Fishing-light complementary unmanned aerial vehicle cooperative intelligent feeding system

Abstract

The invention relates to the technical field of aquaculture intelligent feeding, and discloses a fishing and light complementation unmanned aerial vehicle cooperative intelligent feeding system, which is characterized in that a pond is dynamically divided into feeding sub-regions, a feeding model is arranged, and the feeding model is used for feeding the fish and light complementation unmanned aerial vehicle cooperative intelligent feeding system based on the characteristics of each sub-region, the environment of a breeding region and the characteristics of breeding objects. And outputting a feeding parameter corresponding to each sub-region, thereby planning an optimal feeding path according to the feeding parameters and the current real-time state of the unmanned aerial vehicle group, realizing an intelligent system of multi-unmanned aerial vehicle intelligent scheduling, real-time linkage with breeding environment data, and automatic optimization of a feeding strategy, improving feeding efficiency and precision, reducing labor cost, and improving feeding efficiency. The breeding requirements of a fishing-light complementary scene are met.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent feeding technology for aquaculture, and in particular to a fish-light complementary unmanned aerial vehicle cooperative intelligent feeding system. BACKGROUND

[0002] In the existing aquaculture, intelligent equipment feeding technology has been gradually used instead of simple manual feeding. The most commonly used feeding method is the combination of single unmanned aerial vehicle timed and quantitative feeding and manual fixed equipment feeding. However, this feeding method still has the following shortcomings: first, the coverage of single unmanned aerial vehicle operation is limited, and when multiple unmanned aerial vehicles are operated, there is a lack of cooperative scheduling, which is prone to repeated feeding or missed feeding; second, it cannot be linked with real-time environmental data such as light and water quality of fish-light complementary ponds, and the feeding time and amount are not accurate; third, manual participation is high, which makes the feeding efficiency low and is greatly affected by the environment such as weather. SUMMARY

[0003] Therefore, the present application aims to provide a fish-light complementary unmanned aerial vehicle cooperative intelligent feeding system to solve the technical problems of poor cooperation of the existing unmanned aerial vehicle feeding system, insufficient adaptation to the fish-light complementary environment, resulting in low feeding accuracy and efficiency, and waste of feeding raw materials.

[0004] The present application discloses a fish-light complementary unmanned aerial vehicle cooperative intelligent feeding system, which comprises a data acquisition module, a data transmission module, a cooperative control center and an unmanned aerial vehicle cluster. The data acquisition module is used to acquire real-time aquaculture environment data and aquaculture object data as first acquisition data. The data transmission module is used to upload the first acquisition data to the cooperative control center. The cooperative control center further comprises: The division subunit is used to divide the pond into multiple dynamic subareas according to the first acquisition data, historical aquaculture environment and aquaculture object acquisition data, and historical feeding records. The extraction subunit is used to extract the first feature of each subarea based on the dynamic subarea division and the first acquisition data. The feeding parameter determination subunit is used to input the first feature into a feeding model to determine the feeding parameters of each subarea through the feeding model. The planning subunit is used to generate feeding tasks and path planning instructions of different priorities based on the features and corresponding feeding parameters of each subarea and the real-time state of the unmanned aerial vehicle. The unmanned aerial vehicle cluster comprises a feeding execution unit, which is used to receive and execute the feeding tasks and path planning instructions of the cooperative control center, and to feed back the operation state in real time.

[0005] Further, the process of the division subunit performing the subarea division operation specifically comprises: construct a time-series dynamic feature based on the first collection data, collection data of the historical breeding environment and the breeding object, and historical feeding records; the time-series dynamic feature includes a time-series heat map of the breeding object distribution, periodic statistics of the light intensity and the dissolved oxygen value, and a corresponding relationship between the historical feeding parameters and the aggregation response of the breeding object; perform a division operation on the pond area according to the time-series dynamic feature through a clustering algorithm to determine the sub-region boundary.

[0006] Further, the process of performing the division operation on the pond area according to the time-series dynamic feature through the clustering algorithm to determine the sub-region boundary includes: construct a multi-dimensional feature vector space based on the time-series dynamic feature; the multi-dimensional feature vector includes a breeding object distribution density index, a light intensity fluctuation index, a dissolved oxygen value change index, and a historical feeding response index; perform a clustering operation in the feature vector space to form a preliminary sub-region set; introduce a UAV accessibility constraint, an energy consumption cost constraint, and an environmental risk factor constraint in the clustering iteration update process; when the sub-region boundary violates the constraint, adjust the corresponding clustering center position and / or merge adjacent sub-regions; under the premise of meeting the constraint condition, output the final sub-region boundary after dynamic adjustment.

[0007] Further, the process in which the extraction sub-unit performs the first feature extraction operation specifically includes: determine historical region-related data of each sub-region, and extract first features based on the first collection data and the historical region-related data of each sub-region; The first features include region attribute features, environmental parameter statistical features, environmental parameter trend features, spatial uniformity features, breeding object features, historical feeding response features, task feasibility and operation cost features, and fish-light complementary features; wherein the spatial uniformity features represent the distribution consistency degree of each monitoring parameter in the sub-region.

[0008] Further, the feeding model includes a feature fusion layer, a hierarchical constraint layer, and a two-stage output layer; wherein, The feature fusion layer is configured to channelize each feature in the first features and fuse the processing results into a region feeding demand vector. The hierarchical constraint layer is configured to perform threshold truncation and time window masking on the region feeding demand vector based on an environmental parameter threshold, a time window constraint, and a stability constraint to obtain a first region feeding demand vector.

[0009] Further, the two-stage output layer, when receiving the first area feeding demand vector after the layered constraint layer processing, generates a local feeding demand candidate of each dynamic sub-area based on the first area feeding demand vector and the area attribute feature in the first stage; The second stage globally coordinates and allocates the local feeding demand candidate under the task sustainability constraint of the unmanned aerial vehicle, and outputs the final feeding parameter corresponding to each sub-area.

[0010] Further, the process of generating the feeding task priority by the planning subunit specifically includes: Based on the area attribute feature, the feeding parameter, the breeding object feature, and the environmental parameter stability feature of each dynamic sub-area, a task evaluation index set is constructed; Based on the task evaluation index set, a task priority value of each sub-area is generated, and threshold constraints are set for the breeding object feature and the environmental parameter stability feature in the task priority value generation process. When the threshold constraint is violated, the task priority value of the corresponding sub-area is reduced; According to the task priority value of each sub-area, a task priority queue is established, and a sorted feeding task list is output for execution by the unmanned aerial vehicle cluster.

