Communication and sensing integrated communication method for collaborative operation of intelligent agricultural equipment

Abstract

The invention belongs to the technical field of smart agriculture, and particularly relates to a communication and sensing integrated communication method for collaborative operation of smart agricultural equipment. Comprising the following steps: step 1, alternately training by a re-GDQN method and an IDRL method, and jointly obtaining an optimal unmanned aerial vehicle hovering point, a sensing beam width and an unmanned aerial vehicle trajectory; 2, based on KD-DDPG wave beam excitation current weight optimization, a full-connection network is adopted as a network model of an actor and a critic network, and the area array specification of the unmanned aerial vehicle and the target directions of a sensing wave beam and a communication wave beam at different time points are input into the model. According to the method, reduction of energy consumption and improvement of communication and sensing beam performance are considered at the same time, coverage limitation is introduced to construct a multi-target optimization problem, and minimization of energy consumption of the unmanned aerial vehicle and maximization of the signal-to-noise ratio of the sensing beam are achieved on the premise that it is ensured that the sensing beam completely covers the designated farmland area.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture technology, specifically to a sensor-integrated communication method for collaborative operation of smart agricultural equipment. Background Technology

[0002] With the development and application of communication technology, data in farmland scenarios is becoming increasingly diversified. Real-time monitoring of farmland operations and environmental information is a current necessity in farmland management. Drones, as aerial base stations, have gained significant attention in recent years due to their flexible deployment capabilities. Specifically, drones can monitor data such as soil moisture, crop status, and agricultural machinery location in real time by sensing sensor information in the farmland, thus promoting the development of smart agriculture. In this application scenario, Integrated Sensing and Communication Technology (ISAC) allows sensing and communication to occur simultaneously. Therefore, some current research combines beamforming with ISAC technology to achieve collaborative operation between devices.

[0003] For the application and performance optimization of sensing integration based on UAV base stations in various scenarios, domestic and foreign research has focused on the efficient use of limited UAV resources and the optimization of sensing performance, aiming to reduce the interference between sensing beams and communication beams. However, the codebook design for sensing beams to scan target areas to achieve regional sensing is mostly based on hierarchical codebook design, and does not consider the relationship between the effective coverage of the constructed beam and the movement of mobile user positions and the interference between communication and sensing beams. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] To address the aforementioned technical problems, according to one aspect of the present invention, the present invention provides the following technical solution: A sensor-integrated communication method for collaborative operation of smart agricultural equipment includes the following steps: Step 1: The optimal UAV hovering point, sensing beamwidth and UAV trajectory are jointly obtained by alternating training using the re-GDQN method and the IDRL method. Step 1.1: Propose the IDRL model. IDRL combines diffusion modeling with reinforcement learning, and uses the Q-value to guide the update of the diffusion model to generate the optimal solution; Step 1.2: Regarding the selection of beamwidth, based on the characteristics of the hierarchical codebook, the selectable beamwidth is expressed as a finite number, with the minimum coverage width corresponding to a coverage area radius of 2. nFor a finite set Ψ, a re-GDQN model is proposed to obtain the optimal hovering position and corresponding beamwidth of a UAV. A GNN graph neural network is proposed as the Q-network for reinforcement learning. The message passing layer of the GNN is first reduced in dimensionality by the CNN convolutional layer and then an attention mechanism is added to obtain features to ensure the model is lightweight while improving the feature-grabbing ability of the processor. In addition, the center point position and whether it is covered by the grid are input into the re-GDQN model, and IDRL is called to obtain the current optimal trajectory according to the optimal access order of the generated center points. The reward is obtained according to the trajectory length and coverage efficiency to guide the Q-network update and obtain the optimal UAV hovering position. Step 2: Optimize beam excitation current weights based on KD-DDPG. A fully connected network is used as the network model for both the actor and critic networks. The array specifications of the UAV, the sensing beam and the communication beam, and the target directions at different time points are input into the model to optimize the excitation current weights of the sensing beam and the communication beam. A self-distillation module is added to the critic network to assist in training.

