying. However, the mechanism of the present invention integrates indicators such as energy ratio, position offset, communication density, and task conflict level. After normalization and calculation with the scoring function, the reinforcement learning control function outputs a combination of behaviors including whether to access the relay network and whether to allow task interruption. The system outputs a behavior priority queue based on the strategy, selects the top boats to perform the relay action, and records whether they ultimately participate in the relay behavior. The experimental results are shown in Table 2.

[0079] Table 2 Experimental data table

[0080]

[0081] As can be seen from the data in Table 2, in areas with high energy levels (such as UAVs A and E) and relatively low communication density (favorable for signal propagation), UAVs rank high in behavioral priority and successfully perform relay tasks, demonstrating the effective integration of multi-dimensional indicators by the policy control function of this invention. Traditional scheduling methods often force a UAV to perform relaying even when the task conflict level is high, potentially causing task interruptions and control delays. For example, under traditional rules, UAV C might be selected as a relay due to its close proximity. However, due to its high task conflict level, it is rationally excluded in this invention and does not participate in the link-filling task.

[0082] Furthermore, although unmanned boat F is located near the broken link area, its energy ratio and mission conflict level are both unfavorable, resulting in a low overall score and effectively lowering its priority in the policy function. This "multi-factor dynamic adjustment" capability is not available in existing threshold rule-based methods.

[0083] Furthermore, the policy control mechanism of this invention can output combined behaviors, rather than simply selecting "relay or not." For example, it can determine whether to allow task interruption or request collaboration, ensuring scheduling flexibility and system consistency at the system level. In contrast, traditional rules cannot dynamically generate adaptive behavior combinations, and the resulting scheduling often results in wasted resources or link recovery failures.

[0084] In summary, through the correspondence between the multi-boat scores, priorities and actual execution of behaviors in the table, it can be intuitively verified that the state vector modeling and strategy control function proposed in the present invention has the ability of intelligent judgment, multi-objective adaptation and dynamic decision-making in the unmanned boat relay scheduling process, breaking through the limitations of the static single-factor scheduling scheme in the existing technology, and showing higher adaptability, stability and intelligence in complex communication recovery tasks, fully reflecting its creativity and practical value.

[0085] Example 3, reference Figure 2 , which is an embodiment of the present invention, provides a multifunctional unmanned boat cluster control system for grid deployment, including a real-time communication status perception and broken link location identification module 100, a signal boat relay candidate screening and reinforcement learning decision module 200, and a routing reconstruction and communication chain closed-loop module 300.

[0086] S4: The communication status real-time perception and link break location identification module 100 is used to obtain the communication status data of each unmanned boat and identify the broken connection through link indicator calculation.

[0087] The communication status real-time perception and link break location identification module 100 includes a communication data collection submodule 101 and a link status identification submodule 102 .

[0088] Furthermore, the communication data acquisition submodule 101 is used to periodically acquire communication data between each unmanned vehicle and its communication target, including link performance indicators such as the number of packets sent, the number of responses received, the average round-trip delay, and the packet loss ratio, and organize this data into a communication status data table based on the relationship between the two vehicles. The link status identification submodule 102 is used to construct an adjacency communication matrix based on link quality threshold rules and continuous packet loss detection logic within a time window, determine whether there are broken link paths, and extract the broken link node number, broken link path ID, broken link time and spatial location information to generate a broken link determination result.

[0089] It should be noted that the communication data acquisition submodule 101 is the starting point for the real-time communication status perception and link break location identification module 100. Its sampling results directly determine the integrity of the basic data used for link determination. After receiving data, the link status identification submodule 102 is responsible for performing topological connectivity analysis and identifying broken link areas. It serves as the entry point for subsequent relay screening and path repair tasks.

[0090] It should also be noted that the real-time communication status perception and link break location identification module 100 is the starting module of the entire system perception closed loop, providing a data basis and structural prior for subsequent relay behavior decision-making and link reconstruction processes.

[0091] S5: The signal boat relay candidate screening and reinforcement learning decision module 200 is used to screen boats that meet the conditions for becoming relay boats and build a relay candidate set, and call the strategy control method to generate relay task behavior.

[0092] The signal boat relay candidate screening and reinforcement learning decision module 200 includes a relay screening submodule 201 and a behavior control strategy generation submodule 202 .

[0093] Furthermore, the relay screening submodule 201 extracts multiple state characteristics from boats currently in a mission-idle state, such as their energy ratio, spatial distance from the disconnection point, mission conflict level, and communication density, to construct a state vector. A set of relay candidate connections is constructed based on pre-set multi-dimensional optimization rules. The behavior control strategy generation submodule 202 inputs the state vector into a trained strategy control function, outputting a combination of behaviors, including whether to connect to the relay network, whether to allow interruption of the current mission, and whether to request collaboration with collaborative boats. The order of behavior execution is then determined based on behavior priority.