[0011] Further, the path planning process specifically includes: Based on the task priority queue, and the boundary range and geometric complexity of each dynamic sub-area, a coverage constraint graph is generated; Based on the coverage constraint graph, a path generation operation is performed, and the path is checked for reachability and energy cost evaluation according to real-time state information of the unmanned aerial vehicle; After the path is generated, overlap detection and time conflict detection are performed on different unmanned aerial vehicle paths, and path adjustment operations are performed when conflicts are detected, until the task paths of each unmanned aerial vehicle meet the coverage constraint, energy constraint, and conflict resolution conditions; An adjusted path planning instruction is output.

[0012] Further, the division subunit is also used to set a buffer zone at the boundary of each divided dynamic sub-area; the width of the buffer zone is set based on historical breeding object activity range statistical results; In the task execution stage, when it is detected that the breeding object moves across the boundary of the buffer zone, the feeding parameter elastic adjustment operation is performed between adjacent sub-areas by the cooperative control center according to the crossing proportion of the breeding object.

[0013] Further, in the task execution stage, the data acquisition module is also used to acquire breeding object surface distribution data and breeding object underwater aggregation distribution data; And through the cooperative control center based on the breeding object surface distribution data and the breeding object underwater aggregation distribution data, the position of the breeding object in the sub-region is corrected in real time; Based on the real-time corrected breeding object position, the feeding path and the feeding release position of the current unmanned aerial vehicle are fine-tuned.

[0014] Compared with the prior art, the beneficial effects of the present application are: The present application divides the sub-regions for feeding in the pond dynamically, sets the feeding model, and outputs the corresponding feeding parameters of each sub-region based on the characteristics of each sub-region, the breeding area environment and the characteristics of the breeding objects, so as to plan the optimal feeding path according to the feeding parameters and the real-time state of the current unmanned aerial vehicle group, realize the intelligent system of multi-unmanned aerial vehicle intelligent scheduling, real-time linkage with breeding environment data, automatic optimization of feeding strategy, improve the feeding efficiency and accuracy, reduce the labor cost, and adapt to the breeding demand of fish-light complementary scene. BRIEF DESCRIPTION OF DRAWINGS

[0015] The drawings described herein are used to provide a further understanding of the embodiments of the present application, and form a part of the economic application, and do not constitute a limitation on the embodiments of the present application. In the drawings: Figure 1 The structure diagram of a fish-light complementary unmanned aerial vehicle cooperative intelligent feeding system disclosed by the embodiment of the present application. DETAILED DESCRIPTION

[0016] In order to make the personnel in the technical field better understand the present application scheme, the technical scheme in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. EMBODIMENT

[0017] The present application discloses a fish-light complementary unmanned aerial vehicle cooperative intelligent feeding system, please refer to Figure 1 , Figure 1 The structure diagram of a fish-light complementary unmanned aerial vehicle cooperative intelligent feeding system disclosed by the embodiment of the present application, the system comprises a data acquisition module, a data transmission module, a cooperative control center, an unmanned aerial vehicle cluster; wherein, The data acquisition module is used for real-time acquisition of breeding environment data and breeding object data as first acquisition data; The data transmission module is used for uploading the first acquisition data to the cooperative control center; The cooperative control center further comprises: The division sub-unit is used for dividing the pond into a plurality of dynamic sub-regions according to the first acquisition data, the historical breeding environment and the acquisition data of the breeding objects, and the historical feeding records; The extraction subunit is configured to extract first features of each subregion based on a dynamic subregion division condition and the first collection data; The feeding parameter determination subunit is configured to input the first features into a feeding model and determine feeding parameters of each subregion by the feeding model; The planning subunit is configured to generate feeding tasks and path planning instructions of different priorities based on the features of each subregion, the corresponding feeding parameters and the current real-time state of the unmanned aerial vehicle. The unmanned aerial vehicle cluster comprises a feeding execution subunit configured to receive and execute the feeding tasks and path planning instructions of the collaborative control center and feed back the working state in real time.

[0018] Further, the process of the division subunit performing the subregion division operation specifically comprises: constructing time-series dynamic features based on the first collection data, historical collection data of the breeding environment and the breeding object and historical feeding records; the time-series dynamic features comprise a time-series heat map of the distribution of the breeding object, periodic statistics of the light intensity and the dissolved oxygen value and a corresponding relationship between the historical feeding parameters and the aggregation response of the breeding object; performing a division operation on the pond region according to the time-series dynamic features by a clustering algorithm to determine the subregion boundary.

[0019] Further, the process of performing a division operation on the pond region according to the time-series dynamic features by a clustering algorithm to determine the subregion boundary comprises: constructing a multi-dimensional feature vector space based on the time-series dynamic features; the multi-dimensional feature vector comprises a breeding object distribution density index, a light intensity fluctuation index, a dissolved oxygen value change index and a historical feeding response index; performing a clustering operation in the feature vector space to form a preliminary subregion set; introducing an unmanned aerial vehicle accessibility constraint, an energy consumption cost constraint and an environmental risk factor constraint in the clustering iteration update process; when the subregion boundary violates the constraint, the corresponding clustering center position is adjusted and / or adjacent subregions are merged; under the premise of meeting the constraint condition, the final subregion boundary after dynamic adjustment is output.

[0020] In the embodiment of the present application, the division subunit is configured to divide the entire fish-light complementary pond into multiple dynamic subregions, so as to realize higher precision and differentiated scheduling in subsequent feeding model reasoning and unmanned aerial vehicle task allocation.

[0021] Specifically, first, multi-source data is collected and integrated to form input information for division. Among them, the breeding environment data includes but is not limited to water dissolved oxygen, pH value, water temperature, light intensity and other parameters that change dynamically over time. For example, when the breeding object is fish, the breeding object data can be the distribution density, activity frequency and location heat map of the fish school, and other indicators reflecting the behavior of the breeding organism. The historical feeding record includes the past feeding time, feeding amount and corresponding fish school gathering response during the feeding process, aiming to reveal the causal relationship between feeding and fish feeding. The above multi-source information is fused and analyzed to construct a time sequence dynamic feature.

[0022] After obtaining the time sequence dynamic feature, a multi-dimensional feature vector space is constructed based on the feature. The dimensions of the feature vector space include but are not limited to the breeding object distribution density index, the light intensity fluctuation index, the dissolved oxygen value change index and the historical feeding response index.

