Material delivery method and system for heavy-load unmanned aerial vehicle

Abstract

The invention relates to the technical field of material delivery, in particular to a material delivery method and system for a heavy-load unmanned aerial vehicle, and the method comprises the steps: obtaining a rescue request in a disaster area, determining the position of a rescue point according to the rescue request, and determining a material transportation path of a heavy-load unmanned aerial vehicle cluster according to the position of the rescue point; and controlling the heavy-load unmanned aerial vehicle cluster to transport materials to the sky above the rescue point position according to the material transportation path, and performing image acquisition on the rescue point position according to the heavy-load unmanned aerial vehicles to obtain an environment image of the rescue point. According to the invention, by acquiring the rescue request in the affected area and determining the position of the rescue point, the unmanned aerial vehicle cluster can be ensured to accurately transport the materials to the required area, so that the efficiency of material distribution is improved, and the heavy-load unmanned aerial vehicle cluster is controlled to execute the task according to the preset material transport path, thereby improving the efficiency of material distribution. And no-load flight and unnecessary deviation of the unmanned aerial vehicle can be reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of material delivery, and in particular to a material delivery method and system for a heavy-loaded unmanned aerial vehicle. Background Art

[0002] A heavy-lift drone is a type of drone that can carry and transport heavier items. Unlike conventional consumer-grade drones, heavy-lift drones have a stronger power system, a larger body, and a higher carrying capacity. They are suitable for situations where a large amount of weight needs to be carried when transporting and performing tasks.

[0003] Traditional methods of material delivery, such as relying on transportation tools such as trucks and helicopters, usually require a long transportation time and multiple transfers, and are inefficient. In post-disaster rescue, traditional methods may slow down the response time of material delivery due to problems such as road congestion and traffic restrictions; and traditional methods lack intelligent scheduling and path optimization, which may lead to empty flights and deviations from the path during the material distribution process, increasing the time cost of transportation and reducing efficiency; and traditional methods usually lack real-time environmental recognition capabilities and cannot make quick judgments based on the actual situation on site. Even in complex terrain or environments that are not suitable for landing, traditional methods often cannot identify them in advance, resulting in failure of material delivery or drone malfunctions; and traditional material delivery methods rely on manual operation or fixed-point landing, and may not be able to accurately deliver materials due to misjudgment or an environment that is not suitable for landing, resulting in the failure to deliver materials to where they are needed on time. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a material delivery method and system for heavy-loaded unmanned aerial vehicles.

[0005] The technical solution adopted to solve the above technical problems is: a material delivery method for a heavy-loaded UAV, comprising: Obtaining rescue requests in the disaster-stricken area, determining the location of rescue points based on the rescue requests, and determining the material transportation path of the heavy-loaded drone cluster based on the location of the rescue points; Control the heavy-loaded drone cluster to transport materials to the air above the rescue point according to the material transportation path, and collect images of the rescue point according to the heavy-loaded drone to obtain an environmental image of the rescue point; Performing environmental recognition on the environmental image according to a pre-trained environmental recognition model to obtain the environmental state of the rescue point, wherein the environmental state includes a landable state and a non-landable state; When the environmental state is a landable state, the heavy-load UAV is controlled to land at the rescue point according to the pre-trained landing control model; When the environmental state is in a non-landing state, directly control the heavy-load UAV to land at a preset height and drop the supplies at the rescue point.

[0006] Preferably, determining the location of the rescue point according to the rescue request, and determining the material transportation path of the heavy-load UAV cluster according to the location of the rescue point, including: Obtain a preset list of material transportation paths, where the material transportation path includes at least one rescue point location and a heavy-load UAV base; Randomly select a material transportation path from the list of material transportation paths and add it to the preset current path list, and add the remaining material transportation paths in the list of material transportation paths to the preset remaining path list; Calculate the distance similarity between the material transportation path in the current path list and each material transportation path in the remaining path list; Select the material transportation path with the smallest distance similarity, add the material transportation path with the smallest distance similarity to the current path list, and delete the material transportation path with the smallest distance similarity from the remaining path list; Repeat the above operations until the number of rescue point locations in the current path list reaches a preset quantity threshold to obtain the path grouping of the heavy-load UAV cluster, and add the path grouping to the preset grouping list; Judge whether the remaining path list is empty. If it is empty, output each path grouping in the grouping list, where one heavy-load UAV in the heavy-load UAV cluster is responsible for one path grouping in the grouping list. Otherwise, repeat the above operations.