[0094] It should be noted that the relay screening submodule 201 is the entry point for the entire relay behavior generation process. The quality of its candidate set directly affects the effectiveness of the behavior output by the policy control function. The behavior control strategy generation submodule 202 constructs a behavior mapping function based on reinforcement learning optimization methods, enabling data-driven individual intelligent decision-making. This is one of the key innovations of this invention that distinguishes it from traditional relay rules.

[0095] It should also be noted that the signal boat relay candidate screening and reinforcement learning decision module 200 realizes the coupled modeling and dynamic output of the individual state of the unmanned boat and the relay behavior, and is a key module to ensure that the system has intelligence and scenario adaptability.

[0096] S6: The route reconstruction and communication chain closed-loop module 300 is used to update the communication connection relationship and adjust the adjacency relationship according to the behavior output result of the relay boat to complete the link reconstruction.

[0097] The route reconstruction and communication chain closed-loop module 300 includes an adjacency update submodule 301 and a path reconstruction confirmation submodule 302 .

[0098] Furthermore, the adjacency update submodule 301 is used to initiate adjacency detection after the relay boat enters the target area. Using a handshake request and response feedback mechanism, it identifies boat nodes within its actual communication range, updates the adjacency matrix, and incorporates the new relay node into the communication path structure. The path reconstruction confirmation submodule 302 analyzes changes to the communication graph, determines whether the broken path is closed, and, if successful, broadcasts the relay role status and channel assignment notification to the affected nodes, completing routing table synchronization and reissuing the communication path ID.

[0099] It should be noted that the adjacency update submodule 301 is responsible for performing local perception at the network structure layer and is the entry point for confirming path feasibility. The path reconstruction confirmation submodule 302 completes the closed-loop confirmation of the communication structure and provides anomaly detection and automatic link repair functions, ensuring the system's high robustness in dynamic environments.

[0100] It should also be noted that the routing reconstruction and communication chain closed-loop module 300 not only completes the physical layer connection of the broken link repair process, but also realizes the connection status synchronization of the control layer. It is the core guarantee for the stable operation of the entire system and link consistency.

[0101] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0102] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0103] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0104] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logical functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc. It should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention and are not intended to be limiting. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced with equivalents without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications should be encompassed by the claims of the present invention.

Claims

1. A multifunctional unmanned boat cluster control method for grid deployment, characterized in that: include: Obtain the communication status data of each unmanned boat and identify disconnected connections through link indicator calculation; Screen boats that meet the conditions for becoming relays and build a relay candidate set, and call the policy control method to generate relay task behavior; Update the communication connection relationship and adjust the adjacency relationship according to the behavior output of the relay boat to complete the link reconstruction; Calling the strategy control method to generate relay mission behavior includes extracting the current state vector of each boat as input for the constructed candidate set, and outputting the behavior identification and action instructions of whether to execute the relay mission by calling the decision control function; the decision control function is trained with historical successful relay behavior data, and the input includes energy state, position offset, communication demand intensity, and the remaining time of the current delay budget, and the output is a specific behavior combination; the behavior combination includes whether to access the relay network, the mobile target position, whether to transmit the current payload, whether to request backup boat collaboration, and whether to enable low-power communication.

2. The multifunctional unmanned boat swarm control method for grid deployment according to claim 1, characterized in that: The obtaining of the communication status data of each unmanned boat includes: Periodically sample the number of sent packets, number of response packets, average delay, and frame loss records between all unmanned boats in the cluster and the corresponding communication targets, and establish a communication link status data table; The communication status data table records the data in pairs of boats, and accumulates the average round-trip delay, number of consecutive packet losses, and maximum response time between boat pairs in each sampling period; If the average packet loss rate for three consecutive cycles exceeds the set threshold during the sampling period, it is determined that there is a communication interruption. The corresponding starting boat number, ending boat number, judgment timestamp, and current geographic coordinates of the boat are recorded to form a link break judgment dataset.

3. The multifunctional unmanned boat swarm control method for grid deployment according to claim 1 or 2, characterized in that: The identifying of disconnected connections by link indicator calculation includes: After each round of communication state calculation, a global communication graph is constructed. The communication graph is an adjacency matrix structure, and each element value is filled according to the current effective communication signal strength; If a disconnected path appears between any two nodes in the communication graph, a link break check is automatically performed; The broken link check traverses all connection paths, identifies the non-closed link areas based on the adjacency matrix, and marks the boat pairs corresponding to the paths that fail to close as broken link objects; After the marking is completed, the broken chain distribution map is recorded based on the real-time position of the hull and the upstream and downstream connection conditions.