[0023] To ensure the comparability between different dimensions, first, each index is normalized, and a local extreme point is selected as the initial clustering center in the feature vector space, and then the region is divided based on the clustering algorithm. Specifically, by calculating the similarity measure between the sampling points and the clustering center, each sampling point is assigned to the corresponding clustering center, thereby gradually forming a preliminary sub-region set. The similarity measure not only considers the Euclidean distance of the breeding object distribution density, but also combines the difference measure and cosine similarity between the environmental parameters, so as to more accurately depict the coupling relationship between environmental factors and breeding behavior. With the advancement of iteration, the clustering center is constantly updated, and the sub-region boundary gradually converges, thereby realizing the preliminary division of the pond.

[0024] In the clustering iteration process, to avoid the occurrence of unexecutable or unreasonable division results, the application introduces multiple constraint conditions to dynamically correct the boundaries. Specifically, when the sub-region boundary violates the preset constraint, the correction mechanism is triggered. The above-mentioned preset constraint is that the first unmanned aerial vehicle reachability constraint ensures that the sub-region is within the reach of the unmanned aerial vehicle by calculating the flight radius of the unmanned aerial vehicle and the operation track. If an unreachable region is found, the clustering center position or boundary is adjusted to repair it. The second is the energy consumption cost constraint. By combining the energy consumption model of the unmanned aerial vehicle, the cost of the flight path to each candidate sub-region is calculated. If a boundary division leads to energy consumption exceeding the preset threshold, the clustering center is automatically adjusted, and / or the high-cost sub-region is merged with the adjacent low-cost region, thereby ensuring that the overall energy consumption is within a reasonable range. The third is the environmental risk factor constraint. When there are high-risk situations such as extreme light, drastic water quality fluctuations, or sudden oxygen reduction in a local region, the independence of the region is reduced during the iteration process, and the boundary of the region is contracted to the low-risk region to avoid generating feeding tasks in such regions. Through the combination of the above constraint conditions and the clustering iteration process, dynamic correction of the sub-region boundary can be realized, and the final output division result can meet the environmental rationality and the feasibility of unmanned aerial vehicle operation.

[0025] After multiple iterations and constraint corrections, the dynamically adjusted sub-region boundary is obtained. The dynamic sub-region not only reflects the behavior characteristics of the cultured objects and the state of the culture environment, but also ensures the operability and energy economy of the unmanned aerial vehicle task execution, thereby providing accurate and robust regional basis for subsequent feeding parameter determination and path planning.

[0026] Further, the extraction subunit performs a first feature extraction operation, which specifically includes: determining historical regional related data for each sub-region, and extracting first features based on the first collection data and the historical regional related data of each sub-region; The first features include regional attribute features, environmental parameter statistical features, environmental parameter trend features, spatial uniformity features, cultured object features, historical feeding response features, task feasibility and operation cost features, and fish-light complementary features. The spatial uniformity feature represents the consistency degree of the distribution of each monitoring parameter in the sub-region.

[0027] Specifically, in the embodiments of the present application, the purpose of setting the extraction subunit is to generate the first features for the feeding model facing each sub-region after completing the dynamic sub-region division. In order to ensure the stability and comparability of the features, the extraction subunit first performs unified timestamp alignment and spatial attribution mapping on the real-time collected first collection data and historical region related data on the time axis. Among them, the spatial attribution mapping refers to mapping the historical data back to the current sub-region range according to the geographical coordinates or grid index, so that the historical data and the current sub-region correspond one by one. After the data alignment is completed, the missing data completion and abnormal value identification are performed, the adjacent time point maintenance or interpolation strategy is preferentially used for completion, and the confidence degree generated in the completion process is labeled, and the confidence degree will be output together with the features for subsequent model use.

[0028] The historical region related data in the present embodiment not only includes the historical trajectory of the first collection data in the past period, but also includes multi-source historical data directly related to breeding and operation, including but not limited to historical environmental parameter time series, historical breeding object records such as stocking time and density, fry ratio, body length and weight distribution, seasonal activity law and feeding period, etc., historical feeding records such as feeding time, feeding amount, particle specification and feeding response and leftover feed record after feeding, historical equipment and operation and maintenance records such as aeration, oxygenation, inspection, no-fly or limited-fly period, etc., historical spatial substrate and morphology information such as pool bottom topography, water depth distribution, obstacle and shoreline morphology, etc., and fish-light complementary operation data such as photovoltaic array arrangement, historical power generation curve, shading time sequence and maintenance downtime, etc. In order to facilitate the analysis of trends and periodicity, the historical data is summarized in three scales of short window such as near 1-3 days, medium window such as near 1-2 weeks and long window near 1-3 months, so that the subsequent features can stably represent the sub-region state on different time scales.

[0029] On the basis of the above data preparation, the extraction subunit generates first features. Among them, the regional attribute feature is used to reflect the geometric and static environmental constraints of the sub-region, including but not limited to sub-region area, water depth statistical distribution, boundary shape complexity and shoreline tortuosity, etc. The complexity can be quantified by the turning density and concave-convex degree change of the unit boundary length to describe the curve cost basis of path planning. The environmental parameter statistical feature is used to describe the current state of the environmental monitoring quantity in the sub-region. The average value, variance, extreme value and quantile value in the same moment or in the same short window are used to reduce the influence of instantaneous noise. This kind of feature is directly calculated from the real-time first collection data and compared with the reference interval of the same period in history, which is convenient for subsequent model identification of abnormal deviation. The environmental parameter trend feature is used to represent the change direction and speed of the environment. Based on the first-order change rate, second-order change rate and multi-scale sliding difference constructed from the historical regional related data, at the same time, for light and dissolved oxygen and other parameters with obvious day and night and weekly rhythm, seasonal decomposition and periodic amplitude estimation are performed to obtain the trend strength and periodicity strength index. The spatial uniformity feature is used to describe the spatial consistency of the monitoring parameters in the sub-region. Specifically, the sensor point value is interpolated into a regular grid in the sub-region, and then the dispersion coefficient, maximum spatial gradient amplitude and spatial correlation coefficient of the grid value are calculated. When the spatial uniformity is poor, it means that there is a significant environmental difference in the same sub-region, and the subsequent feeding needs to match the coverage constraint and path refinement.

