Multi-unmanned aerial vehicle multi-task scheduling method and system

Abstract

The invention provides an unmanned aerial vehicle multi-machine multi-task scheduling method and system, and the method comprises the steps: extracting core features correspondingly contained in native protocol data, and creating a corresponding dynamic protocol feature map based on an unsupervised clustering algorithm according to the core features; synchronously executing protocol feature consistency verification, instruction semantic reasonability verification and unmanned aerial vehicle state feedback verification on the multiple types of unmanned aerial vehicles through the dynamic protocol feature atlas; collecting a verification result and an unmanned aerial vehicle verification parameter, and constructing a corresponding multi-target optimization function according to the verification result and the unmanned aerial vehicle verification parameter; and with task completion efficiency maximization, data conversion delay minimization and analysis error rate minimization as targets, an original calling priority in an original task allocation scheme is dynamically adjusted through a multi-target optimization function to generate a target calling priority, and scheduling of multiple types of unmanned aerial vehicles is completed according to the target calling priority. The unmanned aerial vehicle scheduling efficiency can be effectively improved.

Description

TECHNICAL FIELD

[0001] The application relates to the field of unmanned aerial vehicles (UAVs), and particularly to a multi-UAV multi-task scheduling method and system BACKGROUND With the rapid development of UAV technology, its application scenarios have widely covered aerial surveying and mapping, logistics distribution, emergency rescue, agricultural plant protection and other fields. Multi-UAV cooperative operation has become a key direction for improving operation efficiency and expanding application boundaries. Communication between UAVs and between a UAV and a ground control system is the core support for realizing cooperative task scheduling, real-time data transmission and accurate state feedback. As the rule basis for data interaction, a communication protocol directly determines the stability and compatibility of a communication link. At present, there are a large number of UAV manufacturers in the market. Different manufacturers adopt different communication protocols based on their own technical routes, product positioning and application scenario requirements. Common communication protocols include a lightweight message transmission protocol MQTT, a full-duplex communication protocol WebSocket and a self-defined private protocol, forming a diversified protocol application pattern.

[0002] Among them, the heterogeneity of communication protocols of UAVs from different manufacturers leads to a serious protocol incompatibility problem. Because various protocols have significant differences in core technical parameters such as data frame format, transmission rate, interaction logic and check mechanism, UAVs from different manufacturers cannot directly establish an effective communication connection, and it is also difficult to realize data interconnection and uniform scheduling of instructions. To solve this problem, the prior art usually needs to develop a corresponding adaptation module for each protocol, and realizes the cooperation of UAVs from different manufacturers through protocol conversion, data format analysis and instruction mapping. This process not only needs to invest a large amount of research and development resources, but also needs to be customized for the characteristics of different protocols, greatly increasing the design complexity of a multi-UAV cooperative system.

[0003] Further, the additional adaptation requirement caused by protocol incompatibility further brings a series of chain problems. On the one hand, the diversification of adaptation modules leads to a bloated system architecture, increasing the difficulty of later maintenance and upgrading. When a new protocol type of UAV is connected, an adaptation module needs to be redeveloped, which seriously limits the scalability of the system. On the other hand, problems such as data conversion delay and instruction analysis error existing in the multi-protocol adaptation process may lead to a decrease in the real-time performance of UAV cooperative operation and a decrease in the task execution accuracy, and even cause risks such as communication interruption and cooperation failure, restricting the popularization and application of multi-UAV cooperative technology in large-scale and high-complexity scenarios. SUMMARY

[0004] Based on this, the purpose of the present application is to provide a multi-vehicle multi-task scheduling method and system for unmanned aerial vehicles, so as to solve the problem that in the multi-protocol adaptation process of multiple unmanned aerial vehicles, new adaptation modules need to be developed continuously, and there are data conversion delay and instruction parsing error in the prior art.

