Methods and systems for collaborative data interaction between unmanned surface vessels and unmanned aerial vehicles in air-ground converged communication

Abstract

This invention discloses a method and system for collaborative data interaction between unmanned surface vessels (USVs) and unmanned aerial vehicles (UAVs) using air-ground integrated communication, belonging to the field of collaborative communication technology. The system includes a sensing end composed of UAVs and USVs, a fusion base station, and a shore-based host computer. The method includes: the sensing end collecting multi-source data; the host computer optimizing and compressing the sensing data and calculating data scoring indicators; adopting a hierarchical collaborative scheduling architecture based on reinforcement learning algorithms, where the upper layer generates collaborative goals and the lower layer uses an attention mechanism to filter key states; and dynamically adjusting the parameters of the UAVs and USVs in real time through a reinforcement learning closed loop based on the communication link status and task requirements. This invention effectively solves the problem of unstable single communication links in complex aquatic environments, significantly improving the robustness of data transmission and task execution efficiency through a "perception-decision-execution-optimization" closed loop, and is suitable for scenarios such as nearshore patrol, inland waterway monitoring, and emergency rescue.

Description

Technical Field

[0001] This invention belongs to the field of unmanned equipment collaborative communication technology, specifically relating to a method and system for collaborative data interaction between unmanned surface vessels and unmanned aerial vehicles (UAVs) in air-ground integrated communication. Background Technology

[0002] In numerous fields such as water monitoring, emergency rescue, and marine surveying, the collaborative operation of unmanned surface vessels (USVs) and intelligent unmanned devices such as drones has become an indispensable key link. The primary prerequisite for achieving effective collaboration among these intelligent devices is solving the problem of stable communication between the unmanned devices and between the devices and the remote control center. However, the data transmission methods used by traditional USVs have many unavoidable limitations. Specifically, while satellite communication can achieve global coverage, it suffers from high costs, large transmission delays, and limited bandwidth resources, and its signal is easily interfered with and unstable in complex environments. 4G network signal coverage is extremely unstable at sea or in remote waters, and its transmission rate cannot meet the demands of efficient data transmission. Short-range radio communications such as WiFi are limited by their effective transmission distance, making it difficult to support long-distance data transmission. It is worth noting that drones operating alone face similar data transmission challenges: in remote or complex environments such as oceans and mountainous areas, 4G / 5G base stations are sparsely distributed, and WiFi signals are not available, significantly reducing the effectiveness of drone data transmission.

[0003] In traditional applications, unmanned surface vessels (USVs) and unmanned aerial vehicles (UAVs) often operate relatively independently in water-related scenarios. Taking surface inspection and water quality testing as examples, USVs are primarily responsible for collecting information on floating debris and water impurity content in specific water areas; while UAVs, leveraging their aerial advantage, can monitor large areas of water, but are limited by sensor accuracy, making it difficult to accurately collect information from localized areas or obtain crucial data such as water composition. The failure to establish an effective collaborative working model between the two hinders the full utilization of their respective technological advantages.

[0004] Traditionally, communication links between unmanned surface vessels (USVs) and unmanned aerial vehicles (UAVs) are deployed independently, which not only limits data transmission capabilities but also makes it difficult to build efficient collaborative communication links. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the existing technology in which the communication links of unmanned surface vessels and unmanned aerial vehicles are deployed independently, resulting in low communication transmission capacity, so as to provide a method and system for collaborative data interaction between unmanned surface vessels and unmanned aerial vehicles with air-ground integrated communication.

[0006] A method for collaborative data interaction between unmanned surface vessels (USVs) and unmanned aerial vehicles (UAVs) in air-to-ground integrated communication includes the following steps: Step S1: The sensing end collects sensing data; the sensing end includes at least one aerial drone and at least one unmanned surface vessel (USV); the USV and the drone collect and transmit video, image and sensor data through a self-organizing network and a fusion communication link between the USV and the ground host computer to form sensing data. Step S2: Data preprocessing and quality assessment: The host computer receives the sensing data and optimizes and compresses the sensing data; the data scoring index is calculated based on the compressed sensing data; if the data scoring index is lower than a preset threshold, the optimization and compression parameters are adjusted to ensure that the sensing data meets the availability requirements under the minimum transmission rate. Step S4: Layered Cooperative Scheduling of Reinforcement Learning Algorithm: Construct a near-end policy optimization model, generate cooperative goals based on global state through an upper-level coordinator, and control the specific actions of the unmanned surface vessel and the unmanned aerial vehicle through a lower-level actuator; Step S5: Dynamic strategy execution and closed-loop optimization: Based on the output of the lower-level actuator, adjust the motion and relay parameters of the UAV and the data sampling and transmission parameters of the unmanned surface vessel in real time; input the feedback status back into the near-end strategy optimization model to form an adaptive closed loop of "perception-decision-execution-optimization".

[0007] Furthermore, it also includes the following steps: Step S3: In the geographic information system platform, the optimized and compressed perception data is fused with the spatiotemporal trajectories of the unmanned surface vessel and the drone to generate a comprehensive situational view.