[0030] The characteristics of the breeding object are used to describe the breeding object and its behavior state, and the data comes from visual and acoustic perception and historical breeding archives. This kind of feature at least includes current and recent group distribution density, group activity, body length and weight interval proportion and breed composition, and introduces historical feeding period preference and diurnal activity difference parameters to express the timing tendency of feeding willingness.

[0031] The historical feeding response feature is mainly used to reflect the experience relationship between feeding and feeding. The saturation threshold, response slope and lag time are obtained by fitting the historical feeding amount and feeding response record, and the threshold interval and minimum safe interval length of residual food are recorded, so as to provide basis for the subsequent upper limit of feeding and time interval constraint.

[0032] The task feasibility and operation cost feature is used to represent the accessibility and cost of sub-region in execution level. Combined with the location of take-off and landing point, no-fly / limited-fly area, water surface obstacle, historical patrol path and charging turnover record, the arrival cost, path risk level and estimated energy consumption are calculated, and the execution feasibility score is output.

[0033] The fish-light complementary feature is used to consider the time sequence influence of photovoltaic side on breeding and operation, based on historical and predicted power generation power curve and shading ratio sequence, power mean, fluctuation amplitude and shading time ratio are extracted as features; when shading and high irradiance are alternately obvious, the feature can assist in determining the suitable feeding time window and unmanned aerial vehicle operation window.

[0034] In order to ensure that the features can be stably used by the subsequent feeding model, the embodiment performs unified dimension elimination and interval normalization on each feature, and retains the confidence flag bit of missing data completion and abnormality rejection in the feature vector. It needs to be emphasized that the generation of the environmental parameter trend feature must rely on the historical environmental parameter time sequence in the historical regional related data, otherwise the stable first and second order change rate and periodicity index cannot be formed; similarly, the historical feeding response feature must rely on the historical feeding record and feeding feedback, otherwise the key parameters such as saturation and lag cannot be established. These historical and real-time items together constitute the input that has explanatory power for the feeding model.

[0035] In the feature output stage, the extraction subunit concatenates various features in a fixed order into a fixed-length first feature vector, and outputs the confidence flag bit together, for the feature fusion layer of the feeding model to receive and process. Through the above operation, the first feature not only reflects the geometric and environmental state of the sub-region, the breeding object and the feeding rule, the operation and maintenance feasibility and the fish-light complementary time sequence influence, but also has traceability and stability in time and space two dimensions, thereby providing a sufficient and reproducible data basis for subsequent feeding parameter determination and cooperative scheduling.

[0036] Further, the feeding model comprises a feature fusion layer, a hierarchical constraint layer and a two-stage output layer; wherein, The feature fusion layer is used to channelize each feature in the first feature respectively, and fuse the processing result into a regional feeding demand vector; The hierarchical constraint layer is used to perform threshold truncation and time window masking on the regional feeding demand vector based on environmental parameter threshold, time window constraint and stability constraint, to obtain a first regional feeding demand vector.

[0037] Further, when the first regional feeding demand vector processed by the hierarchical constraint layer is received, the two-stage output layer generates a local feeding demand candidate of each dynamic sub-region in the first stage based on the first regional feeding demand vector and regional attribute feature; The second stage performs global coordination and allocation on the local feeding demand candidate under the constraint of unmanned aerial vehicle task sustainability, and outputs the final feeding parameters corresponding to each sub-region.

[0038] Specifically, in the embodiments of the present application, the setting feeding model is sequentially composed of a feature fusion layer, a hierarchical constraint layer and a two-stage output layer, aiming to gradually transform the data-driven feeding demand characterization into executable feeding parameters at the dynamic sub-region level, and coordinate with the execution capability of the unmanned aerial vehicle in the global range to form a closed-loop control instruction. The feeding parameters of the present application include but are not limited to feeding amount, feeding opportunity and feeding position.

[0039] In the feature fusion layer, channelization processing refers to mapping each feature in the first feature into an independent numerical channel, and each channel is subjected to unified time reference alignment, missing data marking and dimensionless processing before entering the fusion operation. Among them, in order to keep the trend information from being overwhelmed by instantaneous noise, the environmental parameter trend channel uses a time sequence segment composed of a fixed window first-order / two-order change rate and a multi-scale sliding difference as an input segment.

[0040] The multi-source vector after channelization is additionally coded with the period position of the intra-day time and the intra-week position to explicitly express the day-night and week rhythm. The fusion operation adopts the way of first local and then global, first performs small-range linear projection and nonlinear interaction within the channel to compress redundancy and extract key patterns within the channel, and then performs interaction mapping between channels to capture the coupling relationship across channels, such as the combination mode of low dissolved oxygen rising trend, high fish activity and weak light. After the above processing, the regional feeding demand vector is output, which is a fixed-length real number representation, organized by semantic sub-segments such as basic demand intensity, time elasticity, feeding saturation risk, spatial distribution sensitivity and execution feasibility confidence, to summarize the feeding tendency of the sub-region at the current time and near future window. It should be noted that the regional feeding demand vector is only a demand characterization in the feasible solution space, not a direct feeding parameter, which will be constrained and projected to executable discrete time periods and magnitudes in the subsequent layer.

[0041] The hierarchical constraint layer is used to limit the above-mentioned feeding demand characterization within the feasible region jointly defined by bio-safety and operation-safety. Among them, the environmental parameter threshold is used to express the hard limit of biology or management, such as the case of dissolved oxygen below the safe lower limit, pH exceeding the allowed interval, water temperature in the feeding inhibition interval, etc. Once the threshold is triggered, the demand component of the corresponding sub-section is immediately truncated to zero or limited to the minimum safety value, ensuring that feeding is not triggered in an unsuitable environment. The time window constraint is used to express the window when it is allowed or more suitable to execute, which is derived from the fishlight complementary shading period, equipment maintenance and no-fly period, historical feeding peak period and management preset window, etc. In implementation, the zeroization of infeasible windows and the suppression of low-optimal windows are realized by applying a masking function to the time sub-section of the demand vector. The stability constraint is used to suppress the dramatic jump of the feeding recommendation in the adjacent period, avoiding frequent switching. In specific implementation, the difference amplitude and the second-order change rate of the recommended intensity of adjacent time slices are limited, and a hysteresis mechanism is introduced. When the recommendation approaches the threshold and swings back and forth, it is required to continuously meet several time slices before allowing the threshold switching. After the hierarchical constraint layer processing, the first area feeding demand vector is obtained. Compared with the original demand, the infeasible components of the environment and time have been eliminated, and a smooth and executable demand profile is formed on the time axis.