4. The multifunctional unmanned boat swarm control method for grid deployment according to claim 3 is characterized in that: The screening of boats that meet the conditions for becoming relays and constructing a relay candidate set includes: Traverse the boats currently in the mission idle state, obtain the idle unmanned boat's geographical location, remaining energy ratio, surrounding connected node density, and Euclidean distance to the disconnection point, and combine these attributes to form the boat state vector; The currently available candidate boats are sorted according to the state vector, with priority given to boats with a remaining energy greater than 60%, a distance from the broken link area less than 50% of the maximum communication range, not currently in a critical mission state, and a surrounding density not exceeding the system-set density threshold. This constructs a candidate set of boats that can currently be scheduled as signal relays, and the set is continuously adjusted as the broken link area is dynamically updated.

5. The multifunctional unmanned boat swarm control method for grid deployment according to any one of claims 1, 2 or 4, characterized in that: The calling strategy control method to generate relay task behavior includes: For the constructed candidate set, the current state vector of each boat is extracted as input. The state vector includes the current energy ratio of the unmanned boat, the position offset distance from the disconnection point, the communication demand density index of the water area, and the response time budget of the system configuration. The state vector is input into the trained decision control function. The decision control function is a behavior strategy mapping structure constructed using historical relay behavior data as training samples. During the training process, the ratio of relay effect to energy consumption is used as the core indicator of the reward function, and the strategy parameters are iteratively updated through the reinforcement learning algorithm. After receiving the state vector, the control function outputs a behavior combination for the UAV. The behavior combination includes whether to immediately access the relay network, move the target location, allow the current mission to be interrupted, enable low-power communication, synchronously request the intervention of the cooperative boat, and retain the current load data. The behavior combination output is accompanied by a timestamp and decision priority, and is broadcast to the boats within the local perception range through the internal communication channel.

6. The multifunctional unmanned boat swarm control method for grid deployment according to claim 5, characterized in that: The updating of the communication connection relationship according to the behavior output result of the relay boat includes: When a boat whose behavior output includes the command to access the relay network arrives at the designated target area, it initiates the adjacency detection process to collect communication response data from all boats within the coverage area. By sending handshake requests one by one and monitoring the return delay, it constructs a list of actual adjacent communication ranges at the current location. The adjacency list is compared with the original communication graph. If it is found that the broken link boat pair has reestablished communication due to the introduction of the unmanned boat, the unmanned boat will be inserted into the broken link path, the communication graph will be updated and the original link interruption path will be opened. At the same time, the number of connection links, link delay, and node routing table of the link-repairing path will be recorded, and state backup will be performed.

7. The multifunctional unmanned boat swarm control method for grid deployment according to any one of claims 1, 2, 4 or 6, characterized in that: The adjusting the adjacency relationship to complete the link reconstruction includes: After the communication graph is updated, a broadcast synchronization notification is sent to all boats currently connected to the new relay node. The notification content includes the role status of the relay node, the communication time slot number, the allowed data length for transmission, and the next allowed connection change time; After the new adjacency relationship is established, all boats connected to the relay node complete the adjacent path registration within the specified period. If the connection is frequently disconnected, the node will be marked as unstable based on the results of the last two rounds of communication, triggering the replacement candidate signal boat to execute the secondary link filling process. After all adjacency relationships are updated, all boats in the broken link path re-allocate communication time slots and re-issue IDs to complete the final link closure.

8. A multifunctional unmanned boat swarm control system for grid deployment, characterized by: It includes a communication status real-time perception and broken link location identification module (100), a signal boat relay candidate screening and reinforcement learning decision module (200), and a routing reconstruction and communication chain closed loop module (300); The communication status real-time perception and broken link position identification module (100) is used to obtain the communication status data of each unmanned boat and identify the broken connection through link index calculation; The signal boat relay candidate screening and reinforcement learning decision module (200) is used to screen boats that meet the conditions for becoming relay boats and construct a relay candidate set, and call a strategy control method to generate a relay task behavior; The routing reconstruction and communication chain closed-loop module (300) is used to update the communication connection relationship and adjust the adjacency relationship according to the behavior output result of the relay boat to complete the link reconstruction.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the multifunctional unmanned boat cluster control method for grid deployment described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the multifunctional unmanned boat cluster control method for grid deployment according to any one of claims 1 to 7 are implemented.

Images