[0008] Furthermore, it also includes the following steps: Based on the comprehensive situational view, the water area is divided into a UAV-dominated area, an unmanned surface vessel-dominated area, and a collaborative boundary area; in the UAV-dominated area, the unmanned surface vessel-dominated area, and the collaborative boundary area, different preset strategies are used to allocate sensing resources.

[0009] Furthermore, the preset strategy includes: In the drone-dominated area, the drones are given priority for wide-area monitoring; In the unmanned surface vessel (USV)-dominated area, the USV is given priority in being assigned to conduct water surface sampling. In the collaborative boundary area, the collaborative relay mode is activated, with the UAV acting as an airborne relay node, establishing a dedicated short-range data link with the unmanned surface vessel, aggregating the perception data of the UAV and the unmanned surface vessel, and transmitting it back through the fusion base station.

[0010] Furthermore, the control strategy of the lower-level actuator for the UAV and the unmanned surface vessel includes: For the UAV, an attention mechanism is embedded to prioritize the filtering of key variables such as channel state and buffer state, and the output action vector includes altitude adjustment, planar position fine-tuning, relay frequency band, data transmission rate and transmit power. For the unmanned surface vessel, the output motion vector includes fine-tuning of waypoints, adjustment of sampling rate of non-critical sensors, adjustment of data priority, and adjustment of communication attitude.

[0011] Furthermore, the proximal policy optimization model employs a dynamic weighting mechanism of core rewards and synergistic rewards, with the reward function expressed as: ; in, The core reward is calculated based on communication quality, task completion rate, and device energy consumption. The collaborative reward is calculated based on the relay success rate and regional coverage effectiveness of the UAV, with α representing the weighting coefficient.

[0012] Furthermore, the integrated communication link integrates 5G communication, satellite communication, terrestrial WiFi, and D2D communication; the system monitors the link status in real time, and when it detects that the current link signal strength is below the threshold or is congested, the algorithm automatically switches the frequency band or communication mode and dynamically adjusts the altitude of the UAV to avoid water surface obstruction.

[0013] Furthermore, the adaptive adjustment of the unmanned surface vessel includes: reducing the sampling specifications of unnecessary sensors when communication bandwidth is insufficient to ensure the transmission of core video streams and control commands; and the unmanned aerial vehicle (UAV) management relay buffer prioritizes forwarding high-priority data based on data value tags.

[0014] A collaborative data interaction system for unmanned surface vessels and unmanned aerial vehicles (UAVs) with integrated air-ground communication, used to implement the aforementioned collaborative communication method, characterized in that it includes: Sensing end: including at least one aerial drone and at least one unmanned surface vessel, equipped with high-definition cameras, environmental sensors and multimodal communication modules for data acquisition and D2D networking; Converged base station: As a ground data receiving hub, it supports 5G, WiFi and satellite signal access, and is used to realize data aggregation and forwarding; The host computer processing center is equipped with a GPU acceleration module, a GIS platform, and a reinforcement learning decision-making module. The host computer processing center is configured to: receive data forwarded by the fusion base station, perform video preprocessing and quality assessment, run a near-end strategy optimization model and generate control commands, and send the adjustment commands to the sensing end through the fusion base station.

[0015] Furthermore, the communication connection and interaction process of the above system components includes: the fusion base station communicates with the host computer and the UAV respectively; the host computer communicates with the fusion base station, the UAV and the unmanned surface vessel respectively; the UAV communicates with the fusion base station, the host computer and the unmanned surface vessel respectively. During operation, the fusion base station reports the link status to the host computer, and the host computer sends frequency band switching instructions and rate adjustment parameters to the fusion base station; the host computer obtains the altitude, speed and mission level from the UAV and sends a status adjustment instruction to the UAV; the unmanned surface vessel sends its speed and position to the host computer; the control center (host computer) sends a service flow priority flag to the unmanned surface vessel, and the unmanned surface vessel sends a rate adjustment ACK to the control center.

[0016] Furthermore, the reinforcement learning decision module includes: State space input interface: used to receive the status of UAV, unmanned surface vessel, communication link, and environment; Policy networks and value networks: used to output the probability distribution of cooperative actions and evaluate the value of the current state; Command conversion module: Converts the action vectors output by the algorithm into specific device control commands.

[0017] Beneficial effects: This invention utilizes GPUs to reconstruct and compress video data, effectively reducing data transmission bandwidth consumption. It also combines SSIM, VMAF, or MOS metrics to detect data transmission quality in real time, triggering a feedback mechanism when metrics are abnormal. This solves the problems of limited bandwidth, high latency, and unreliable data quality in traditional data transmission. At the same time, it uses a geographic information system platform to construct multi-dimensional spatiotemporal data, linking and integrating device trajectories with corresponding time-based sensing data. This enables spatiotemporal and structured data management, providing precise data support for collaborative decision-making processes.

[0018] Based on the Proximal Policy Optimization (PPO) algorithm, a hierarchical control architecture of "upper-level coordinator-lower-level actuator" is constructed. The upper level outputs high-level collaborative goals, while the lower level precisely optimizes the operating status of equipment, overcoming the limitations of poor equipment coordination and insufficient control precision in traditional independent operations. In particular, the lower-level actuator can control the flight parameters of UAVs and the sensor sampling rate of UAVs in real time, realizing efficient collaborative linkage between UAVs and UAVs, improving the quality of task completion, optimizing collaborative control precision, and improving response speed.