[0042] After receiving the first area feeding demand vector, the two-stage output layer first generates the local feeding demand candidate of each dynamic sub-area in the first stage. This stage uses the first area feeding demand vector and the area attribute characteristics together to project the continuous demand intensity onto the discrete candidate time period and candidate magnitude. Specifically, select several candidate time slices corresponding to the demand peak on the preset discrete time grid, and estimate the reasonable magnitude interval of single feeding according to the area area, water depth statistical distribution and breeding population density. At the same time, the saturation threshold and the minimum safety interval in the historical feeding response are used as the upper limit and interval constraint, forming a candidate pair of time and magnitude on each candidate time slice. To avoid the feeding deviation caused by uneven spatial distribution, the candidate pair also carries a coverage sensitivity index, which is used to indicate whether the subsequent path needs to encrypt the route inside the sub-area to achieve uniform feeding. This stage only solves independently within the sub-area and does not involve resource coordination between unmanned aerial vehicles.

[0043] In the second stage, the candidate pairs of all sub-regions are globally coordinated under the task execution sustainability constraints of the UAVs, and the final feeding parameters are output. The task execution sustainability constraints depict the execution boundaries at the single machine and cluster levels, including at least the closed-loop constraints of current and predicted power and load, availability and turnaround time of take-off and landing points and charging sites, predicted flight energy consumption and operation time length, accessibility and flight path conflict restrictions, etc. The global coordination process regards each candidate pair of sub-regions as a to-be-allocated operation unit, and jointly solves in the time axis and spatial topology to make each UAV task chain meet the power and load simultaneous constraints, operation path coverage rate constraints and time window constraints. When the constraints cannot be met, the candidate pairs are preferentially reduced in magnitude or delayed to a suboptimal window, and if necessary, cross-machine transfer or split execution is triggered. After solving in this stage, the final feeding parameters of each sub-region are obtained, i.e., the explicit feeding time and corresponding feeding amount, which are issued together with the corresponding UAV task chain and operation path.

[0044] The layered structure of the above feeding model can derive reliable feeding parameters from multiple source features. The principle is to decouple the abstract demand representation, feasible region clipping and resource coordination into three stages. The feature fusion layer purifies and interacts among channels under the premise of spatiotemporal consistency, so that the generated regional feeding demand vector carries the essence of what is needed and when it is more suitable without mixing execution side constraints. The layered constraint layer limits the demand within the biological-operation double safety boundary to avoid unsuitable time periods or unstable suggestions from entering the scheduling. The two-stage output layer first lands the continuous demand into discrete candidates within the region, and then filters and rearranges them at the cluster level with sustainability constraints, and finally obtains the time-amount pairs that meet the executability and resource boundaries. Thus, the final feeding parameters for UAV collaborative feeding can be stably generated.

[0045] In the embodiment of the present application, the feeding model is modeled by combining multi-source historical data and constraint conditions, and is updated online and adaptively. Specifically, historical environmental data, related data of the breeding objects and historical feeding records are used to construct training samples, which are organized in the form of sub-regions and time slices to ensure that the training samples can cover the differences of different regions and time periods. The model is pre-trained in a self-supervised manner at the beginning of training, and the mask reconstruction and cross-modal contrast learning methods are used to enable the model to automatically capture the internal laws of environmental changes and breeding behaviors.

[0046] Then, a supervised training phase is entered, and the feeding response, residual feed amount, and dissolved oxygen disturbance amplitude of the cultured object group after feeding are used as supervision labels to guide the model to learn the corresponding relationship between the regional feeding demand and the actual response. In this process, the hierarchical constraints of the model are embedded in the training in a differentiable manner, so that the constraint conditions can participate in back propagation, thereby guiding the model to gradually converge to the solution space that meets the environmental safety threshold and time period regularity. To deal with the drift problem caused by seasonal changes and the growth cycle of the cultured object group, the invention deploys online fine-tuning based on a sliding time window. When the over-feeding rate or under-feeding rate deviates from the preset threshold, the model's partial parameters are updated in small steps, thereby maintaining the adaptability and stability of the prediction.

[0047] In the specific implementation process, each functional layer of the feeding model can be implemented based on various deep neural network architectures to adapt to different feature types and computing requirements. For example, when processing the channelization of spatial features, the feature fusion layer can adopt a convolutional neural network (CNN) structure to extract local statistical patterns and spatial correlations within the region through convolution kernels. When modeling time-series dynamic features, structures such as recurrent neural networks (RNN), long short-term memory networks (LSTM), or gated recurrent units (GRU) can be used to capture the nonlinear regularity of changes in environmental parameters and cultured object group behavior over time. In cases where both cross-regional feature interaction and global dependency need to be considered, a Transformer structure based on attention mechanisms can be used to achieve global interaction and weighted combination between features through multi-head attention mechanisms, thereby improving the expression ability of the feeding model for complex scenarios. For the final mapping and nonlinear combination of feeding parameters, a multi-layer perceptron (MLP) can be used as the output layer to realize the joint mapping of local feeding demand and global scheduling constraints. It should be noted that the selection of the above network architecture is not unique, and the invention is not limited to a specific neural network form. Any network structure that can realize the fusion and modeling of input features and output feeding parameters that meet the requirements of the invention belongs to the protection scope of the invention.

[0048] Further, the process of generating feeding task priorities by the planning subunit specifically includes: Based on the regional attribute features, feeding parameters, cultured object features, and environmental parameter stability features of each dynamic sub-region, a task evaluation index set is constructed; generate a task priority value of each sub-region based on the task evaluation index set, and set threshold constraints on the breeding object features and the environmental parameter stability features in the task priority value generation process, and reduce the task priority value of the corresponding sub-region when the threshold constraints are violated; establish a task priority queue according to the task priority values of the sub-regions, and output a sorted feeding task list for execution by the UAV cluster.

[0049] Further, the path planning process specifically includes: generate a coverage constraint graph based on the task priority queue, and the boundary range and geometric complexity of each dynamic sub-region; perform a path generation operation based on the coverage constraint graph, and perform reachability verification and energy consumption cost evaluation on the path according to real-time state information of the UAVs; After the path is generated, overlap detection and time conflict detection are performed on different UAV paths, and path adjustment operations are performed when conflicts are detected, until the task paths of the UAVs meet the coverage constraint, energy consumption constraint and conflict resolution conditions; output the adjusted path planning instructions.