[0005] The first aspect of the embodiment of the present application proposes: A multi-vehicle multi-task scheduling method for unmanned aerial vehicles, specifically comprising the following steps Collecting native protocol data of multiple types of unmanned aerial vehicles to be scheduled, and extracting a plurality of core features contained in the native protocol data, to create a corresponding dynamic protocol feature map based on an unsupervised clustering algorithm according to the plurality of core features; When receiving an original task allocation scheme, performing protocol feature consistency verification, instruction semantic rationality verification and unmanned aerial vehicle state feedback verification on the multiple types of unmanned aerial vehicles simultaneously through the dynamic protocol feature map; Collecting verification results and unmanned aerial vehicle verification parameters, and constructing a corresponding multi-objective optimization function according to the verification results and the unmanned aerial vehicle verification parameters; Taking the maximization of task completion efficiency, the minimization of data conversion delay and the minimization of parsing error rate as the goal, dynamically adjusting the original call priority in the original task allocation scheme through the multi-objective optimization function to generate a target call priority, and completing the scheduling of the multiple types of unmanned aerial vehicles according to the target call priority.

[0006] The beneficial effects of the present application are: by collecting native protocol data of multiple types of unmanned aerial vehicles, a plurality of core features for subsequent analysis can be extracted, based on which a corresponding dynamic protocol feature map is created, when receiving an original task allocation scheme, multiple verifications can be performed, and an adaptive multi-objective optimization function can be generated, so that the original task allocation scheme can be dynamically adjusted to generate a corresponding target call priority, so that no additional adaptation module is needed, and the data processing can be quickly and effectively completed, and the scheduling efficiency of the unmanned aerial vehicle is improved.

[0007] Further, the step of creating a corresponding dynamic protocol feature map based on an unsupervised clustering algorithm according to a plurality of core features comprises: Dividing a plurality of core features into a plurality of feature subsets according to the protocol type of the unmanned aerial vehicle, and using a heterogeneous feature adaptation coding algorithm based on protocol attributes to map the core features of different feature subsets to a unified dimension feature space to generate a corresponding standard core feature set; The instruction format features in the standard core feature set are clustered in a first layer by using a hierarchical density clustering algorithm to obtain a plurality of basic feature clusters, and the cooperative triggering frequency of the core features inside each basic feature cluster is calculated, and a second layer of density clustering is performed based on the cooperative triggering frequency to obtain a plurality of protocol feature clusters; An initial protocol feature map is constructed based on the plurality of protocol feature clusters, and the initial protocol feature map is dynamically updated to generate the dynamic protocol feature map.

[0008] Further, the step of constructing an initial protocol feature map based on a plurality of protocol feature clusters and dynamically updating the initial protocol feature map to generate the dynamic protocol feature map comprises: The functional semantic labels of the core features in each protocol feature cluster are extracted, and the semantic similarity and triggering time sequence correlation between different protocol feature clusters are calculated based on the instruction interaction time sequence relationship in the multi-task scenario of the unmanned aerial vehicle, so as to correspondingly create the initial protocol feature map; New protocol data corresponding to the unmanned aerial vehicle performing tasks is collected, and the new protocol data is locally clustered to generate a temporary feature cluster; The semantic compatibility and time sequence matching degree between the temporary feature cluster and each graph node in the initial protocol feature map are calculated, and the graph topology structure of the initial protocol feature map is adjusted based on an incremental update algorithm according to the semantic compatibility and the time sequence matching degree to generate the dynamic protocol feature map.

[0009] Further, the step of constructing a corresponding multi-objective optimization function according to the verification result and the unmanned aerial vehicle verification parameter comprises: The feature matching degree, semantic confidence and state adaptation rate in the verification result are extracted, a corresponding coupling coefficient is calculated, and an adaptive optimization target is generated; The fluctuation variance of the unmanned aerial vehicle inspection parameter is calculated based on a sliding time window, an adaptive compensation term is added to the optimization target according to the fluctuation variance, and a historical scheduling data feedback model is constructed to extract the association rule of historical verification results and scheduling effects; The initial optimization function is adaptively adjusted according to the adaptive compensation term and the association rule to generate the multi-objective optimization function.

[0010] Further, the step of constructing a corresponding multi-objective optimization function according to the verification result and the unmanned aerial vehicle verification parameter comprises: According to the nonlinear correlation coefficient in the correlation law, a plurality of target balance effective samples are extracted from historical scheduling data, and a weight distribution matrix is generated by mapping and quantifying the contribution weight of each optimization target through a Gaussian kernel function, so as to give adaptive weights to the objective terms of the initial optimization function; According to the fluctuation variance classification of the unmanned aerial vehicle verification parameters, a differentiated compensation coefficient is matched for different fluctuation levels, and the adaptive compensation term is converted into a targeted constraint condition; The adaptive weights, the differentiated compensation coefficient and the targeted constraint condition are embedded into the initial optimization function, and a non-dominated sorting genetic algorithm is used for multiple rounds of iterative solving to generate the multi-objective optimization function.