[0019] By dividing the water area into UAV-dominated areas, unmanned surface vessel-dominated areas, and collaborative boundary areas through a geographic information system platform, and by allocating sensing resources in a targeted manner according to the operational characteristics of different areas, the advantages of UAVs in large-scale monitoring and the advantages of unmanned surface vessels in close-range and precise data collection are fully utilized. Furthermore, by integrating base stations to support multi-network access of 5G, WiFi, and satellite signals, the problem of the inability to fully utilize the advantages of equipment in traditional operations is solved, the rationality of resource allocation is improved, and the operational coverage is expanded.

[0020] By introducing a reinforcement learning mechanism and embedding an attention mechanism into the drone control, key channel states are prioritized and selected. Based on communication feedback information, the drone's altitude, pose, and relay frequency band are adjusted in real time. This can dynamically adapt to the communication needs in remote areas, areas with poor public network coverage, or complex water environments. It solves the pain points of traditional communication methods in complex environments where signals are easily interfered with and communication performance is unstable, thus enhancing the stability and adaptability of communication in complex environments.

[0021] By using a closed-loop feedback mechanism of "perception-decision-execution-optimization" to perceive communication performance in real time, and combining reinforcement learning mechanism and hierarchical control strategy, a dynamic optimization link of data acquisition-processing-decision-adjustment is formed. This not only ensures the stability of collaborative communication, but also adaptively adjusts the equipment status according to changes in the operation scenario, improves the versatility and flexibility under different water environments and operation tasks, and realizes dynamic adaptive optimization of communication and operation. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic flowchart of the collaborative communication method of the present invention; Figure 2 This is a schematic diagram of the real-time decision-making process of the near-end policy optimization algorithm of the present invention; Figure 3 This is a schematic diagram of the collaborative communication system structure of the present invention; Figure 4 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0024] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0025] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0026] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "joining," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0027] Example 1: Reference Figure 1 and Figure 2 As shown, this embodiment provides a method for collaborative data interaction between unmanned surface vessels (USVs) and unmanned aerial vehicles (UAVs) in air-to-ground integrated communication, including the following steps: Step S1: Multimodal data acquisition and multi-link transmission at the sensing end; the sensing end includes at least one aerial UAV and at least one unmanned surface vessel (USV); during operation, the USV and the UAV acquire and transmit multimodal sensing data, including video, images, and sensor data, through a self-organizing network and a fused communication link between the USV and a ground-based host computer; specifically, the USV and the UAV acquire high-definition video, images, GNSS positioning information, meteorological / hydrological sensor data, etc.; the USV and the UAV transmit the aforementioned sensing data back to the host computer through a wireless self-organizing network and a fused communication link between the USV and the shore-based host computer. The host computer has a built-in multi-channel data buffer module to achieve synchronous reception and buffering, providing a stable data source for the communication link.

[0028] Multi-link fusion and switching: The fused communication link integrates 5G communication, satellite communication, terrestrial WiFi and D2D communication; the system monitors the link status in real time, and when the current link signal strength is detected to be lower than the threshold or congested, the algorithm automatically switches the frequency band or communication mode, and dynamically adjusts the altitude of the UAV to avoid water surface obstruction.

[0029] Step S2: Intelligent Preprocessing, Quality Assessment, and Temporal Integration: The host computer receives the perceived data and uses the GPU to perform super-resolution reconstruction and intelligent compression on the video stream data to reduce the backhaul bandwidth pressure. A data scoring index is calculated based on the compressed perceived data; specifically, Structural Similarity (SSIM), Video Multi-Method Evaluation Fusion (VMAF), and Mean Subjective Opinion Score (MOS) are used to objectively assess video quality. If the data scoring index is lower than a preset threshold, a system feedback mechanism is triggered to adjust the optimization and compression parameters, ensuring that the perceived data meets the availability requirements under the minimum transmission rate. The preprocessed perceived data is seamlessly integrated with the sensor data stream through classification, labeling, and temporal alignment.

[0030] Step S3: GIS Spatiotemporal Multidimensional Fusion and Dynamic Resource Allocation: The Geographic Information System (GIS) platform acquires the preprocessed data and forms multidimensional spatiotemporal data based on the preprocessed data. The spatiotemporal multidimensional data includes the trajectories of UAVs and unmanned surface vessels and the corresponding time-sensing data. As a further improvement to this embodiment, the geographic information system platform integrates spatiotemporal multidimensional data to achieve unified overlay display of unmanned surface vessel / unmanned aerial vehicle (USV) trajectory, attitude, and environmental parameters, generating a real-time interactive comprehensive situational view; it further supports operators to make manual decisions and interventions based on hot zones, historical data, and abnormal events.