[0006] As a preferred embodiment of the integrated sensing communication method for collaborative operation of smart agricultural equipment described in this invention, the alternating training method in step one is as follows: First, the IDRL model is trained so that it can obtain the optimal UAV trajectory for a set of hovering positions and corresponding communication, sensing beamwidths and target directions, so that it can avoid overlap between sensing beams and communication beams while minimizing distance; after p rounds, the IDRL model is fixed and the re-GDQN method is trained so that it can obtain reasonable UAV hovering points and corresponding sensing beamwidths for irregular farmland, so that the UAV's movement energy consumption is minimized while the sensing beam can scan the entire target area.

[0007] As a preferred embodiment of the integrated sensing communication method for collaborative operation of smart agricultural equipment described in this invention, in step 1.1, the proposed IDRL uses the reverse generation process of the diffusion model as the policy network for reinforcement learning. In each training process, the diffusion model network composed of U-Net generates the probability of selecting each path, and then uses the H operator to convert the probability matrix into a Hamiltonian loop. Subsequently, the corresponding Q value is calculated and passed to guide the update of the diffusion model, thereby obtaining the optimal UAV access order so that the UAV movement energy consumption is minimized and the target direction of the sensing beam and the communication beam overlaps.

[0008] As a preferred embodiment of the integrated sensing communication method for collaborative operation of smart agricultural equipment described in this invention, in step 1.1, the environment in which the intelligent agent is located includes the hovering position X specified by the UAV, the sensing beam coverage range φt corresponding to the UAV's hovering position, and the target position aligned with the communication beam in each time slot. and the corresponding communication beam coverage area , and the selected Hamiltonian cycle A represented by the adjacency matrix tsp_t The movement trajectory of the agricultural machinery, the time allocation T of the overall environment, and the time step parameter t required by the diffusion model are used to represent the environmental state space as follows: , where A t For the corresponding state s in different models t The action of making a selection; The agent's state space is set as a continuous space consisting of edge probability distributions of floating-point numbers from 0 to 1, and the reward function is set as follows: , where R collection This is an overlapping penalty.

[0009] As a preferred embodiment of the integrated sensing communication method for collaborative operation of smart agricultural equipment described in this invention, in step 1.2, the description of the intelligent agent environment first introduces a quadtree to divide the target farmland area into uniform grids. The side length of the leaf node of the divided grid is equal to the side length of the inscribed square of the beam coverage area corresponding to the smallest beam that the UAV can construct. Based on this, the environment is constructed as a graph structure, specifically including two types of nodes: the center point of the grid divided by the quadtree and the center point corresponding to the area covered by the sensing beam. The edges include three types: grid point-grid point, coverage area-grid point, and coverage area-coverage area. The graph structure shows the topological relationship.

[0010] As a preferred embodiment of the integrated sensor communication method for collaborative operation of smart agricultural equipment described in this invention, in step 1.2, the state is represented as follows: Where G is the grid topology, V is the node feature matrix containing the coordinates and coverage status of nodes, E is the edge relationship represented by the adjacency matrix, and A is the action mask matrix, identifying the legal action space. The action space is discrete and represents the set of grid centers of any coverage width in the current state that can be connected to the covered sensing area. Each action is represented as... node i The center coordinates of the selected grid belonging to the optional action space. This refers to the beamwidth selected for the corresponding location. The reward is returned to the agent only after the re-GDQN completely covers the target area, i.e., sparse reward is set. , where μ is the coverage efficiency of the currently selected scheme, which is the area of ​​the covered region and the total area covered by the currently set sensing beam, and D is the total energy consumption of the drone calculated by the optimal access order under the current scheme obtained by IDRL.

[0011] As a preferred embodiment of the sensor-integrated communication method for collaborative operation of smart agricultural equipment described in this invention, in step two, the state is represented as... ,in The depression and elevation angles represent the target's angle of arrival corresponding to the sensing or communication beam. This indicates the depression and elevation angles of the target arrival angle for the communication / sensing beam in the same time slot. This is the selected action, and the action space is represented as a 1×N vector consisting of floating-point numbers from 0 to 1, where N=N x ×N y , Construct the reward function The optimal excitation current weight allocation for the sensing beam and communication beam under different time slots is obtained sequentially. At this point, the drone hovering position and the corresponding sensing beam width, the drone access sequence, i.e. the drone trajectory and the excitation current weight of the beam are optimized.