[0011] Further, the step of dynamically adjusting the original call priority in the original task allocation scheme through the multi-objective optimization function to generate a target call priority comprises: Based on the analysis error rate distribution in the verification result and the task processing capability difference in the unmanned aerial vehicle verification parameters, an entropy weight-analytic hierarchy process is used to dynamically allocate optimization weights of task completion efficiency, data conversion delay and analysis error rate; According to the optimization weights and the dynamic protocol feature map, a corresponding task-unmanned aerial vehicle bidirectional adaptation matrix is constructed, and a comprehensive optimization value of each element in the task-unmanned aerial vehicle bidirectional adaptation matrix is calculated through the multi-objective optimization function; The comprehensive optimization value is sorted to generate the target call priority.

[0012] Further, the step of sorting the comprehensive optimization value to generate the target call priority comprises: In combination with the updated features in the dynamic protocol feature map and the remaining load margin in the unmanned aerial vehicle verification parameters, the comprehensive optimization value is time-effectively weighted and corrected to obtain a corrected optimization value; Based on the corrected optimization value, a hierarchical matching queue is constructed by fusing task urgency weight and unmanned aerial vehicle protocol adaptation stability score; Each element in the hierarchical matching queue is locally optimally sorted to generate the target call priority.

[0013] The second aspect of the embodiment of the application provides: A multi-machine multi-task scheduling system of unmanned aerial vehicles, wherein the system comprises: A collection module is configured to collect original protocol data of a plurality of types of unmanned aerial vehicles to be scheduled, and extract a plurality of core features contained in the original protocol data, so as to create a dynamic protocol feature map based on the plurality of core features according to an unsupervised clustering algorithm; The verification module is configured to perform protocol feature consistency verification, instruction semantic rationality verification and unmanned aerial vehicle state feedback verification on the multi-type unmanned aerial vehicles simultaneously through the dynamic protocol feature map when the original task allocation scheme is received. The construction module is configured to collect verification results and unmanned aerial vehicle verification parameters, and construct a corresponding multi-objective optimization function according to the verification results and the unmanned aerial vehicle verification parameters. The calling module is configured to dynamically adjust an original calling priority in the original task allocation scheme through the multi-objective optimization function to generate a target calling priority, and complete the scheduling of the multi-type unmanned aerial vehicles according to the target calling priority, with the goal of maximizing task completion efficiency, minimizing data conversion delay and minimizing error rate.

[0014] Further, the collection module is specifically configured to: The core features are divided into a plurality of feature subsets according to the protocol types of the unmanned aerial vehicles, and a heterogeneous feature adaptation coding algorithm based on protocol attributes is used to map the core features of different feature subsets to a feature space of a uniform dimension to generate a corresponding standard core feature set. A hierarchical density clustering algorithm is used to perform first layer clustering on the instruction format features in the standard core feature set to obtain a plurality of basic feature clusters, and the cooperative triggering frequency of the core features inside each basic feature cluster is calculated, and second layer density clustering is performed based on the cooperative triggering frequency to obtain a plurality of protocol feature clusters. An initial protocol feature map is constructed based on the plurality of protocol feature clusters, and the initial protocol feature map is dynamically updated to generate the dynamic protocol feature map.

[0015] Further, the collection module is specifically configured to: The functional semantic labels of the core features in each protocol feature cluster are extracted, and the semantic similarity and triggering timing correlation between different protocol feature clusters are calculated based on the instruction interaction timing relationship in the unmanned aerial vehicle multi-task scenario, so as to correspondingly create the initial protocol feature map. New protocol data generated by the unmanned aerial vehicles in the process of executing tasks is collected, and the new protocol data is locally clustered to generate temporary feature clusters. The semantic compatibility and timing matching degree between the temporary feature clusters and each graph node in the initial protocol feature map are calculated, and the graph topology structure of the initial protocol feature map is adjusted based on an incremental update algorithm according to the semantic compatibility and the timing matching degree to generate the dynamic protocol feature map.