[0007] Preferably, the calculation formula of the distance similarity is as follows: ; Wherein, represents the distance similarity, and represent paths and path , and represent paths and path The number of rescue points and heavy-load UAV bases included in, and represent the heavy-load UAV bases of paths and path , represents a random number uniformly distributed in the range of 0-1, represents the rescue point and the rescue point The distance between...

[0008] Preferably, the environment recognition model adopts an improved YOLO model, and the backbone network of the YOLO model is improved according to the feature extraction module and the feature fusion module. Among them, the feature extraction module includes four branches, each branch corresponding to a different receptive field size. Among them, the first branch adopts a residual structure. Before executing the remaining three branches, the number of channels of the input feature map is reduced from 2C to C by using a 1×1 convolution. The second branch extracts small receptive field features through a 3×3 convolution to obtain a first output feature map. The third branch uses a DRB module with K = 7 to extract medium receptive field features to obtain a second output feature map. The fourth branch uses a DRB module with K = 15 to extract large receptive field features to obtain a third output feature map. The first output feature map is added element-wise to the output feature map of the first branch to obtain a fourth output feature map. The second output feature map is added element-wise to the output feature map of the first branch to obtain a fifth output feature map. The third output feature map is added element-wise to the output feature map of the first branch to obtain a sixth output feature map.

[0009] Preferably, the feature fusion module uses a parallel structure of max pooling and average pooling to downsample the fourth output feature map to obtain a seventh output feature map. The sixth output feature map is upsampled using the nearest neighbor interpolation method to obtain an eighth output feature map. The seventh output feature map, the fifth output feature map, and the eighth output feature map are divided into three equal sub-feature maps in the channel dimension. The Sigmoid function is applied to the three equal channel sub-feature maps of the seventh output feature map, the fifth output feature map, and the eighth output feature map respectively through a feature adaptive selection fusion mechanism to calculate the selection weights. The three equal channel sub-feature maps of the seventh output feature map, the fifth output feature map, and the eighth output feature map are fused according to the selection weights to obtain a first fusion feature, a second fusion feature, and a third fusion feature. The first fusion feature, the second fusion feature, and the third fusion feature are concatenated in the channel dimension to obtain an output fusion feature.

[0010] Preferably, the landing control model includes a Markov decision model and a deep reinforcement learning model. Among them, the Markov decision model includes a global state space set, an action space set of the heavy-load UAV cluster, a state transition probability, a discount factor, and a reward value. Among them, the global state space set includes a first state, a second state, and a third state. The action space set includes a height control action and a steering angle control action. The state transition probability represents the state transition probability that the global state jumps to the next state after the heavy-load UAV cluster takes an action. The discount factor represents the influence of subsequent states on the current action decision. The reward value includes a first reward, a second reward, and a third reward. The deep reinforcement learning model is used to solve the Markov decision model to obtain an optimal landing strategy.

[0011] Preferably, the first state represents the relative position information between the heavy-load UAV and the rescue point. The first state includes the distance and relative direction angle between the heavy-load UAV and the rescue point. The second state represents the relative information of the neighboring heavy-load UAVs. The second state includes the distance and relative direction angle between the two heavy-load UAVs with the closest distance within the communication range. The third state represents the obstacle distance obtained by lidar. The expression of the steering angle control action is as follows: ; Wherein, represents the steering angle of the th heavy-load UAV, represents the th heavy-load UAV, represents the absolute value of the maximum steering angle of the heavy-load UAV, represents the number of equal divisions of the steering range of the heavy-load UAV.

[0012] Preferably, the expression of the first reward is as follows: ; Wherein, represents the first reward, and represent preset weights, and , represents the set of rescue points near the heavy-load UAV, represents at the distance between the heavy-load UAV and the target point at the moment; The expression of the second reward is as follows: ; Wherein, represents the second reward, Represents the minimum distance from the overloaded drone to the center of the influence range of surrounding obstacles, Represents the influence radius of the obstacle, Represents the obstacle avoidance buffer distance, Represents the safety distance of the overloaded drone, Represents the allowable radius of the obstacle, Represents the allowable straight-line distance between the overloaded drone and the obstacle, Represents the first preset reward value, Represents the second preset reward value; The expression of the third reward is as follows: ; Wherein, Represents the third reward, Represents the minimum distance between the overloaded drone and other overloaded drones, Represents the minimum distance that should be maintained between overloaded drones.