[0031] Specifically, in terms of dynamic task collaboration, the geographic information system divides the unmanned surface vessel (USV)-dominated area, unmanned aerial vehicle (UAV)-dominated area, and collaborative area according to the characteristics of the water area, and adaptively allocates roles; in terms of communication perception path planning, offline trajectory pre-assessment of communication link feasibility is adopted, and online replanning is based on reinforcement learning feedback, which fine-tunes the UAV / USV pose to optimize relay quality while ensuring the core measurement tasks; in terms of intelligent sampling linkage, USV dynamically downgrades the sampling specifications of non-critical sensors according to communication bandwidth and task priority to alleviate transmission pressure, while UAV intelligently manages the relay buffer, prioritizes forwarding high-value data, and ensures that critical data meets the backhaul requirements.

[0032] The adaptive adjustment of the unmanned surface vessel includes: reducing the sampling specifications of unnecessary sensors when communication bandwidth is insufficient to ensure the transmission of core video streams and control commands; and the unmanned aerial vehicle management relay buffer prioritizes forwarding high-priority data based on data value tags.

[0033] The geographic information system platform divides the water area into UAV-dominated areas, unmanned surface vessel-dominated areas, and collaborative boundary areas, and allocates sensing resources based on the water area division; Based on the comprehensive situational view, the water area is divided into a UAV-dominated area, an unmanned surface vessel-dominated area, and a collaborative boundary area; in the UAV-dominated area, the unmanned surface vessel-dominated area, and the collaborative boundary area, different preset strategies are used to allocate sensing resources.

[0034] In some embodiments of this example, water area demarcation may include: The main areas for drones are those that are far from the shore, with open waters, and require rapid, large-scale coverage (such as search and rescue, patrols); The main areas for unmanned surface vessels are those that are close to the shore, have complex water conditions, and require detailed underwater / surface data (such as water quality sampling and nearshore mapping). The collaborative boundary zone is an area with moderate offshore distance, moderate environmental conditions, and mission requirements that combine "high-altitude + surface" methods (such as both large-scale monitoring and localized detailed sampling).

[0035] In this embodiment, when allocating sensing resources, drones are made to be more inclined towards drone-dominated areas and less inclined towards unmanned surface vessels (USVs)-dominated areas, while USVs are made to be more inclined towards USV-dominated areas and less inclined towards drone-dominated areas.

[0036] In the drone-dominated area, the drones are given priority for wide-area monitoring; In the unmanned surface vessel (USV)-dominated area, the USV is given priority in being assigned to conduct water surface sampling. In the collaborative boundary area, the collaborative relay mode is activated, with the UAV acting as an airborne relay node, establishing a dedicated short-range data link with the unmanned surface vessel, aggregating the perception data of the UAV and the unmanned surface vessel, and transmitting it back through the fusion base station.

[0037] Step S4: Hierarchical Cooperative Scheduling and Offline / Online Integration Based on PPO Algorithm: Construct a Proximal Policy Optimization (PPO) model. The upper-level coordinator generates cooperative objectives based on the global state, and the lower-level actuators control the specific actions of the unmanned surface vessel (USV) and the unmanned aerial vehicle (UAV). Based on the Proximal Policy Optimization algorithm, the upper-level coordinator outputs high-level cooperative objectives based on the current global state in the spatiotemporal multidimensional data, and the lower-level controller optimizes the control of the UAV and the USV. Specifically, the PPO algorithm architecture is used to achieve collaborative scheduling of unmanned surface vessels (USVs) and unmanned aerial vehicles (UAVs). The upper-level coordinator generates collaborative goals based on the global state; the lower-level actuator has separate policy modules for UAVs and USVs. For drones, key states are filtered based on an embedded attention mechanism: the output action vector includes: dynamically adjusting flight altitude / position, relay strategy, data rate, and transmission power; For unmanned surface vessels, the output action vector includes: dynamically adjusting the waypoints based on the link status, adjusting the sampling rate of non-critical sensors, prioritizing data transmission, and fine-tuning the path points within a safe range to optimize communication attitude.

[0038] The model uses a dynamic weighting of core rewards and collaborative rewards to balance communication quality, task execution, and system energy consumption.

[0039] Step S5: Dynamic Interaction, Business Flow Adaptation and Closed-Loop Optimization: Based on the output of the lower-level actuator, adjust the motion and relay parameters of the UAV and the data sampling and transmission parameters of the UAV in real time, including the UAV's flight altitude, position, relay frequency and the sampling rate of non-critical sensors of the UAV and the data transmission priority; input the feedback status back into the near-end strategy optimization model.

[0040] For the UAV, an attention mechanism is embedded to prioritize the filtering of key variables such as channel state and buffer state, and the output action vector includes altitude adjustment, planar position fine-tuning, relay frequency band, data transmission rate and transmit power. For the unmanned surface vessel, the output motion vector includes fine-tuning of waypoints, adjustment of sampling rate of non-critical sensors, adjustment of data priority, and adjustment of communication attitude.

[0041] For UAVs, the embedded attention mechanism adjusts the UAV's flight altitude, pose, and intermediate frequency based on channel state and buffer; for unmanned surface vessels, the sensor sampling rate and data transmission priority are adjusted, and the waterway points are adjusted based on communication angle.