[0016] Further, the construction module is specifically configured to: Extracting the feature matching degree, semantic confidence and state adaptation rate in the verification result, calculating the corresponding coupling coefficient, and generating an adaptive optimization target; Based on the sliding time window, the fluctuation variance of the unmanned aerial vehicle inspection parameters is calculated, an adaptive compensation term is added to the optimization target according to the fluctuation variance, and a historical scheduling data feedback model is constructed to extract the correlation between historical verification results and scheduling effects; According to the adaptive compensation term and the correlation, the initial optimization function is adaptively adjusted to generate the multi-objective optimization function.

[0017] Further, the construction module is specifically used for: According to the nonlinear correlation coefficient in the correlation, a multi-objective balanced effective sample is extracted from historical scheduling data, and the contribution weight of each optimization target is mapped and quantified by a Gaussian kernel function to generate a corresponding weight distribution matrix and give the adaptive weight to the objective term of the initial optimization function; According to the fluctuation variance of the unmanned aerial vehicle inspection parameters, the fluctuation variance is graded, and the adaptive compensation term is converted into a targeted constraint condition by matching different compensation coefficients for different fluctuation levels; The adaptive weight, the differentiated compensation coefficient and the targeted constraint condition are embedded in the initial optimization function, and a non-dominated sorting genetic algorithm is used for multi-round iteration to generate the multi-objective optimization function.

[0018] Further, the calling module is specifically used for: Based on the error rate distribution in the verification result and the task processing capability difference in the unmanned aerial vehicle inspection parameters, the optimization weights of task completion efficiency, data conversion delay and error rate are dynamically allocated by using the entropy weight-hierarchical analysis method; According to the optimization weight and the dynamic protocol feature map, a corresponding task-unmanned aerial vehicle bidirectional adaptation matrix is constructed, and the comprehensive optimization value of each element in the task-unmanned aerial vehicle bidirectional adaptation matrix is calculated by the multi-objective optimization function; The comprehensive optimization value is sorted to generate the target calling priority.

[0019] Further, the calling module is specifically used for: Combined with the updated features in the dynamic protocol feature map and the remaining load margin in the unmanned aerial vehicle inspection parameters, the comprehensive optimization value is time-weighted corrected to obtain a corrected optimization value; Based on the corrected optimization value, a hierarchical matching queue is constructed by fusing the task urgency weight and the unmanned aerial vehicle protocol adaptation stability score; Each element in the hierarchical matching queue is locally optimized and sorted to generate the target calling priority.

[0020] The third aspect of the embodiments of the present application provides: A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the unmanned aerial vehicle multi-machine multi-task scheduling method as described above when executing the computer program.

[0021] The fourth aspect of the embodiments of the present application provides: A readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement the unmanned aerial vehicle multi-machine multi-task scheduling method as described above.

[0022] Additional aspects and advantages of the present application will be made apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 A flow chart of the unmanned aerial vehicle multi-machine multi-task scheduling method provided by the first embodiment of the present application; Figure 2 A structure block diagram of the unmanned aerial vehicle multi-machine multi-task scheduling system provided by the third embodiment of the present application.

[0024] The following detailed description will further illustrate the present application in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0025] In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the accompanying drawings. The drawings show several embodiments of the present application. However, the present application can be realized in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present application more thorough and comprehensive.

[0026] It should be noted that when an element is referred to as being "fixedly attached" to another element, it can be directly on the other element or there can be an intervening element. When an element is referred to as being "connected" to another element, it can be directly connected to the other element or intervening elements can be present. As used herein the terms "vertical", "horizontal", "left", "right", and the like are merely used for the purpose of illustration.

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0028] Referring to Figure 1 The unmanned aerial vehicle multi-machine multi-task scheduling method provided by the first embodiment of the application can quickly and effectively complete the scheduling of multiple unmanned aerial vehicles without an adaptation module, thereby improving the scheduling efficiency of the unmanned aerial vehicles.