[0050] Specifically, in the embodiments of the present application, the task evaluation index set is preferably constructed using the region attribute features, the breeding object features and the environmental parameter stability features, and the feeding parameters, because the region attribute features represent the objective conditions of the sub-regions, such as area, water depth, geometric complexity, etc., which can affect the coverage range and path length of the UAV feeding operation. The breeding object features can directly reflect the necessity and priority of feeding. The environmental parameter stability features are used to describe the fluctuation amplitude of parameters such as dissolved oxygen, pH, water temperature, etc. within a short time. When the environment is unstable, even if the breeding object population density is high, it is not suitable to feed immediately to prevent stress reaction. By including the above features and the feeding parameters of the corresponding sub-regions in the index set, a comprehensive evaluation of the environmental conditions, breeding needs and feeding resources can be achieved.

[0051] In the process of generating the task priority value, the planning sub-unit converts each index into a comparable score according to a predetermined measurement standard, and combines to form the task priority value of each sub-region. In this process, threshold constraints are set for the breeding object features and the environmental parameter stability features, for example, when the fish population density is below the minimum feeding threshold, or the dissolved oxygen fluctuation exceeds the safe range, the task priority value of the sub-region is automatically reduced. Through this threshold constraint mechanism, it can be prevented that the feeding priority of some regions is mistakenly raised due to abnormal environment or unsuitable conditions, thereby improving the rationality of task scheduling.

[0052] In the path planning stage, the planning subunit first generates a coverage constraint graph based on the task priority queue, the boundary range of each dynamic sub-region, and the geometric complexity. Specifically, the coverage constraint graph is constructed by discretizing the sub-region boundary into a grid of cells and labeling each cell with a task demand weight that is positively correlated with the task priority value of the sub-region to ensure that high-priority regions are preferentially covered during path generation. In this constraint graph, different cells are connected by edges, forming a weighted graph structure, where the edge weights represent the spatial cost of moving between two cells. By performing a shortest path or heuristic search on this graph model, a set of candidate paths that cover all target cells is obtained.

[0053] At the same time as path generation, real-time state information of the UAV is combined to perform reachability verification and energy consumption cost evaluation. Reachability verification includes two aspects: whether the path length exceeds the flight radius supported by the current remaining power of the UAV, and whether the path segment crosses a region with geographical obstacles or beyond line of sight. Energy consumption cost evaluation takes the battery power, current load weight, and spatial length of the path as input, calculates the energy consumption prediction value of a single path based on an empirical energy consumption function, and compares it with the upper limit of the available energy of the UAV. If a candidate path does not meet the conditions in terms of reachability or energy consumption, the planning subunit adjusts the path, including shortening the path range, executing tasks in segments, or reassigning tasks to other UAVs, to ensure that the final path is executable.

[0054] After path generation is complete, the planning subunit further performs overlap detection and time conflict detection on multiple UAV paths. Overlap detection is performed by calculating the spatial intersection and time overlap of different UAV paths. If it is found that two paths have spatial overlap in the same region and overlap in time, it is determined that there is a potential risk of repeated feeding. Time conflict detection checks whether different UAVs will enter adjacent or the same sub-region at the same time based on the path timeline. If such a situation exists, it is determined to be a conflict.

[0055] When a conflict is detected, the planning subunit performs correction based on a path adjustment algorithm. The correction method specifically solves time conflicts by delaying the start time of part of the tasks, eliminates spatial overlap by re-planning the local path in the coverage constraint graph, and transfers part of the tasks of a sub-region to other paths or UAVs through dynamic task splitting. This correction process is performed iteratively, and each adjustment requires re-verification of coverage, energy consumption, and conflict status until the task paths of all UAVs meet the coverage constraint, energy consumption constraint, and conflict resolution conditions. Finally, the planning subunit outputs the optimized path planning instructions to ensure that the UAV swarm can efficiently and cooperatively complete the feeding task under limited energy consumption and dynamic environmental conditions.

[0056] Through the above operation, the planning subunit can organically combine task priority generation and path planning, the former ensures the rationality and urgency of task allocation, and the latter guarantees the feasibility and safety of the unmanned aerial vehicle at the execution level, so as to realize efficient and collaborative operation of the whole system in a dynamic environment.

[0057] Further, the dividing subunit is further used for setting a buffer zone at the boundary of each divided dynamic subregion; and the width of the buffer zone is set based on statistical results of the historical activity range of the breeding object. In the task execution phase, when it is detected that the breeding object moves across the boundary of the buffer zone, the cooperative control center performs an elastic adjustment operation of feeding parameters between adjacent subregions according to the crossing ratio of the breeding object.

[0058] In the embodiment of the application, the dividing subunit is not only used for dividing the pond into multiple dynamic subregions based on time sequence dynamic characteristics, but also further sets a buffer zone at the boundary of each dynamic subregion to adapt to the actual swimming behavior of the breeding object. Specifically, the width of the buffer zone is not fixedly set, but is dynamically determined according to statistical results of the historical activity range of the breeding object. For example, by long-term collection of fish group position distribution data, a frequency distribution curve of the activity of the fish group in the adjacent region of each boundary is obtained, and the width of the buffer zone is set according to the 95% coverage range of the distribution, so as to ensure that the buffer zone can cover most of the common cross-border swimming behaviors.

[0059] In the task execution phase, the cooperative control center continuously receives real-time breeding object group position data uploaded from the visual monitoring and / or underwater acoustic detection device. When it is detected that the breeding object moves across the boundary of the buffer zone, the crossing ratio, i.e. the ratio of the number of breeding objects entering the adjacent subregion to the total number in the buffer zone, is calculated. The crossing ratio reflects the dynamic trend of the distribution of the breeding objects between the two subregions.

[0060] After the crossing ratio is determined, the cooperative control center triggers the elastic adjustment mechanism of the feeding parameters. Specifically, if it is detected that a of the breeding objects in a subregion cross into the adjacent subregion, the feeding demand value of the subregion is reduced by a proportion of 1-a, and the feeding demand value of the adjacent subregion is supplemented by a proportion of a. The adjustment process is based on real-time statistical results in the buffer zone, ensuring that the feeding amount can dynamically change with the actual position distribution of the fish group, and will not be mispositioned or missed due to the static nature of the boundary division.