[0042] In the near-end policy optimization algorithm, base stations, drones, and unmanned surface vessels are integrated as the perception layer to provide state data; the host computer (deploying the PPO model) serves as the decision-making layer to generate policies, translate decisions into specific instructions, and collect feedback data. The remote control center acts as the control layer to issue instructions.

[0043] Signaling interaction process: The fused base station reports link status to the host computer, and the host computer sends frequency band switching instructions and rate adjustment parameters to the fused base station; the host computer obtains altitude, speed, and mission level from the UAV, obtains status information from the UAV, and sends altitude adjustment instructions to the UAV; the fused base station sends retransmission policy configuration to the UAV, and the UAV sends a new altitude confirmation signal to the fused base station; the unmanned surface vessel (USV) sends its speed and position to the host computer and receives service flow priority markers from the control center, and sends a rate adjustment ACK to the control center; the host computer inputs a state vector to the near-end policy optimization model, and the near-end policy optimization model outputs the optimal action to the host computer.

[0044] The link status includes reference signal received power, bandwidth, and delay.

[0045] The communication feedback information of the UAV and the unmanned surface vessel is obtained. If the communication performance in the communication feedback information is lower than a preset threshold, the altitude and trajectory of the UAV are adjusted in real time based on the communication feedback information, and the pose of the unmanned surface vessel is adjusted, based on reinforcement learning. Specifically, during the offline phase, the feasibility of communication links along historical paths is evaluated; during online operation, if the communication performance is below a threshold, the reinforcement learning module fine-tunes the drone's altitude and trajectory or the unmanned surface vessel's attitude in real time based on feedback information, optimizing relay quality while meeting the measurement task and safe altitude requirements.

[0046] Adaptive degradation of service flow: The unmanned surface vessel adjusts the sensor sampling specifications based on real-time bandwidth and mission criticality; the unmanned aerial vehicle manages the relay buffer and schedules forwarding according to data value priority.

[0047] Specifically, when communication bandwidth is insufficient, the unmanned surface vessel (USV) adaptively downgrades the sampling specifications of non-essential sensors based on real-time bandwidth and mission criticality, prioritizing the transmission of core video streams or control commands. Simultaneously, the USV intelligently manages the relay buffer, prioritizing the forwarding of high-priority data based on data value tags. Finally, the new transmitted status data is used as feedback to re-input the PPO model of the host computer, and a new round of transmitted data is reinjected into the processing flow, forming a closed loop of "perception-decision-execution-feedback," continuously optimizing system performance.

[0048] The architecture of this embodiment includes a sender and a receiver.

[0049] The unmanned surface vessel (USV) and drone in the transmitting end collect data including images, environmental perception, and location information through cameras, sensors, and GPS, and then stream the data via a streaming module. The communication module transmits the collected data and relays it to the receiving end via a drone.

[0050] The receiving end includes a host computer and a control center. The data receiving module receives signals relayed back from the UAV and buffers the data for subsequent processing. The host computer uses a GPU to perform preprocessing operations such as super-resolution reconstruction and intelligent compression on the video data, effectively alleviating bandwidth pressure and improving subsequent processing efficiency. The transmitted data is evaluated from multiple dimensions using metrics such as Structural Similarity (SSIM), Video Multi-Method Evaluation Fusion (VMAF), or Mean Subjective Opinion Score (MOS) fed back from the receiving end. Data meeting the requirements undergoes spatiotemporal fusion, integrating temporal and spatial information to provide comprehensive quality support for decision-making, and is then sent to the reinforcement learning module. The reinforcement learning module generates optimization decisions based on the evaluation metrics fed back from the receiving end. In the optimization decision-making process, the receiver's rate, frequency band, and altitude are adjusted via instructions, while task objectives are set for lower-level actuators, achieving task-level collaborative scheduling.

[0051] Specifically, in this embodiment, to achieve the hierarchical collaborative scheduling described in step S4, a hierarchical control architecture based on Proximal Policy Optimization (PPO) is constructed. The upper-layer coordinator mainly consists of a policy network of the PPO model, responsible for generating high-dimensional collaborative policies based on the global state; the lower-layer executor is responsible for mapping the policies to specific device control actions and embedding an attention mechanism to optimize feature extraction of key states.

[0052] The near-end policy optimization model consists of a policy network and a value network. The policy network incorporates a multi-head attention mechanism to prioritize key variables such as channel state and buffer state when processing UAV states, assigning them higher weights to output the optimal action probability distribution. The value network evaluates the value of the current state to assist the policy network in optimization.

[0053] The state space design for constructing the proximal policy optimization model covers all dimensions of information, including UAV state, such as perception, communication, and task. These include the UAV's three-dimensional position, flight speed, remaining battery power, relay buffer occupancy rate, and whether it is in relay status; and the UAV's status. These include the unmanned surface vessel's two-dimensional position, speed, remaining battery power, sampling rate of non-critical sensors, data transmission priority, and mission area coverage; as well as the communication link status. These are, respectively, the reference signal received power, real-time bandwidth, transmission delay, packet loss rate, video structural similarity, video subjective score, and video multi-method evaluation fusion index; task status. These include task type, task priority, task completion progress, and whether it is in a critical hotspot; environmental status. These are wind speed, wave height, and visibility, respectively. Normalizing and concatenating the above states, we obtain the global state integration vector as follows: This serves as input to the upper-level coordinator (policy network).