[0029] Specifically, the embodiment provides: An unmanned aerial vehicle multi-machine multi-task scheduling method, specifically comprising the following steps Step S10, collecting native protocol data of multiple types of unmanned aerial vehicles to be scheduled, and extracting a plurality of core features contained in the native protocol data to create a corresponding dynamic protocol feature map based on an unsupervised clustering algorithm according to the plurality of core features; It should be noted that first, the native protocol data of the multiple types of unmanned aerial vehicles to be scheduled needs to be collected. Specifically, such data contains core information such as the communication rules, instruction format, and state feedback of the unmanned aerial vehicles, and is the basis for scheduling adaptation. Since the protocols of unmanned aerial vehicles of different manufacturers and models are heterogeneous (such as differences in instruction encoding format and data transmission rate), a plurality of core features need to be extracted, including protocol instruction format, data transmission delay threshold, state feedback field type, and task execution capability identifier. These features directly determine the feasibility of the adaptation of the unmanned aerial vehicles to the tasks. Based on the unsupervised clustering algorithm (which does not require manual annotation of categories and can automatically mine feature similarity), the core features are clustered to create a dynamic protocol feature map. Specifically, the map classifies unmanned aerial vehicles with similar protocol features, directly presents the protocol compatibility range and interaction characteristics of different unmanned aerial vehicles, and provides a "protocol dictionary" for subsequent verification and scheduling.

[0030] Step S20, when receiving an original task allocation scheme, performing protocol feature consistency verification, instruction semantic rationality verification, and unmanned aerial vehicle state feedback verification on the multiple types of unmanned aerial vehicles simultaneously through the dynamic protocol feature map; It should be noted that when receiving the original task allocation scheme, the dynamic protocol feature map becomes the core basis for verification, and the three verifications are performed simultaneously to avoid scheduling risks: protocol feature consistency verification (judging whether the unmanned aerial vehicle protocol supports the task instruction format to avoid the inability to parse the instruction), instruction semantic rationality verification (verifying the matching degree of the task instruction and the unmanned aerial vehicle function, such as prohibiting the allocation of a bombing task to a reconnaissance unmanned aerial vehicle), and unmanned aerial vehicle state feedback verification (combining the real-time power, load, and position of the unmanned aerial vehicle to judge whether the execution condition is met). The three verifications form a "safety threshold" for scheduling, avoiding task failure due to protocol incompatibility, instruction errors, or abnormal unmanned aerial vehicle state. This facilitates subsequent processing.

[0031] In step S30, the verification result and the UAV verification parameter are collected, and a corresponding multi-objective optimization function is constructed according to the verification result and the UAV verification parameter. It should be noted that after collecting the verification result (such as feature matching rate, semantic compliance score, state adaptation level) and the UAV verification parameter (such as real-time load, communication delay, historical execution error), the multi-objective optimization function is constructed, and specifically, these parameters directly reflect the constraint conditions and optimization direction of scheduling. In order to facilitate subsequent processing.

[0032] In step S40, the original call priority in the original task allocation scheme is dynamically adjusted by the multi-objective optimization function to generate a target call priority, and the scheduling of the multi-type UAV is completed according to the target call priority, with the goal of maximizing task completion efficiency, minimizing data conversion delay, and minimizing error rate.

[0033] It should be noted that finally, the goal is to maximize task completion efficiency (improve scheduling throughput), minimize data conversion delay (ensure real-time instruction), and minimize error rate (reduce communication risk), and the call priority of the original task allocation scheme is dynamically adjusted by the optimization function (such as assigning high emergency tasks to UAVs with good protocol adaptability and good state), to generate a target call priority and complete scheduling, to achieve optimal matching of resources and tasks. In order to facilitate subsequent processing.

[0034] Second embodiment Further, the step of creating a corresponding dynamic protocol feature map based on the unsupervised clustering algorithm includes: The core features are divided into several feature subsets according to the protocol type of the UAV, and a heterogeneous feature adaptation coding algorithm based on protocol attributes is used to map the core features of different feature subsets to a unified dimension feature space to generate a corresponding standard core feature set; The instruction format features in the standard core feature set are clustered in the first layer by using a hierarchical density clustering algorithm to obtain several basic feature clusters, and the cooperative triggering frequency of the core features inside each basic feature cluster is calculated, and the second layer density clustering is performed based on the cooperative triggering frequency to obtain several protocol feature clusters; An initial protocol feature map is constructed based on the several protocol feature clusters, and the initial protocol feature map is dynamically updated to generate the dynamic protocol feature map.