[0013] Preferably, the deep reinforcement learning model is used to solve the Markov decision model to obtain an optimal landing strategy, including: Initializing the departure position and target position of the drone; Selecting an action uniformly at random from the global state space set with a preset probability; Executing the action while obtaining the next moment state and reward; Putting the current trajectory into the experience replay buffer, randomly and uniformly collecting multiple trajectories, and updating the main network parameters according to the loss function for each trajectory, where the trajectory includes the current state, the current action, the next moment state, and the reward; Updating the target network parameters using the main network parameters every preset number of training steps.

[0014] The technical solution adopted to solve the above technical problems is: A material delivery system for an overloaded drone, which is applicable to the above-mentioned material delivery method for an overloaded drone, including: A path planning unit, which is used to obtain a rescue request in the disaster area, determine the rescue point location according to the rescue request, and determine the material transportation path of the overloaded drone cluster according to the rescue point location; An environment acquisition unit, which is used to control the overloaded drone cluster to transport materials to the air above the rescue point location according to the material transportation path, and perform image acquisition on the rescue point location according to the overloaded drone to obtain an environmental image of the rescue point; An environment recognition unit, which is used to perform environment recognition on the environment image according to a pre-trained environment recognition model to obtain the environmental state of the rescue point, where the environmental state includes a landable state and a non-landable state; A landing control unit, which is used to control the heavy-duty UAV to land at the rescue point according to a pre-trained landing control model when the environmental state is a landable state; A material delivery unit, which is used to directly control the heavy-duty UAV to land at a preset height and deliver the materials to the rescue point when the environmental state is a non-landable state.

[0015] The beneficial effects of the present invention are as follows: (1) By obtaining the rescue request in the disaster area and determining the location of the rescue point, the present invention can ensure that the UAV cluster accurately transports materials to the areas in need, thereby improving the efficiency of material distribution. And controlling the heavy-duty UAV cluster to execute tasks according to the preset material transportation path can reduce the empty flight of the UAVs and unnecessary deviations, optimize the path planning, thereby saving time and improving the accuracy of material distribution; (2) By collecting images of the environment of the rescue point and applying a pre-trained environment recognition model, the system can identify and judge the environmental state of the rescue point. This intelligent environment recognition and judgment can avoid failures or accidents caused by unsafe landings of the UAVs, improving the safety and reliability of material delivery; (3) When the environmental state is "landable state", the heavy-duty UAV can land safely and complete material delivery; while in the "non-landable state", the UAV will choose to land directly at the preset height and deliver materials, which avoids the risk of the UAV being unable to complete the task due to unsuitable landing environment, and ensures that the materials can be successfully delivered under different environmental conditions. Whether it can land or not, it can ensure that the materials reach the required rescue point in time. Description of the Drawings

[0016] Figure 1 It is a schematic flow chart of the steps of the overall method in an embodiment proposed by the present invention; Figure 2 It is a schematic system architecture diagram of the overall system in an embodiment proposed by the present invention.

[0017] Reference numerals: 1, path planning unit; 2, environment collection unit; 3, environment recognition unit; 4, landing control unit; 5, material delivery unit. Detailed Embodiments

[0018] Embodiment 1, as Figure 1 shown, a method for delivering materials for a heavy-duty UAV proposed by the present invention includes: S1. Obtain the rescue requests within the disaster area, determine the locations of the rescue points according to the rescue requests, and determine the material transportation routes of the heavy-load UAV cluster based on the locations of the rescue points; S2. Control the heavy-load UAV cluster to transport materials to the airspace above the locations of the rescue points according to the material transportation routes, and perform image acquisition on the locations of the rescue points by the heavy-load UAVs to obtain the environmental images of the rescue points; S3. Perform environmental recognition on the environmental images according to the pre-trained environmental recognition model to obtain the environmental states of the rescue points, where the environmental states include a landable state and a non-landable state; S4. When the environmental state is the landable state, control the heavy-load UAVs to land at the rescue points according to the pre-trained landing control model; S5. When the environmental state is the non-landable state, directly control the heavy-load UAVs to land at a preset height and drop the materials at the rescue points.