[0012] Compared with existing technologies, the advantages of this invention are as follows: This invention simultaneously considers reducing energy consumption and improving the performance of communication and sensing beams, and introduces coverage constraints to construct a multi-objective optimization problem, which is then decomposed into three sub-problems, and corresponding algorithms are proposed for collaborative optimization of each. Specifically, the proposed re-GDQN model and IDRL model are trained alternately to jointly optimize the hovering position of the UAV, the sensing beamwidth at each position, and the UAV's access sequence (i.e., the UAV trajectory); based on this, the KD-DDPG algorithm is further used in each time slot to optimize the excitation current weight configuration of the sensing and communication beams to improve the beam signal-to-noise ratio (SNR). In terms of methodology, the IDRL algorithm uses a diffusion model as the policy network to generate actions, while the re-GDQN algorithm models the environment as a graph structure and introduces convolutional layers to construct a DQN network. Both serve as lightweight models, and through alternating training, they determine the optimal UAV trajectory and related beam parameters, resulting in faster convergence and shorter training time. Subsequently, the KD-DDPG algorithm, through fine-tuning the excitation current weights, ensures that the gain of the communication beam and the sensing beam in each other's directions is close to zero, thereby effectively suppressing mutual interference. Ultimately, while ensuring complete coverage of the designated farmland area by the sensing beam, the algorithm minimizes UAV energy consumption and maximizes the signal-to-noise ratio of the sensing beam. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a time allocation diagram for the integrated sensing and communication method for collaborative operation of smart agricultural equipment according to the present invention. Figure 2 This is an overall flowchart of a sensor-integrated communication method for collaborative operation of smart agricultural equipment according to the present invention; Figure 3 This is a flowchart of the alternating training of the model in step one of the sensor-integrated communication method for collaborative operation of smart agricultural equipment according to the present invention. Figure 4 This is a flowchart of the IDRL training process for a sensor-integrated communication method for collaborative operation of smart agricultural equipment according to the present invention. Figure 5 This is a flowchart of the re-GDQN training process for a sensor-integrated communication method for collaborative operation of smart agricultural equipment according to the present invention. Figure 6 This is a flowchart of the KD-DDPG training process for a sensor-integrated communication method for collaborative operation of smart agricultural equipment according to the present invention. Detailed Implementation

[0014] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0015] Secondly, the present invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of the present invention. In addition, actual fabrication should include three-dimensional spatial dimensions of length, width, and depth.

[0016] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0017] This invention proposes a sensor-integrated communication method for collaborative operation of smart agricultural equipment. Specifically, firstly, an arbitrary polygon is divided into a uniform grid using a quadtree algorithm. Then, improved re-GDQN and IDRL are alternately optimized to optimize the hovering position, corresponding sensing beamwidth, and drone access sequence (i.e., drone trajectory) of the drone in each flight cycle. This allows the drone to fully perceive the entire target farmland area through sensing beam scanning within a single cycle, while minimizing drone movement energy consumption while avoiding overlap between the sensing and communication beams. Subsequently, for each hovering point of the drone, based on zero-forcing beamforming, the proposed KD-DDPG is used to optimize the beam weights of the sensing and communication beams, further reducing interference between them.

[0018] The method of this invention includes: alternating training of the proposed re-GDQN model and the proposed IDRL model to obtain the optimal UAV hovering position, the UAV's sensing beamwidth at the corresponding position, and the UAV's access order, i.e., the UAV trajectory; subsequently, for each time slot, the excitation current weight configuration of the sensing and communication beams is optimized using the KD-DDPG model to improve the beam signal-to-noise ratio. This invention models and deconstructs the multi-objective NP-hard problem of reducing energy consumption and improving communication and sensing beam performance into three sub-problems, and proposes alternating training of the IDRL algorithm, the re-GDQN algorithm, and the KD-DDPG algorithm to solve these problems respectively. The IDRL algorithm and the re-GDQN algorithm jointly optimize the drone trajectory and the width of the sensing beam corresponding to the drone hovering point. The IDRL algorithm uses the diffusion model as the policy network generation action, and the re-GDQN algorithm models the environment as a graph structure and combines it with convolutional layers to form the DQN algorithm. The two lightweight models alternately optimize to obtain the optimal drone hovering position, the drone's sensing beam width at the corresponding position, and the drone's access order, i.e., the drone trajectory. Then, the KD-DDPG algorithm is used to optimize the excitation current weight of the beam, so that the gain of the communication beam and the sensing beam in each other's directions approaches 0, thereby reducing the interference between them. Finally, it can minimize the drone's energy consumption and maximize the sensing beam SNR while ensuring that the sensing beam can completely cover the designated farmland.