[0035] It should be noted that, first, the core features are divided into several feature subsets according to the unmanned aerial vehicle protocol type (such as MavLink, STANAG4586 standard protocol, or manufacturer-defined protocol), and specifically, the feature subsets of the same protocol type have commonality in instruction structure, data coding, etc., and the classification processing can reduce the subsequent coding difficulty. Due to the differences in feature dimensions and dimensions of different feature subsets (such as “instruction length” is a numerical type, and “instruction type” is a category type), a heterogeneous feature adaptation coding algorithm based on protocol attributes is adopted to map them to a unified dimension feature space (such as processing category features through one-hot encoding and processing numerical features through standardization), to generate a standard core feature set. Specifically, this step ensures that features of different protocols can be uniformly clustered and analyzed, avoiding clustering bias caused by differences in feature form.

[0036] A two-step clustering is performed using a hierarchical density clustering algorithm: the first layer is for clustering the “instruction format features” (such as instruction header identification, parameter field number, and check bit rule) in the standard core feature set, to obtain a basic feature cluster. Specifically, the instruction format is the core of protocol interaction, and unmanned aerial vehicles of the same format can directly reuse the instruction parsing logic. The second layer calculates the “cooperative triggering frequency” (such as the number of associated triggers of “take-off instruction” and “height feedback instruction”) of the core features within each basic feature cluster, and performs secondary clustering based on the frequency to obtain a protocol feature cluster. Specifically, the cooperative triggering frequency reflects the interaction logic commonality of the unmanned aerial vehicle protocol, and unmanned aerial vehicles of the same cluster have more compatibility in task instruction sequence execution.

[0037] An initial protocol feature map is constructed based on the protocol feature clusters (with clusters as nodes and protocol compatibility between clusters as edges), and a dynamic protocol feature map is generated through dynamic updating. Specifically, unmanned aerial vehicle protocols may be adjusted due to firmware upgrades and changes in task scenarios, and dynamic updating can ensure the timeliness of the map and provide accurate protocol basis for subsequent verification and scheduling.

[0038] Further, the step of constructing an initial protocol feature map based on a plurality of protocol feature clusters and dynamically updating the initial protocol feature map to generate the dynamic protocol feature map comprises: Extracting the functional semantic labels of the core features in each protocol feature cluster, and calculating the semantic similarity and triggering timing correlation between different protocol feature clusters based on the instruction interaction timing relationship in the multi-task scenario of the unmanned aerial vehicle, to correspondingly create the initial protocol feature map; Collecting newly generated protocol data corresponding to the execution of the task by the unmanned aerial vehicle, and performing local clustering on the new protocol data to generate a temporary feature cluster; Calculate the semantic compatibility and timing matching degree between the temporary feature cluster and each node in the initial protocol feature graph, and adjust the graph topology structure of the initial protocol feature graph based on the incremental update algorithm according to the semantic compatibility and the timing matching degree, to generate the dynamic protocol feature graph.

[0039] It should be noted that first, the "functional semantic label" (such as "navigation type", "task execution type", "state feedback type") of each protocol feature cluster core feature is extracted, combined with the instruction interaction timing relationship (such as the timing flow of "task allocation-confirmation reception-execution feedback") in the multi-task scene of the unmanned aerial vehicle, the "semantic similarity" (whether the functions overlap) and "trigger timing correlation" (whether the interaction logic matches) between different protocol feature clusters are calculated, and the initial protocol feature graph is constructed based on the two indexes. Specifically, the semantic similarity and timing correlation directly determine the feasibility of the cooperative scheduling of the unmanned aerial vehicle between clusters, and the graph can directly present this correlation.

[0040] During the execution of the task of the unmanned aerial vehicle, new protocol data (such as new instructions after firmware upgrade, new feedback fields after task extension) will be generated. These data are locally clustered (only for new data, to avoid the efficiency loss of re-clustering of full data) to generate temporary feature clusters. Specifically, local clustering can quickly capture protocol change characteristics and reduce update costs. The "semantic compatibility" (the degree of functional matching of new features and original features) and "timing matching degree" (the degree of fit of the interaction logic of new features and original clusters) of the temporary feature cluster and each node (original protocol feature cluster) in the initial graph are calculated. These two indexes determine the attribution of the temporary feature cluster: if the compatibility is high, it is merged into the node; if it is a new feature, a new node is added.