[0061] In the process of executing the elastic adjustment, the cooperative control center also performs secondary optimization according to the execution task sustainability constraint of the unmanned aerial vehicle. If the adjacent subregion after the supplement exceeds the upper limit of the feeding load of the unmanned aerial vehicle, the supplement amount is split for shared execution by multiple unmanned aerial vehicles, or part of the supplement task is delayed to the next time window, so as to ensure the feasibility of the overall feeding plan.

[0062] Through the above buffer zone and elastic adjustment operation, the application realizes the flexibility of sub-area division, ensures the spatial accuracy brought by the division of the region boundary, avoids the boundary effect caused by the natural swimming of the breeding objects, ensures that the feeding strategy can dynamically adapt to the real distribution state of the breeding objects, and thus improves the feeding efficiency and resource utilization.

[0063] Further, in the task execution phase, the data acquisition module is further used to acquire surface layer distribution data of the breeding objects and underwater aggregation distribution data of the breeding objects; And through the cooperative control center, the position of the breeding objects in the sub-area is corrected in real time based on the surface layer distribution data of the breeding objects and the underwater aggregation distribution data of the breeding objects; Based on the real-time corrected position of the breeding objects, fine adjustment operation is performed on the feeding path and the feeding release position of the current unmanned aerial vehicle.

[0064] In the embodiment of the application, the data acquisition module is not only used to acquire conventional environmental parameters and breeding object related data, but also further used to acquire surface layer distribution data and underwater aggregation distribution data of the breeding objects in the task execution phase. Specifically, the surface layer distribution data mainly reflects the activity density and position distribution of the breeding object group in the surface layer of the water body, and the underwater aggregation distribution data can reflect the aggregation area and depth distribution characteristics of the underwater breeding object group.

[0065] After receiving the above data, the cooperative control center corrects the position of the breeding objects in the sub-area in real time by using a data fusion algorithm. Specifically, first, the surface layer distribution data and the underwater distribution data are aligned in spatial coordinates, and the overall three-dimensional distribution of the breeding object group is obtained through depth weighted matching. If it is detected that there is a significant difference between the surface layer and the underwater distribution, the underwater aggregation distribution is given priority and the surface layer distribution is given priority, and a weighted correction result is formed, so as to avoid the deviation caused by a single data source.

[0066] After completing the real-time correction of the position of the breeding objects, the cooperative control center fine adjusts the feeding path and the feeding release position of the unmanned aerial vehicle based on the corrected position. Specifically, in the process of path generation, the track nodes of the unmanned aerial vehicle are adjusted locally, so that the feeding release point of the unmanned aerial vehicle and the corrected aggregation center position of the breeding object group coincide as much as possible, and when the group as a whole dives or floats, the feeding release height and the release timing are fine adjusted to ensure that the feed can be effectively fed to the aggregation area of the breeding object group.

[0067] Through the above operation, the application can realize real-time correction of the position of the fish group in the sub-area under the condition that the distribution of the breeding objects changes dynamically with time and environment, so that the feeding path and the release point of the unmanned aerial vehicle always coincide with the actual distribution of the breeding object group, thereby improving the feeding accuracy and reducing the waste of feed.

[0068] As a further preferred, the cooperative control center is not only used for task allocation and path planning, but also further performs homeward journey and supplementary feeding operations based on the feedback information of the UAV cluster. Specifically, during the task execution phase, the UAV cluster periodically reports task feedback information to the cooperative control center, including but not limited to the current remaining load capacity, the remaining power level, the task completion progress, and the environmental monitoring data.

[0069] When the cooperative control center detects that the remaining load of a UAV is insufficient to complete the unexecuted feeding task, or its remaining power is lower than the preset threshold, the homeward journey decision is triggered. At this time, the cooperative control center calculates the optimal homeward journey path according to the current position of the UAV and the geographical position of each landing point. The generation process of the path considers the flight distance, power consumption model and obstacle avoidance rules to ensure that the UAV can safely return to the nearest landing point to complete the supply or charging operation before the power is completely consumed.

[0070] At the same time, the cooperative control center also monitors the task execution results of the sub-regions. When it detects that the actual feeding amount of a certain sub-region is lower than the preset target, or identifies through the on-board visual and acoustic sensors that the target feeding state of the aquaculture object group has not been reached, the cooperative control center will trigger the supplementary feeding operation. Specifically, based on the uncompleted task amount, a supplementary feeding task is generated, and then the available UAVs are scheduled again in the global range. In this scheduling process, the cooperative control center will comprehensively consider the remaining load capacity of the UAV, the flight accessibility and the pond environment safety factor. For example, if a certain region is judged as a high-risk region due to abnormal water quality or excessive light, the UAV is restricted from performing supplementary feeding in this region, and other regional UAVs or standby UAVs are used to perform the operation.

[0071] After completing the above-mentioned secondary scheduling, the cooperative control center outputs the supplementary feeding task list and generates new path planning and feeding instructions for each supplementary feeding task. The new path planning not only needs to meet the coverage rate and energy consumption constraints, but also needs to coordinate with the original task path to avoid introducing path conflicts or time conflicts due to the supplementary feeding operation. The execution results of the supplementary feeding task are fed back to the cooperative control center again, forming a closed-loop monitoring and scheduling process.

[0072] Through the above operation, the present application can realize dynamic homeward journey and supplementary feeding scheduling during the task execution of the UAV, so that the system can still maintain the continuity and effectiveness of task execution when facing power shortage, insufficient load or unsatisfactory feeding effect and other sudden situations. This operation significantly improves the task completion rate of the UAV cluster in the fish-light complementary feeding scene, and reduces the insufficient feeding and feed waste caused by mid-way failure or environmental abnormalities.