[0019] Specifically, in the scenario of this invention, the drone actively constructs a sensing beam to perform sensing scans and perceive farmland environmental information and the locations of sensors requiring linking. This invention treats the target farmland as a polygon. Each time the drone hovers directly above the sensing codebook, it constructs a sensing beam. Simultaneously, the drone constructs a communication beam to communicate with the moving farm machinery. That is, the construction of each sensing beam is synchronized with the construction of the communication beam. Furthermore, during each beam sensing operation, the drone hovers directly above the perceived target's location. Then, in the next time slot, the drone moves to the next target sensing direction. The specific time allocation is as follows: Figure 1 As shown. Based on this, the present invention proposes a scheme for sensing codebook, trajectory optimization, and dynamic adjustment of beam weights, the process of which is as follows: Figure 2 As shown.

[0020] This invention optimizes the 3dB width of the sensing beam, the UAV trajectory, and the beam excitation current weight to maximize communication rate while ensuring optimal sensing performance and minimizing UAV energy consumption. In the scenario, the UAV uses an Nx×Ny array to transmit ISAC signals and receives reflected ISAC signals from the sensing target to perceive farmland. Simultaneously, the UAV communicates with agricultural machinery using its Nx×Ny array. Since this invention focuses on a farmland scenario, the channel is modeled as a Loss-of-Stake (LoS) ideal channel, and the channel models for the sensing and communication beams are independent. Specifically, for any chosen transmit beam forming a communication beam with weight wcom, the signal received at the agricultural machinery can be expressed as: (1) Where P represents the transmission power. , represents the channel coefficient vector, N=Nx×Ny, where x is the normalized power carrying information signal, i.e., E[|x[n]]. 2 |]=1, w c Let n be the communication beamforming vector. c This represents additive white Gaussian noise, satisfying... ,in For transmitting target information from drones to agricultural machinery, The information transmitted by the drone to other non-target agricultural machinery mainly consists of reference information regarding the direction of the sensing beam in this scenario. Therefore, the signal-to-noise ratio received at the target agricultural machinery is: (2) For the aforementioned ISAC system, beam scanning is considered for target sensing, where a UAV acts as an aerial base station, constructing a directional, ground-perpendicular beam. The UAV moves over time to a location of interest to achieve target area sensing. The scanning scheme allows the UAV to sense over a wide area, and precise beam guidance each time helps to achieve greater beamforming gain in the direction of interest while reducing clutter from other directions. The sensing process lasts for a specified time Δt. Therefore, for each duration... For time t, the sensing beam weight w is selected. t The sensing beam receives the following signal at the target direction: (3) in This is represented as the channel coefficient vector corresponding to the sensing beam, where θ and ψ are the elevation and depression angles of the angle of arrival (AoA) and the angle of departure (AoD), respectively. It is assumed that AoA and AoD are the same because the target distance is usually much larger than the array aperture. s w is the reference signal for normalized power. s Let n be the sensing beam forming vector. s This represents additive white Gaussian noise, satisfying... Therefore, the beam gain for target sensing can be expressed as: (4) The corresponding echo signal-to-noise ratio of the perceived target is: (5) The energy loss of a drone during hovering and movement can be modeled as follows: (6) Where PH represents the drone's hovering energy consumption, and PF represents its mobile energy consumption D. i The distance traveled in the i-th time slot is calculated as follows: (7) To reduce drone energy consumption while maintaining sensing performance, the optimization objectives are as follows: Objective 1: Improving communication quality requires maximizing the SNR between the base station and the user; this is the primary optimization objective. Therefore, our first objective can be expressed as: (8) Objective Two: Return rate is a commonly used metric for evaluating sensing performance in ISAC systems. Therefore, the second objective aims to maximize the return rate of the target system, expressed as: (9) Objective 3: Drones are energy-constrained aerial platforms; reducing their energy consumption can extend the operational time of the systems under consideration. Therefore, the third objective focuses on minimizing drone energy consumption, expressed as follows: (10) Based on the above optimization variables and objectives, the multi-objective optimization problem can be described as follows: (11) In the proposed scheme, the target angles of the sensing and communication beams constructed by the UAV need to be in different directions. If the other two beams both point to the same area, the dual-beam configuration will degenerate into a single beam, causing the sensing and communication functions to only match one, resulting in a performance degradation of that function. Furthermore, all areas of the target farmland need to be scanned by the sensing beam, and the UAV's movement distance within a time slot must comply with its speed limits. This is an NP-hard problem, and the three objectives are interconnected. To simplify the problem, this invention proposes a phased, lightweight optimization framework based on reinforcement learning.