[0041] Based on the incremental update algorithm (only adjusting the changed part of the graph structure, not reconstructing the full graph), the topology structure of the initial graph is adjusted (such as adding a node, updating the association strength of the edge, and merging similar nodes) according to the semantic compatibility and the timing matching degree, to generate the dynamic protocol feature graph. Specifically, this updating method not only ensures the real-time performance of the graph, but also avoids the resource consumption caused by full update, ensuring that the subsequent scheduling can adapt to the dynamic changes of the unmanned aerial vehicle protocol. In order to facilitate subsequent processing.

[0042] Further, the step of constructing a corresponding multi-objective optimization function according to the verification result and the unmanned aerial vehicle verification parameter comprises: Extract the feature matching degree, semantic confidence and state adaptation rate in the verification result, calculate the corresponding coupling coefficient, and generate an adaptive optimization target; statistical fluctuation variance of the unmanned aerial vehicle verification parameter based on a sliding time window, adding an adaptive compensation term to the optimization target according to the fluctuation variance, and constructing a historical scheduling data feedback model to extract the correlation law of historical verification results and scheduling effects; According to the adaptive compensation term and the correlation law, the initial optimization function is adaptively adjusted to generate the multi-objective optimization function.

[0043] Need to be explained first, from the verification results, three core indicators are extracted: feature matching degree (the matching ratio of protocol features and task requirements, reflecting the adaptation basis), semantic confidence (the matching probability of instruction semantics and unmanned aerial vehicle function, reflecting the feasibility of the instruction), and state adaptation rate (the degree to which the real-time state of the unmanned aerial vehicle meets the task requirements, reflecting the execution conditions). The coupling coefficient of the three indicators (reflecting the correlation between the indicators, such as low feature matching degree reducing the actual utility of semantic confidence) is calculated by weighting to generate an adaptive optimization target. Specifically, these targets directly correspond to the core needs of scheduling, avoiding deviation of the optimization direction from the actual scene.

[0044] The unmanned aerial vehicle verification parameters (such as communication delay, load fluctuation, and power consumption rate) are dynamic, and the fluctuation variance (reflecting the stability of the parameters, the larger the variance, the more intense the parameter fluctuation) is calculated by using a sliding time window. According to the fluctuation variance, an adaptive compensation term is added to the optimization target. Specifically, for example, for unmanned aerial vehicles with large communication delay fluctuations, a compensation coefficient needs to be added in the "data conversion delay minimization" target to avoid invalidation of the optimization result due to parameter mutation. At the same time, a historical scheduling data feedback model is constructed to mine the correlation law of historical verification results (such as the relationship between the feature matching degree of a certain type of unmanned aerial vehicle and the scheduling success rate) and scheduling effects (such as task completion rate and error rate), providing an empirical basis for the optimization function and avoiding repeated past scheduling mistakes.

[0045] The adaptive compensation term (to cope with parameter dynamics) and the correlation law (to learn from historical experience) are integrated into the initial optimization function, and adaptive adjustment is completed by adjusting the target weight and supplementing the constraint conditions to generate the final multi-objective optimization function. Specifically, this function contains clear optimization direction and can adapt to the dynamic changes of unmanned aerial vehicle state and protocol, providing a scientific quantitative tool for priority adjustment.

[0046] Further, the step of adaptively adjusting the initial optimization function according to the adaptive compensation term and the correlation law to generate the multi-objective optimization function comprises: According to the nonlinear correlation coefficient in the correlation law, multiple target balancing effective samples are extracted from historical scheduling data, the contribution weight of each optimization target is quantified by a Gaussian kernel function, a corresponding weight distribution matrix is generated, and adaptive weights are assigned to the target terms of the initial optimization function. According to the fluctuation variance of the unmanned aerial vehicle verification parameter, different compensation coefficients are matched for different fluctuation levels, and the adaptive compensation term is converted into a targeted constraint condition; The adaptive weight, the differentiated compensation coefficient and the targeted constraint condition are embedded into the initial optimization function, and a non-dominated sorting genetic algorithm is used for multiple rounds of iterative solving to generate the multi-objective optimization function.