[0054] The action integration vector in the action space is represented as The specific execution is carried out by the lower-level actuators; among them, the drone's motion space The corresponding height adjustment, planar position fine-tuning, relay band switching, data transmission rate, and transmission power adjustment are described in the claims; the unmanned surface vessel's maneuvering space. This corresponds to the track point fine-tuning, non-critical sensor sampling rate adjustment, data priority adjustment, and communication attitude adjustment described in the claims.

[0055] The reward function design employs a dynamic weighted mechanism of core reward + synergistic reward to guide the strategy optimization objective. ; Among them, core rewards This is used to focus on communication quality and task execution, and is represented as: ; Communication quality awards Based on link-state quantization, it is represented as: ; Where max_bw is the maximum bandwidth and max_delay is the maximum acceptable delay.

[0056] Task execution rewards Based on task progress and key data feedback, it is represented as follows: ; Energy consumption penalty To prevent the device's battery from running out quickly, it is represented as: ; Collaborative Rewards Used to focus on equipment coordination efficiency, expressed as: ; Relay coordination reward This is used to reward unmanned surface vessels (USVs) for improved data transmission success rates when they act as relays, and is expressed as: ; Regional coverage collaborative rewards For regions defined by a geographic information system (UAV-dominated areas / Unmanned surface vessel-dominated areas / cooperative boundary areas), reward equipment for efficient operation in the corresponding areas is represented as follows: ; Higher rewards are given when the device is located in the UAV-dominant area, USV-dominant area, or collaborative boundary area defined by GIS and performs the corresponding task.

[0057] In the strategy update logic, UAVs and unmanned surface vessels interact with the environment under the current strategy, collecting multiple sets of "state-action-reward-next state" data. The reward R is based on the dynamic weighted formula of the core reward and collaborative reward described in claim 3. The calculation yields the value V(S) of the network's current state, which is then evaluated using temporal differencing to estimate the advantage function A(S, a). The advantage function measures the degree of advantage of an action relative to the average level and is expressed as follows: Where R is the reward for the current step, γ is the discount factor, and V(S') is the estimated value of the next state.

[0058] The policy loss function uses a clip mechanism to limit the update magnitude of the old and new policies, preventing excessively rapid policy updates from causing convergence instability. The loss function is expressed as: ; Among them, the ratio of the new strategy to the old strategy , The clip parameter is used to limit the magnitude of policy updates. This term represents entropy and is used to encourage the exploration of diverse actions. is the entropy regularization coefficient.

[0059] Minimize the value network loss function, expressed as: .

[0060] In other embodiments of this example, the hyperparameters of the near-end strategy optimization model can be adaptively designed according to the actual needs of the task.

[0061] Working Principle: In this embodiment, high-definition video and sensor data are collected at the sensing end → the host computer uses GPU for super-resolution reconstruction and quality assessment (SSIM / VMAF / MOS) (step S2) → the geographic information system fuses the trajectory and data to generate a situation map and divide the dominant area → the operator can intervene in decision-making as needed → the reinforcement learning module integrates the communication link status (RSRP / bandwidth / latency), equipment status, and task objectives to generate hierarchical collaborative optimization instructions → the UAV / unmanned surface vessel executes actions such as altitude adjustment, relay frequency band switching, and sampling rate adjustment → a new round of communication and feedback status forms a closed-loop feedback, driving reinforcement learning to continuously optimize the strategy. This process forms an adaptive collaborative closed loop of "perception-decision-execution-optimization", significantly improving the efficiency of water operations and robustness in complex environments.

[0062] The air-to-ground integrated communication method for collaborative data interaction between unmanned surface vessels (USVs) and unmanned aerial vehicles (UAVs) provided in this embodiment monitors the communication link status in real time through algorithms and dynamically adjusts the data transmission strategy based on parameters such as signal strength, bandwidth usage, and transmission delay. Specific adjustment measures include: adjusting the UAV's flight altitude to avoid water surface obstructions, adaptively switching between 5G / WiFi / satellite or D2D relay frequency bands, and dynamically changing the data transmission rate and sampling rate of non-critical sensors. This ensures that the execution of core tasks is not interrupted due to communication blind spots or congestion. Compared to traditional methods with fixed parameters, this method significantly improves the continuity of task execution and optimizes the overall operational efficiency of the air-to-ground collaborative system.

[0063] Example 2: Reference Figure 3 and Figure 4 As shown, this embodiment provides a collaborative data interaction system for unmanned surface vessels (USVs) and unmanned aerial vehicles (UAVs) with integrated air-ground communication, used to implement the collaborative communication method described in Embodiment 1. The system is characterized by comprising: a fusion base station, a host computer, a UAV, and an USV; the fusion base station is communicatively connected to the host computer and the UAV; the host computer is communicatively connected to the fusion base station, the UAV, and the USV; and the UAV is communicatively connected to the fusion base station, the host computer, and the USV.