[0019] In the present invention, first, it is necessary to collect the rescue requests within the disaster area. These requests usually come from the rescue personnel or relevant departments in the disaster area, indicating the need to urgently transport materials to specific locations. After receiving the rescue requests, the system will determine the locations of the rescue points where the materials need to be dropped according to the content of the requests. These locations are usually calibrated by positioning technologies such as GPS coordinates to accurately determine the specific locations where the materials need to be dropped. The UAV cluster starts to execute the transportation task according to the material transportation routes, carrying the rescue materials and flying to the designated rescue points. During the transportation process, the UAV cluster will fly according to the predetermined routes to ensure timely and accurate arrival at the target locations. When the UAV cluster reaches the airspace above the rescue points, the UAVs will perform image acquisition, obtaining the environmental images around the rescue points through the on-board cameras or sensors. If the environmental recognition model determines that the conditions in the rescue point area are suitable for landing (for example, there are no obstacles, the ground is flat, and there are no power lines, etc.), it is regarded as the landable state. If it is determined that the environmental conditions are not suitable for landing (for example, the ground is uneven, there are obstacles, or there are other safety hazards), it is regarded as the non-landable state. If it is determined that the environment is suitable for landing, the pre-trained landing control model will be used to control the landing process of the heavy-load UAVs. The landing control model will adjust the attitude, speed, etc. of the UAVs according to the real-time environmental information to ensure that the UAVs can land safely at the rescue points. If the environmental recognition model determines that the environment is not suitable for landing, the system will control the UAVs to land at a preset safe height. Then, the heavy-load UAVs will perform material dropping at the preset height, and the materials will be accurately dropped into the rescue point area by means of aerial delivery. In this way, even if landing is not possible, the UAVs can still deliver the required materials.

[0020] Embodiment 2. A method for delivering supplies to a heavy-duty unmanned aerial vehicle proposed by the present invention. Compared with Embodiment 1, this embodiment further includes: determining the location of the rescue point according to the rescue request, and determining the supply transportation path of the heavy-duty unmanned aerial vehicle cluster according to the location of the rescue point, including: A1. Obtain a preset list of supply transportation paths, where the supply transportation path includes at least one rescue point location and a heavy-duty unmanned aerial vehicle base; A2. Randomly select a supply transportation path from the list of supply transportation paths and add it to the preset current path list, and add the remaining supply transportation paths in the list of supply transportation paths to the preset remaining path list; A3. Calculate the distance similarity between the supply transportation path in the current path list and each supply transportation path in the remaining path list; A4. Select the supply transportation path with the smallest distance similarity, add the supply transportation path with the smallest distance similarity to the current path list, and delete the supply transportation path with the smallest distance similarity from the remaining path list; A5. Repeat the above operations until the number of rescue point locations in the current path list reaches a preset quantity threshold to obtain path groups of the heavy-duty unmanned aerial vehicle cluster, and add the path groups to the preset group list; A6. Determine whether the remaining path list is empty. If it is empty, output each path group in the group list, where one heavy-duty unmanned aerial vehicle in the heavy-duty unmanned aerial vehicle cluster is responsible for one path group in the group list. Otherwise, repeat the above operations.

[0021] In this embodiment, the rescue point location is the target area, the place where supplies need to be delivered; the heavy-lift UAV base refers to the starting point of material transportation, usually the take-off location of the heavy-lift UAV; from the list of material transportation paths, randomly select a path and add it to the current path list, and the remaining paths (the unselected paths) will be added to the remaining path list; calculate the distance similarity between each path in the current path list and each path in the remaining path list, and this similarity may be based on factors such as geographical distance, flight time, path complexity, etc.; select the path with the minimum similarity to the paths in the current path list from the remaining path list, which means selecting the path that is most different from the paths in the current path list. Usually, this is to avoid path repetition or optimize resource scheduling; repeat the above steps: each time, select the path with the minimum similarity to the paths in the current path list from the remaining path list, add it to the current path list and delete it. This process will continue until the preset number of rescue points is included in the current path list; when the number of paths in the current path list reaches the preset quantity threshold, regard the current path list as a path group and add it to the group list; determine whether the remaining path list is empty. If it is empty, it means that all paths have been allocated and the path grouping is completed; if the remaining path list is empty, output the group list, where each path group can be assigned to a UAV in a heavy-lift UAV cluster.