[0021] A sensor-integrated communication method for collaborative operation of smart agricultural equipment includes the following steps: Step 1: Joint optimization of sensing beamwidth and UAV trajectory: In the first stage, the proposed re-GDQN method and IDRL method are trained alternately to jointly obtain the optimal UAV hovering point, sensing beamwidth, and UAV trajectory. Specifically, in the proposed scheme, IDRL is first trained so that it can obtain the optimal UAV trajectory for a set of hovering positions and corresponding communication, sensing beamwidths, and target directions, minimizing distance while avoiding overlap between sensing and communication beams. After p rounds, the IDRL model is fixed, and the re-GDQN method is trained so that it can obtain reasonable UAV hovering points and corresponding sensing beamwidths for irregular farmland, minimizing UAV movement energy consumption while allowing the sensing beam to scan the entire target area. The algorithm flow is as follows: Figure 3 As shown.

[0022] Step 1.1: Optimization of Drone Access Sequence Based on IDRL: In this section, the present invention proposes an IDRL model to minimize UAV trajectory while avoiding overlap between the per-second coverage areas of the sensing and communication beams. IDRL combines a diffusion model with reinforcement learning, using Q-values ​​to guide the diffusion model updates to generate the optimal solution. The proposed IDRL uses the reverse generation process of the diffusion model as the policy network for reinforcement learning. In each training iteration, the diffusion model network, composed of U-Nets, generates probabilities for selecting various paths. The H operator is then used to convert the probability matrix into a Hamiltonian loop. Subsequently, the Q-value corresponding to that "action" is calculated and passed to guide the diffusion model updates, thereby obtaining the optimal UAV access sequence that minimizes UAV movement energy consumption and prevents overlap between the target directions of the sensing and communication beams. The specific process is as follows: Figure 4 As shown.

[0023] In the algorithm, the environment in which the agent is located includes the hovering position X specified by the UAV, the sensing beam coverage area φt corresponding to the UAV's hovering position, and the target position aligned with the communication beam in each time slot. and the corresponding communication beam coverage area , and the selected Hamiltonian cycle A represented by the adjacency matrix tsp_t The movement trajectory of the agricultural machinery, the time allocation T of the overall environment, and the time step parameter t required by the diffusion model. Therefore, the environmental state space can be represented as: (12) The agent's state space is set as a continuous space consisting of a continuous probability distribution of floating-point numbers from 0 to 1.

[0024] Based on a combination of Formulas 6 and 10, in this part, the invention sets the reward function as follows: (13) Where R collection The overlap penalty is that the present invention adds the overlap restriction in Formula 10 to the reward function in the form of a penalty function to guide the agent to make decisions.

[0025] Step 2.2: Optimization of UAV hovering position and sensing beamwidth based on re-GDQN: In this section, regarding the selection of beamwidth, based on the characteristics of the hierarchical codebook, the selectable beamwidth is expressed as a finite number, with the minimum coverage width corresponding to a coverage area radius of 2. n The finite set Ψ is multiple of the given set. To simplify the calculation, the scanning problem of the sensing beam is modeled as a square coverage problem in this section, that is, the coverage range of the beam is simplified to the inscribed square of the corresponding coverage area.

[0026] Specifically, this invention proposes a re-GDQN model to obtain the optimal hovering position and corresponding beamwidth of a UAV. In this method, for describing the agent's environment, this invention first introduces a quadtree to divide the target farmland area into a uniform grid. The side length of each leaf node in the quadtree is equal to the side length of the inscribed square of the beam coverage area corresponding to the smallest beam that the UAV can construct. Based on this, the environment is constructed as a graph structure, specifically containing two types of nodes: the center point of the grid divided by the quadtree, and the center point corresponding to the area covered by the sensing beam. Edges include three types: grid point-grid point, coverage area-grid point, and coverage area-coverage area, with the graph structure indicating the topological relationships. Based on this, this invention proposes a GNN (Graph Neural Network) as the Q-network for reinforcement learning. The message passing layer of the GNN is first dimensionality-reduced by a CNN convolutional layer before an attention mechanism is added to acquire features, ensuring the model's lightweight nature while improving its feature-grabbing ability. In addition, the center point positions and whether they are covered are input into the re-GDQN model, and IDRL is called to obtain the current optimal trajectory based on the optimal access order of the generated center points. A reward is obtained based on the trajectory length and coverage efficiency to guide the Q-network update and obtain the optimal drone hovering position. The specific process is as follows: Figure 5 As shown.