[0073] Finally, it should be noted that the above-described embodiments include a plurality of parallel embodiments of the present application, and deleting or otherwise adjusting one or more embodiments does not affect the implementation of the scheme. In addition, the disclosed fishing light complementary unmanned aerial vehicle cooperative intelligent feeding system of the embodiments of the present application is only the preferred embodiment of the present application, and is only used to illustrate the technical solutions of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand; the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A collaborative intelligent feeding system for fishery-solar hybrid unmanned aerial vehicles, characterized in that, The system includes a data acquisition module, a data transmission module, a collaborative control center, and a drone swarm; wherein, The data acquisition module is used to collect real-time data on the aquaculture environment and the aquaculture objects, serving as the primary data to be collected. The data transmission module is used to upload the first collected data to the collaborative control center; The collaborative control center also includes: The sub-unit division is used to divide the pond into multiple dynamic sub-regions based on the first collected data, historical aquaculture environment and aquaculture object collected data, and historical feeding records; The extraction sub-unit is used to extract the first feature of each sub-region based on the dynamic sub-region division and the first acquired data; The feeding parameter determination subunit is used to input the first feature into the feeding model and determine the feeding parameters for each sub-region through the feeding model; The planning sub-unit is used to generate feeding tasks and path planning instructions with different priorities based on the characteristics of each sub-region, the corresponding feeding parameters, and the current real-time status of the UAV. The drone swarm includes a feeding execution unit, which receives and executes the feeding tasks and path planning instructions from the collaborative control center, and provides real-time feedback on the operation status.

2. The fishery-solar hybrid drone-assisted intelligent feeding system according to claim 1, characterized in that, The process of performing sub-region division operation in the sub-unit specifically includes: Based on the first collected data, historical aquaculture environment and aquaculture object data, and historical feeding records, a time-series dynamic feature is constructed; the time-series dynamic feature includes a time-series heat map of the distribution of aquaculture objects, periodic statistics of light intensity and dissolved oxygen value, and the correspondence between historical feeding parameters and the aggregation response of aquaculture objects; The pond region is divided using a clustering algorithm based on the aforementioned temporal dynamic characteristics to determine the boundaries of the sub-regions.

3. The fishery-solar hybrid drone-assisted intelligent feeding system according to claim 2, characterized in that, The process of dividing the pond region according to the time-series dynamic features using a clustering algorithm and determining the boundaries of the sub-regions includes: A multidimensional feature vector space is constructed based on the aforementioned time-series dynamic features; the multidimensional feature vector includes the distribution density index of the cultured objects, the light intensity fluctuation index, the dissolved oxygen value change index, and the historical feeding response index; Clustering operations are performed in the feature vector space to form a preliminary set of sub-regions; During the clustering iterative update process, constraints on drone accessibility, energy consumption cost, and environmental risk factors are introduced. When the sub-region boundary violates the constraints, the corresponding cluster center position is adjusted and / or adjacent sub-regions are merged. Under the premise of satisfying the constraints, the dynamically adjusted final sub-region boundary is output.

4. The fishery-solar hybrid drone-assisted intelligent feeding system according to claim 1, characterized in that, The process of the extraction subunit performing the first feature extraction operation specifically includes: Determine the historical region-related data for each sub-region, and extract the first feature based on the first collected data and historical region-related data for each sub-region; The first feature includes regional attribute features, environmental parameter statistical features, environmental parameter trend features, spatial uniformity features, aquaculture object features, historical feeding response features, task feasibility and operation and maintenance cost features, and aquaculture-solar complementarity features; wherein, the spatial uniformity feature represents the degree of consistency in the distribution of each monitoring parameter within the sub-region.

5. The fishery-solar hybrid drone-assisted intelligent feeding system according to claim 4, characterized in that, The feeding model includes a feature fusion layer, a hierarchical constraint layer, and a two-stage output layer; wherein... The feature fusion layer is used to perform channelization processing on each feature in the first feature, and fuse the processing results into a regional feeding demand vector. The hierarchical constraint layer is used to perform threshold truncation and time window masking on the regional feeding demand vector based on environmental parameter thresholds, time window constraints, and stability constraints to obtain the first regional feeding demand vector.

6. The fishery-solar hybrid drone-assisted intelligent feeding system according to claim 5, characterized in that, When the dual-stage output layer receives the first region feeding demand vector after it has been processed by the hierarchical constraint layer, it generates local feeding demand candidates for each dynamic sub-region based on the first region feeding demand vector and the region attribute features in the first stage. In the second stage, under the constraint of the sustainability of UAV mission execution, the local feeding demand candidates are globally coordinated and allocated, and the final feeding parameters corresponding to each sub-region are output.

7. The fishery-solar hybrid drone-assisted intelligent feeding system according to claim 3, characterized in that, The process of generating feeding task priorities in the planning sub-unit specifically includes: Based on the regional attribute characteristics, feeding parameters, characteristics of aquaculture objects, and stability characteristics of environmental parameters of each dynamic sub-region, a set of task evaluation indicators is constructed. Based on the set of task evaluation indicators, a task priority value is generated for each sub-region. During the task priority value generation process, threshold constraints are set on the characteristics of the aquaculture object and the stability characteristics of environmental parameters. When the threshold constraints are violated, the task priority value of the corresponding sub-region is reduced. A task priority queue is established based on the task priority value of each sub-region, and a sorted list of feeding tasks is output for the drone cluster to execute.

8. The fishery-solar hybrid drone-assisted intelligent feeding system according to claim 7, characterized in that, The path planning process specifically includes: Based on the task priority queue and the boundary range and geometric complexity of each dynamic sub-region, a coverage constraint graph is generated. The path generation operation is performed based on the coverage constraint graph, and the reachability and energy consumption cost of the path are verified based on the real-time status information of the UAV. After the path is generated, overlap detection and time conflict detection are performed on different UAV paths. When a conflict is detected, the path adjustment operation is performed until the mission path of each UAV meets the coverage constraint, energy consumption constraint and conflict resolution condition. Output the adjusted path planning instructions.

9. The fishery-solar hybrid drone-assisted intelligent feeding system according to any one of claims 1-8, characterized in that, The division sub-unit is also used to set buffer zones at the boundaries of each divided dynamic sub-region; the width of the buffer zone is set based on the statistical results of the historical activity range of the aquaculture objects. During the task execution phase, when it is detected that the aquaculture object is moving across the boundary of the buffer zone, the feeding parameters are flexibly adjusted between adjacent sub-regions according to the crossing ratio of the aquaculture object through the collaborative control center.

10. The fishery-solar hybrid UAV-assisted intelligent feeding system according to claim 9, characterized in that, During the task execution phase, the data acquisition module is also used to collect surface distribution data and underwater aggregation distribution data of the aquaculture objects. The collaborative control center then performs real-time corrections on the location of the aquaculture objects within the sub-region based on the surface distribution data and underwater aggregation distribution data of the aquaculture objects. Based on the real-time correction of the location of the aquaculture object, fine-tuning is performed on the current feeding path and feeding release location of the drone.

Images