[0022] In an alternative embodiment, the calculation formula for the distance similarity is as follows: ; Where, represents the distance similarity, and represent paths and path , and represent paths and path include the number of rescue points and heavy-lift UAV bases, and represent the heavy-lift UAV bases of paths and path , represents a random number uniformly distributed in the range of 0 - 1, represents the rescue point and the rescue point between the distances.

[0023] In an alternative embodiment, the environment recognition model adopts an improved YOLO model, and the backbone network of the YOLO model is improved according to the feature extraction module and the feature fusion module. Among them, the feature extraction module includes four branches, each branch corresponding to a different receptive field size. Among them, the first branch adopts a residual structure. Before executing the remaining three branches, the number of channels of the input feature map is reduced from 2C to C by using a 1×1 convolution. The second branch extracts small receptive field features through a 3×3 convolution to obtain a first output feature map. The third branch uses a DRB module with K = 7 to extract medium receptive field features to obtain a second output feature map. The fourth branch uses a DRB module with K = 15 to extract large receptive field features to obtain a third output feature map. The first output feature map is added element-wise to the output feature map of the first branch to obtain a fourth output feature map. The second output feature map is added element-wise to the output feature map of the first branch to obtain a fifth output feature map. The third output feature map is added element-wise to the output feature map of the first branch to obtain a sixth output feature map.

[0024] It should be noted that the goal of the feature extraction module is to extract useful features from the input image to assist in environment recognition. To improve the effect of feature extraction, this module uses four branches, each branch targeting a different receptive field size (i.e., the size of the image information area captured by each branch); DRB with K = 7 represents a dynamic convolution module with a convolution kernel size of 7×7; the DRB module with K = 15 has a convolution kernel size of 15×15.

[0025] In an alternative embodiment, the feature fusion module uses a parallel structure of max pooling and average pooling to downsample the fourth output feature map to obtain a seventh output feature map, and uses the nearest neighbor interpolation method to upsample the sixth output feature map to obtain an eighth output feature map. The seventh output feature map, the fifth output feature map, and the eighth output feature map are divided into three equal sub-feature maps in the channel dimension. The sigmoid function is applied to the three equal channel sub-feature maps of the seventh output feature map, the fifth output feature map, and the eighth output feature map respectively through a feature adaptive selection fusion mechanism to calculate the selection weights. The three equal channel sub-feature maps of the seventh output feature map, the fifth output feature map, and the eighth output feature map are fused according to the selection weights to obtain a first fusion feature, a second fusion feature, and a third fusion feature. The first fusion feature, the second fusion feature, and the third fusion feature are concatenated in the channel dimension to obtain an output fusion feature.

[0026] It should be noted that max pooling is used to extract the local strongest response in the feature map, while average pooling retains the overall information of the feature map. The combination of the two can effectively retain different types of feature information. The Sigmoid function is applied to each sub-feature map to calculate the selection weights. The output of the Sigmoid function will be a value between 0 and 1, which is used to represent the importance or weight of each channel sub-feature map.

[0027] In an alternative embodiment, the landing control model includes a Markov decision model and a deep reinforcement learning model. Among them, the Markov decision model includes a global state space set, an action space set of the heavy-load UAV cluster, a state transition probability, a discount factor, and a reward value. Among them, the global state space set includes a first state, a second state, and a third state. The action space set includes a height control action and a steering angle control action. The state transition probability represents the state transition probability that the global state jumps to the next state when the heavy-load UAV cluster takes an action. The discount factor represents the influence of subsequent states on the current action decision. The reward value includes a first reward, a second reward, and a third reward. The deep reinforcement learning model is used to solve the Markov decision model to obtain an optimal landing strategy.

[0028] It should be noted that the Markov decision process is a mathematical framework for modeling decision-making problems, especially in the presence of uncertainty. Deep reinforcement learning is a technology that combines deep learning and reinforcement learning, which can automatically learn optimal strategies in complex environments. In this landing control model, deep reinforcement learning is used to solve the Markov decision process (MDP) and optimize the landing strategy of the UAV by interacting with the environment.