[0027] In the above scheme, the state is represented as: (14) Where G represents the grid topology, V is the node feature matrix, containing the coordinates of nodes and whether they are covered, E represents edge relationships, expressed by an adjacency matrix, and A is the action mask matrix, identifying the legal action space. In this scheme, the action space is discrete, representing a set of grid centers of arbitrary coverage width in the current state, which can be connected to the covered sensing area. Each action is represented as: (15) node i The center coordinates of the selected grid belonging to the optional action space. This refers to the beamwidth selected at the corresponding location. In this scenario, the sensing beam needs to fully cover the target area. Therefore, the re-GDQN returns a reward to the agent only after it has fully covered the target area; that is, a sparse reward is set, which is specifically expressed in Formula 10 as follows: (16) Where μ represents the coverage efficiency of the currently selected scheme, which is the ratio of the covered area to the total area covered by the currently set sensing beam, and D represents the total power consumption of the UAV calculated from the optimal access order under the current scheme obtained by IDRL. Based on this model, re-GDQN training obtains the optimal UAV hovering position and the corresponding sensing beam coverage range, so that the sensing beam constructed by the UAV minimizes the UAV's power consumption while fully scanning the target area.

[0028] Step 2: Beam excitation current weighting optimization based on KD-DDPG: Based on the hovering positions of the UAV at different times obtained from the model trained in the first stage, combined with the movement trajectory of the agricultural machinery, the angle of arrival of the UAV relative to the agricultural machinery at different time points can be calculated. Therefore, the actual target direction of the communication beam can be obtained. In order to further improve the sensing performance, I optimized the excitation current weight in the beam weight in this part, so that the sensing beam and the communication beam form a zero trap with mutually secure gains.

[0029] In the method described, KD-DDPG uses a fully connected network as the network model for both the actor and critic networks. The model inputs the UAV's array specifications, sensing beam, and communication beam's target direction at different time points. The excitation current weights of the sensing and communication beams are optimized. To ensure model lightweighting during training, a self-distillation module is added to the critic network to assist in training. The process is as follows: Figure 6 As shown.

[0030] In this section, the state is represented as: (17) in The depression and elevation angles represent the target's angle of arrival corresponding to the sensing or communication beam. This indicates the depression and elevation angles of the target arrival angle for the communication / sensing beam in the same time slot. The selected action is then represented in the algorithm as a 1×N vector consisting of floating-point numbers from 0 to 1, where N = N x ×N y (18) Construct the reward function using Formula 10: (19) The scheme sequentially obtains the optimal excitation current weight allocation of the sensing beam and the communication beam under different time slots. At this point, the hovering position of the UAV and the corresponding sensing beam width, the UAV access sequence, i.e. the UAV trajectory and the excitation current weight of the beam are optimized, and the optimization problem corresponding to formula (10) is solved.