[0064] The specific components of the system include: Sensing end: including at least one aerial drone and at least one unmanned surface vessel, equipped with high-definition cameras, environmental sensors and multimodal communication modules for data acquisition and D2D networking; Converged base station: As a ground data receiving hub, it supports 5G, WiFi and satellite signal access, and is used to realize data aggregation and forwarding; The host computer processing center is equipped with a GPU acceleration module, a GIS platform, and a reinforcement learning decision-making module. The host computer processing center is configured to: receive data forwarded by the fusion base station, perform video preprocessing and quality assessment, run a near-end strategy optimization model and generate control commands, and send adjustment commands (including altitude, trajectory and sampling rate) to the sensing end through the fusion base station.

[0065] The host computer processing center is configured to: receive data forwarded by the fusion base station, perform video preprocessing and quality assessment, run the near-end policy optimization (PPO) model and generate control commands, and send adjustment commands (including altitude, trajectory and sampling rate) to the sensing end through the fusion base station.

[0066] The reinforcement learning decision module specifically includes: State space input interface: used to receive UAV status (position, speed, battery level, buffer), unmanned surface vessel status (position, sampling rate), communication link status (RSRP, bandwidth, latency, SSIM, MOS) and environmental status (wind speed, wave height); Policy networks and value networks: used to output the probability distribution of cooperative actions and evaluate the value of the current state; Command conversion module: Converts the action vectors output by the algorithm into specific device control commands.

[0067] In the system's signaling interaction and operation mechanism: the fused base station, as the data transmission hub, forwards uplink status and sensing data to the host computer processing center, and issues frequency band switching instructions and rate adjustment parameters generated by the host computer; the host computer processing center, as the system's adaptive decision control center, obtains altitude, speed, and mission level from the UAV, runs reinforcement learning algorithms to generate altitude adjustment instructions and issues them through the fused base station, and simultaneously obtains speed and position from the unmanned surface vessel; the UAV obtains the aforementioned adjustment instructions from the host computer processing center and sends a new altitude confirmation signal; the unmanned surface vessel obtains service flow priority markers from the host computer processing center and sends a rate adjustment ACK.

[0068] In the actual operation of this collaborative communication system, each component plays the following collaborative advantages: the UAV, with its flexibility and maneuverability, can quickly adjust its flight attitude and position in the air, optimize the communication link with the unmanned surface vessel, and at the same time use its own high-performance communication module as a data aggregation node to transmit the fused multi-source data back to the base station at high speed and stably through the communication network. Its flight altitude can be dynamically adjusted according to the communication demand command issued by the host computer, effectively avoiding communication blockage on the water surface and expanding the data transmission coverage.

[0069] Unmanned surface vessels (USVs) are dedicated to water monitoring missions. Their sensors collect a wide variety of data, covering multi-dimensional water information. Although their communication modules have limited power, once a dedicated short-range data link is established with unmanned aerial vehicles (UAVs), they can continuously transmit monitoring data to the UAVs in real time, ensuring the original collection and initial aggregation of core water data. Furthermore, the USVs can leverage their inherent stability to provide UAVs with a long-term, continuous surface data source. Simultaneously, they can also directly transmit core and critical data back to a shore-based host computer processing center via a converged communication link.

[0070] The converged base station possesses high-bandwidth, low-latency signal processing capabilities. It can simultaneously receive large amounts of data transmitted from multiple drones, perform efficient signal conversion and forwarding, ensuring high-speed and high-reliability data transmission to the shore-based host computer. Its network architecture supports dynamic resource allocation, prioritizing data channels for collaborative operation systems based on task load strategies issued by the host computer processing center, maintaining the smoothness of the entire water monitoring data chain.

[0071] In summary, the collaborative communication system between surface and aerial equipment provided in this embodiment not only solves the problem of reliable long-distance data transmission from unmanned surface vessels (USVs), but also fully leverages the advantages of integrating multiple communication technologies. Through the reinforcement learning optimization and control decision-making module built into the host computer processing center, a strict "perception-decision-execution-feedback" closed loop is formed, greatly improving the robustness and resilience of the closed-loop video stream transmission system in complex environments. The system can transmit high-definition video data in real time and accurately transmit the location information of the air and surface equipment, achieving true air-ground collaborative operations and providing comprehensive data support for emergency rescue, wide-area patrols, and other missions.

[0072] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0073] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for collaborative data interaction between unmanned surface vessels (USVs) and unmanned aerial vehicles (UAVs) in air-to-ground integrated communication, characterized in that, Includes the following steps: The sensing end collects sensing data; the sensing end includes at least one aerial drone and at least one unmanned surface vessel; the unmanned surface vessel and the drone collect and transmit sensing data, including video, images and sensor data, through a self-organizing network and a fusion communication link between the unmanned surface vessel and the ground host computer. The host computer receives the sensing data and optimizes and compresses it; it calculates a data scoring index based on the compressed sensing data; if the data scoring index is lower than a preset threshold, it provides feedback to adjust the optimization and compression parameters so that the sensing data meets the availability requirements under the minimum transmission rate. A near-end strategy optimization model is constructed, which generates a cooperative target based on the global state through an upper-level coordinator and controls the specific actions of the unmanned surface vessel and the unmanned aerial vehicle through a lower-level actuator. Based on the output of the lower-level actuator, the motion and relay parameters of the UAV and the data sampling and transmission parameters of the unmanned surface vessel are adjusted in real time; the feedback status is input into the near-end strategy optimization model.