[0029] In an alternative embodiment, the first state represents the relative position information between the heavy-load UAV and the rescue point. The first state includes the distance and relative direction angle between the heavy-load UAV and the rescue point. The second state represents the relative information of the neighboring heavy-load UAVs. The second state includes the distance and relative direction angle of the two heavy-load UAVs with the shortest distance within the communication range. The third state represents the obstacle distance obtained by lidar. The expression of the steering angle control action is as follows: ; where represents the steering angle of the th heavy-load UAV, represents the th heavy-load UAV, represents the absolute value of the maximum steering angle of the heavy-load UAV,

[0030] In an alternative embodiment, the expression of the first reward is as follows: ; Among them, represents the first reward, and represent preset weights, and , represents the set of rescue points near the heavy-lift drone, represents at the distance between the heavy-lift drone and the target point at the moment, represents the first preset reward value, represents the second preset reward value; The expression of the second reward is as follows: ; Among them, represents the second reward, represents the minimum value of the distance between the heavy-lift drone and the center of the influence range of surrounding obstacles, represents the influence radius of the obstacle, represents the obstacle avoidance buffer distance, represents the safety distance of the heavy-lift drone, represents the allowable radius of the obstacle, represents the allowable straight-line distance between the heavy-lift drone and the obstacle; The expression of the third reward is as follows: ; Among them, represents the third reward, represents the minimum distance between the heavy-lift drone and other heavy-lift drones, represents the minimum distance that should be maintained between heavy-lift drones.

[0031] In an alternative embodiment, a deep reinforcement learning model is used to solve the Markov decision model to obtain an optimal landing strategy, including: B1. Initialize the departure position and target position of the drone; B2. Select an action uniformly at random from the global state space set with a preset probability; B3. Execute the action, and at the same time obtain the next moment state and reward; B4. Put the current trajectory into the experience replay buffer, randomly and uniformly collect multiple trajectories, and update the main network parameters according to the loss function for each trajectory, where the trajectory includes the current state, the current action, the next moment state, and the reward; B5. Update the target network parameters using the main network parameters every preset number of training steps.

[0032] It should be noted that in reinforcement learning, a main network and a target network are usually used for training. The parameters of the target network are a copy of the main network and are updated regularly to maintain the stability of training. After every preset number of training steps, the parameters of the target network are updated according to the parameters of the main network. This is done to avoid unstable training caused by frequent updates.

[0033] Embodiment 3, as Figure 2 shown, a material delivery system for a heavy-duty unmanned aerial vehicle proposed by the present invention, and a method for delivering materials for a heavy-duty unmanned aerial vehicle applicable thereto, include: A path planning unit 1, which is used to obtain rescue requests within the disaster area, determine the location of the rescue point according to the rescue requests, and determine the material transportation path of the heavy-duty unmanned aerial vehicle cluster according to the location of the rescue point; An environment acquisition unit 2, which is used to control the heavy-duty unmanned aerial vehicle cluster to transport materials to the air above the rescue point location according to the material transportation path, and collect images of the rescue point location by the heavy-duty unmanned aerial vehicle to obtain the environmental image of the rescue point; An environment recognition unit 3, which is used to perform environment recognition on the environmental image according to a pre-trained environment recognition model to obtain the environmental state of the rescue point, where the environmental state includes a landable state and a non-landable state; A landing control unit 4, which is used to control the heavy-duty unmanned aerial vehicle to land at the rescue point according to a pre-trained landing control model when the environmental state is a landable state; A material delivery unit 5, which is used to directly control the heavy-duty unmanned aerial vehicle to land at a preset height and deliver the materials to the rescue point when the environmental state is a non-landable state.

[0034] The above has described the embodiments of the present invention in detail with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made without departing from the spirit of the present invention within the scope of knowledge possessed by those skilled in the art to which the present invention pertains.