[0031] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for integrated communication of sensing and coordination of collaborative work of smart agricultural equipment, characterized in that, Includes the following steps: Step 1: Train alternately using the re-GDQN method and the IDRL method to jointly obtain the optimal UAV hovering point, sensing beamwidth, and UAV trajectory; Step 1.1: Propose the IDRL model. IDRL combines diffusion modeling with reinforcement learning, and uses the Q-value to guide the update of the diffusion model to generate the optimal solution; Step 1.2: A re-GDQN model is proposed to obtain the optimal hovering position of the UAV and the corresponding beamwidth. A Q-network with a GNN graph neural network as reinforcement learning is proposed. The information of the center point position and whether it is covered by the grid is input into the re-GDQN model, and IDRL is called to obtain the current optimal trajectory according to the optimal access order of the generated center points. The reward is obtained according to the trajectory length and coverage efficiency to guide the Q-network update and obtain the optimal hovering position of the UAV. Step 2: Optimize beam excitation current weights based on KD-DDPG. A fully connected network is used as the network model for both the actor and critic networks. The array specifications of the UAV, the sensing beam and the communication beam, and the target directions at different time points are input into the model to optimize the excitation current weights of the sensing beam and the communication beam. A self-distillation module is added to the critic network to assist in training. 2.The method of claim 1, wherein, The alternating training method in step one is as follows: First, train IDRL so that it can obtain the optimal UAV trajectory for a set of hovering positions and corresponding communication, sensing beamwidths and target directions, so that it can avoid overlap between sensing beams and communication beams while minimizing the distance; after p rounds, fix the IDRL model and train the re-GDQN method so that it can obtain reasonable UAV hovering points and corresponding sensing beamwidths for irregular farmland, so that the UAV's movement energy consumption is minimized while the sensing beam can scan the entire target area. 3.The method of claim 1, wherein, In step 1.1, the proposed IDRL uses the reverse generation process of the diffusion model as the policy network for reinforcement learning. In each training process, the diffusion model network composed of U-Net generates the probability of selecting each path, and then uses the H operator to convert the probability matrix into a Hamiltonian loop. Subsequently, the corresponding Q value is calculated and passed to guide the diffusion model update, thereby obtaining the optimal UAV access order so that the UAV movement energy consumption is minimized and the target direction of the variable sensing beam and the communication beam overlaps.

4. The sensor-integrated communication method for collaborative operation of smart agricultural equipment according to claim 1, characterized in that, In step 1.1, the environment in which the agent is located contains the UAV designated hovering position X, the perception beam coverage range φt corresponding to the UAV hovering position, and the target position of the communication beam alignment in each time slot And the corresponding communication beam coverage range , and the selected Hamiltonian loop A represented by the adjacency matrix tsp_t , the moving trajectory of the agricultural machine, the time allocation T of the overall environment, and the time step parameter t required by the diffusion model, so the environment state space is represented as: , wherein A t is the selected action in the corresponding state s t in different models; The agent state space is set as a continuous space of edge probability distribution consisting of a floating-point array from 0 to 1, and the reward function is set as: , wherein R collection is an overlap penalty. 5.The method of claim 1, wherein, In step 1.2, the description of the intelligent agent's environment first involves introducing a quadtree to divide the target farmland area into uniform grids. The side length of each leaf node in the quadtree grid is equal to the side length of the inscribed square of the beam coverage area corresponding to the smallest beam that the UAV can construct. Based on this, the environment is constructed as a graph structure, specifically containing two types of nodes: the center point of the grid divided by the quadtree and the center point corresponding to the area covered by the sensing beam. The edges include three types: grid point-grid point, coverage area-grid point, and coverage area-coverage area. The graph structure shows the topological relationships.

6. The sensor-integrated communication method for collaborative operation of smart agricultural equipment according to claim 1, characterized in that, The state in step 1.2 is represented as , where G is the grid topology, V is the node feature matrix containing the coordinate position of the node and the feature of whether it is covered, E is the edge relationship represented by the adjacency matrix, A is the action mask matrix identifying the legal action space, and the action space is discrete, which is the set of arbitrary coverage widths that can construct a grid center connected to the covered perception area under the current state, and each action is represented as , node i is the selected center coordinate of the grid belonging to the optional action space, is the selected beam width corresponding to the position, and the re-GDQN returns a reward to the agent after completely covering the target area, that is, a sparse reward is set, , where μ is the coverage efficiency of the current selected scheme, that is, the ratio of the area covered to the total area covered by the current set of perception beams, and D is the total energy consumption of the UAV calculated by the optimal access order under the current scheme obtained by IDRL.

7. The sensor-integrated communication method for collaborative operation of smart agricultural equipment according to claim 1, characterized in that, In step two, the state is represented as , where represents the target azimuth and elevation angles of the perception or communication beam, represents the target azimuth and elevation angles of the corresponding communication / perception beam in the same time slot, is the selected action, and the action space is represented as a 1xN vector composed of 0~1 floating-point numbers, where N=N x xN y , , and the reward function is constructed , and the optimal excitation current weight distribution of the perception beam and the communication beam in different time slots is obtained in turn. At this point, the UAV hovering position and the corresponding perception beam width, the UAV access order, i.e., the UAV trajectory and the beam excitation current weight optimization are completed.

Images