2. The method for collaborative data interaction between unmanned surface vessels and unmanned aerial vehicles (UAVs) in air-to-ground integrated communication according to claim 1, characterized in that, It also includes the following steps: The optimized and compressed perception data is fused with the spatiotemporal trajectories of the unmanned surface vessel and the unmanned aerial vehicle in the geographic information system platform to generate a comprehensive situational view.

3. The method for collaborative data interaction between unmanned surface vessels and unmanned aerial vehicles (UAVs) in air-to-ground integrated communication according to claim 2, characterized in that, It also includes the following steps: Based on the comprehensive situational view, the water area is divided into a UAV-dominated area, an unmanned surface vessel-dominated area, and a collaborative boundary area; in the UAV-dominated area, the unmanned surface vessel-dominated area, and the collaborative boundary area, different preset strategies are used to allocate sensing resources.

4. The method for collaborative data interaction between unmanned surface vessels and unmanned aerial vehicles (UAVs) in air-to-ground integrated communication according to claim 3, characterized in that, The preset strategy includes: In the drone-dominated area, the drones are given priority for wide-area monitoring; In the unmanned surface vessel (USV)-dominated area, the USV is given priority in being assigned to conduct water surface sampling. In the collaborative boundary area, the collaborative relay mode is activated, with the UAV acting as an airborne relay node, establishing a dedicated short-range data link with the unmanned surface vessel, aggregating the perception data of the UAV and the unmanned surface vessel, and transmitting it back through the fusion base station.

5. The method for collaborative data interaction between unmanned surface vessels and unmanned aerial vehicles (UAVs) in air-to-ground integrated communication according to claim 1, characterized in that, The control strategy of the lower-level actuator over the UAV and the unmanned surface vessel includes: For the UAV, an attention mechanism is embedded to prioritize the filtering of key variables such as channel state and buffer state, and the output action vector includes altitude adjustment, planar position fine-tuning, relay frequency band, data transmission rate and transmit power. For the unmanned surface vessel, the output motion vector includes fine-tuning of waypoints, adjustment of sampling rate of non-critical sensors, adjustment of data priority, and adjustment of communication attitude.

6. The method for collaborative data interaction between unmanned surface vessels and unmanned aerial vehicles (UAVs) in air-to-ground integrated communication according to claim 1, characterized in that, The proximal policy optimization model employs a dynamic weighting mechanism of core rewards and synergistic rewards, and the reward function is expressed as: ； in, The core reward is calculated based on communication quality, task completion rate, and device energy consumption. The collaborative reward is calculated based on the relay success rate and regional coverage effectiveness of the UAV, with α representing the weighting coefficient.

7. The method for collaborative data interaction between unmanned surface vessels and unmanned aerial vehicles (UAVs) in air-to-ground integrated communication according to claim 1, characterized in that, The integrated communication link integrates 5G communication, satellite communication, terrestrial WiFi and D2D communication; the system monitors the link status in real time, and when the current link signal strength is detected to be lower than the threshold or congested, the algorithm automatically switches the frequency band or communication mode and dynamically adjusts the altitude of the UAV.

8. The method for collaborative data interaction between unmanned surface vessels and unmanned aerial vehicles (UAVs) in air-to-ground integrated communication according to claim 1, characterized in that, The adaptive adjustment of the unmanned surface vessel includes: reducing the sampling specifications of unnecessary sensors when communication bandwidth is insufficient to ensure the transmission of core video streams and control commands; and the unmanned aerial vehicle management relay buffer prioritizes forwarding high-priority data based on data value tags.

9. A collaborative data interaction system for unmanned surface vessels and unmanned aerial vehicles (UAVs) with integrated air-ground communication, used to implement the collaborative communication method according to any one of claims 1 to 8, characterized in that, include: Sensing end: including at least one aerial drone and at least one unmanned surface vessel, equipped with high-definition cameras, environmental sensors and multimodal communication modules for data acquisition and D2D networking; Converged base station: As a ground data receiving hub, it supports 5G, WiFi and satellite signal access, and is used to realize data aggregation and forwarding; The host computer processing center is equipped with a GPU acceleration module, a GIS platform, and a reinforcement learning decision-making module. The host computer processing center is configured to: receive data forwarded by the fusion base station, perform video preprocessing and quality assessment, run a near-end strategy optimization model and generate control commands, and send the adjustment commands to the sensing end through the fusion base station.

10. A collaborative data interaction system for unmanned surface vessels and unmanned aerial vehicles (UAVs) with integrated air-to-ground communication as described in claim 9, characterized in that, The reinforcement learning decision module includes: State space input interface: used to receive the status of UAV, unmanned surface vessel, communication link, and environment; Policy networks and value networks: used to output the probability distribution of cooperative actions and evaluate the value of the current state; Command conversion module: Converts the action vectors output by the algorithm into specific device control commands.

Images