Claims

1. A material delivery method for a heavy-loaded UAV, characterized in that: include: Obtaining rescue requests in the disaster-stricken area, determining the location of rescue points based on the rescue requests, and determining the material transportation path of the heavy-loaded drone cluster based on the location of the rescue points; Control the heavy-loaded drone cluster to transport materials to the air above the rescue point according to the material transportation path, and collect images of the rescue point according to the heavy-loaded drone to obtain an environmental image of the rescue point; Performing environmental recognition on the environmental image according to a pre-trained environmental recognition model to obtain the environmental state of the rescue point, wherein the environmental state includes a landable state and a non-landable state; When the environmental state is a landable state, the heavy-load UAV is controlled to land at the rescue point according to the pre-trained landing control model; When the environmental state is a non-landing state, the heavy-loaded UAV is directly controlled to land at a preset height, and the supplies are delivered to the rescue point.

2. A material delivery method for a heavy-loaded UAV according to claim 1, characterized in that: Determining the location of the rescue point according to the rescue request, and determining the material transportation path of the heavy-loaded drone cluster according to the location of the rescue point, including: Obtaining a preset material transportation route list, wherein the material transportation route includes at least one rescue point location and a heavy-load drone base; Randomly selecting a material transportation path from the material transportation path list and adding it to a preset current path list, and adding the remaining material transportation paths in the material transportation path list to a preset remaining path list; Calculate the distance similarity between the material transportation path in the current path list and each material transportation path in the remaining path list; Selecting the material transportation path with the smallest distance similarity, adding the material transportation path with the smallest distance similarity to the current path list, and deleting the material transportation path with the smallest distance similarity from the remaining path list; Repeat the above operation until the rescue point positions in the current path list reach a preset quantity threshold, so as to obtain the path grouping of the heavy-loaded drone cluster, and add the path grouping to the preset grouping list; Determine whether the remaining path list is empty. If so, output each path grouping in the grouping list, wherein one heavy-loaded drone in the heavy-loaded drone cluster is responsible for one path grouping in the grouping list. Otherwise, repeat the above operation.

3. A material delivery method for a heavy-loaded UAV according to claim 2, characterized in that: The calculation formula of the distance similarity is as follows: ； in, Represents the distance similarity, and Indicates the path and path , and Indicates the path and path The number of rescue points and heavy-load drone bases included in and Indicates the path and path The heavy-load UAV base, Indicates a random number in the range of 0-1 that follows a uniform distribution. Indicates rescue point and rescue points The distance between.

4. A material delivery method for a heavy-loaded UAV according to claim 3, characterized in that: The environment recognition model adopts an improved YOLO model, and improves the backbone network of the YOLO model according to a feature extraction module and a feature fusion module, wherein the feature extraction module includes four branches, each branch corresponds to a different receptive field size, wherein the first branch adopts a residual structure, and before executing the remaining three branches, the number of channels of the input feature map is reduced from 2C to C by using a 1×1 convolution, the second branch extracts small receptive field features by a 3×3 convolution to obtain a first output feature map, the third branch uses a DRB module with K=7 to extract medium receptive field features to obtain a second output feature map, the fourth branch uses a DRB module with K=15 to extract large receptive field features to obtain a third output feature map, the first output feature map is element-wise added to the output feature map of the first branch to obtain a fourth output feature map, the second output feature map is element-wise added to the output feature map of the first branch to obtain a fifth output feature map, and the third output feature map is element-wise added to the output feature map of the first branch to obtain a sixth output feature map.

5. A material delivery method for a heavy-loaded UAV according to claim 4, characterized in that: The feature fusion module downsamples the fourth output feature map using a parallel structure of maximum pooling and average pooling to obtain a seventh output feature map, upsamples the sixth output feature map using a nearest neighbor interpolation method to obtain an eighth output feature map, divides the seventh output feature map, the fifth output feature map, and the eighth output feature map into three equal sub-feature maps in the channel dimension, and applies a Sigmoid function to the three equal channel sub-feature maps of the seventh output feature map, the fifth output feature map, and the eighth output feature map through a feature adaptive selection fusion mechanism to calculate selection weights, fuses the three equal channel sub-feature maps of the seventh output feature map, the fifth output feature map, and the eighth output feature map according to the selection weights to obtain a first fusion feature, a second fusion feature, and a third fusion feature, and splices the first fusion feature, the second fusion feature, and the third fusion feature in the channel dimension to obtain an output fusion feature.

6. A material delivery method for a heavy-loaded UAV according to claim 5, characterized in that: The landing control model includes a Markov decision model and a deep reinforcement learning model, wherein the Markov decision model includes a global state space set, an action space set of a heavy-loaded UAV cluster, a state transition probability, a discount factor and a reward value, wherein the global state space set includes, the global state space set includes a first state, a second state and a third state, the action space set includes an altitude control action and a steering angle control action, the state transition probability represents the state transition probability of the global state jumping to the next state after the heavy-loaded UAV cluster takes an action, the discount factor represents the influence of subsequent states on the current action decision, the reward value includes a first reward, a second reward and a third reward, and the deep reinforcement learning model is used to solve the Markov decision model to obtain the optimal landing strategy.

7. A material delivery method for a heavy-loaded UAV according to claim 6, characterized in that: in, The first state represents the relative position information between the heavy-loaded UAV and the rescue point, including the distance and relative direction angle between the heavy-loaded UAV and the rescue point. The second state represents the relative information of the adjacent heavy-loaded UAVs, including the distance and relative direction angle between the two closest heavy-loaded UAVs within the communication range. The third state represents the obstacle distance obtained by the laser radar. The expression of the steering angle control action is as follows: ； in, Indicates The steering angle of a heavy-loaded UAV, Indicates A heavy-loaded drone. Indicates the absolute value of the maximum steering angle of a heavy-loaded UAV. Indicates the number of equal parts of the steering range of the heavy-load drone.

8. A material delivery method for a heavy-loaded UAV according to claim 7, characterized in that: The expression of the first reward is as follows: ； in, Indicates the first reward, and represents the preset weight, and , Indicates the set of rescue points near the heavy-loaded drone. Indicated in Always reload the distance between the drone and the target point; The expression of the second reward is as follows: ； in, Indicates the second reward, Indicates the minimum value of the distance between the heavy-loaded UAV and the center of the surrounding obstacle influence range. Indicates the influence radius of the obstacle, Indicates the obstacle avoidance buffer distance, Indicates the safe distance for heavy-loaded drones. Indicates the allowed radius of obstacles, Indicates the straight-line allowable distance between the heavy-loaded drone and the obstacle. represents the first preset reward value, represents a second preset reward value; The expression of the third reward is as follows: ； in, Indicates the third reward, Indicates the minimum distance between a heavy-loaded drone and other heavy-loaded drones. Indicates the minimum distance that heavy-loaded drones should maintain.

9. A material delivery method for a heavy-loaded UAV according to claim 8, characterized in that: The deep reinforcement learning model is used to solve the Markov decision model to obtain the optimal landing strategy, including: Initialize the starting position and target position of the drone; Under a preset probability, uniformly and randomly select an action from the global state space set; Execute the action and obtain the next state and reward at the same time; Put the current trajectory into the experience playback buffer, randomly and evenly collect multiple trajectories, and update the main network parameters for each trajectory according to the loss function, where the trajectory includes the current state, current action, next moment state and reward; The target network parameters are updated using the main network parameters every preset number of training steps.

10. A material delivery system for a heavy-loaded unmanned aerial vehicle, which is applicable to the material delivery method for a heavy-loaded unmanned aerial vehicle according to claim 9, characterized in that: include: A path planning unit (1), the path planning unit (1) being used to obtain a rescue request in a disaster-stricken area, determine a rescue point location according to the rescue request, and determine a material transportation path for a heavy-loaded drone cluster according to the rescue point location; An environment collection unit (2), the environment collection unit (2) being used to control the heavy-load drone cluster to transport materials to the air above the rescue point according to the material transportation path, and to collect images of the rescue point according to the heavy-load drone to obtain an environment image of the rescue point; An environment recognition unit (3), the environment recognition unit (3) being used to perform environment recognition on the environment image according to a pre-trained environment recognition model to obtain the environment state of the rescue point, wherein the environment state includes a landable state and a non-landable state; A landing control unit (4), the landing control unit (4) being used to control the heavy-loaded UAV to land at the rescue point according to a pre-trained landing control model when the environmental state is a landable state; A material delivery unit (5), the material delivery unit (5) is used to directly control the heavy-loaded drone to land to a preset height and deliver the material to the rescue point when the environmental state is a non-landing state.

Images