[0047] Need to be explained first, from the correlation law, the "nonlinear correlation coefficient" (reflecting the non-simple nonlinear relationship between the verification index and the scheduling effect, such as the success rate of scheduling will increase significantly after the feature matching degree exceeds 80%) is extracted, and based on the coefficient, the multi-objective balanced effective sample (i.e. high-quality scheduling cases that meet the efficiency, delay, and error rate targets) is selected from historical scheduling data. The contribution weight of each optimization target is quantified through Gaussian kernel function mapping (good at processing nonlinear data quantization). Specifically, for example, the contribution weight of "task completion efficiency" is higher than that of "data conversion delay" in emergency tasks, a weight distribution matrix is generated, and adaptive weights are given to the three target items of the initial optimization function to avoid insufficient scene adaptation caused by fixed weights.

[0048] According to the fluctuation variance of the unmanned aerial vehicle verification parameter, different compensation coefficients are matched for different fluctuation levels (such as low fluctuation: variance <5%; medium fluctuation: 5%≤variance≤15%; high fluctuation: variance>15%), and different compensation coefficients are matched for different fluctuation levels (high fluctuation level corresponds to higher compensation coefficient, which enhances the tolerance of the function to parameter mutation), and the adaptive compensation term is converted into a targeted constraint condition (such as "data conversion delay≤reference delay×(1+compensation coefficient)"), and the targeted constraint condition can avoid the optimization result exceeding the actual capacity of the unmanned aerial vehicle.

[0049] The adaptive weight (dynamically matches the scene demand), the differentiated compensation coefficient (adapts to the parameter fluctuation), and the targeted constraint condition (clearly defines the capability boundary) are embedded into the initial optimization function, and a non-dominated sorting genetic algorithm is used for multiple rounds of iterative solving. Specifically, this algorithm is good at handling multi-objective optimization problems, and by simulating the natural selection process, it finds the Pareto optimal solution that takes into account multiple objectives, and finally generates a multi-objective optimization function that meets the scheduling requirements and adapts to the actual constraints. To facilitate subsequent processing.

[0050] Further, the step of dynamically adjusting the original call priority in the original task allocation scheme through the multi-objective optimization function to generate a target call priority comprises: Based on the error rate distribution in the verification result and the task processing capability difference in the unmanned aerial vehicle verification parameter, the entropy weight-hierarchical analysis method is used to dynamically allocate the optimization weights of task completion efficiency, data conversion delay, and error rate; According to the optimization weight and the dynamic protocol feature map, a corresponding task-drone bidirectional adaptation matrix is constructed, and a comprehensive optimization value of each element in the task-drone bidirectional adaptation matrix is calculated through the multi-objective optimization function; The comprehensive optimization value is sorted to generate the target calling priority.

[0051] It should be noted that first, the entropy weight-AHP method is used to dynamically allocate the weights of the three optimization objectives according to the "analysis error rate distribution" in the verification result (for example, the analysis error rate of a certain type of drone for a specific task instruction is concentrated above 10%, which needs to reduce its priority) and the "task processing capability difference" in the drone verification parameter (for example, the task processing capability of a load-carrying drone is lower than that of a reconnaissance drone, which needs to be matched with the corresponding task). The entropy weight method objectively determines the weight through data distribution (for example, when the analysis error rate fluctuates greatly, its weight is automatically increased), and the AHP method combines expert experience for subjective correction (for example, the efficiency weight is artificially increased in an emergency task). The combination of the two ensures the objectivity and scene adaptability of the weight allocation.

[0052] According to the optimization weight and the dynamic protocol feature map (which provides the basis for protocol adaptation), a task-drone bidirectional adaptation matrix is constructed, specifically, the rows of the matrix represent tasks, the columns represent drones, and the elements are "task demand-drone capability" matching scores (comprehensive optimization target weight and protocol adaptation degree). The comprehensive optimization value of each element in the matrix is calculated through a multi-objective optimization function (quantifying the matching effect, the higher the value, the better the adaptability), and specifically, this value is the core basis for priority sorting, avoiding the subjectivity of priority allocation.

[0053] The comprehensive optimization value is sorted from high to low to generate the target calling priority, specifically, the "task-drone" combination with high priority will be executed first, ensuring that resources are tilted towards the best combination, improving the overall scheduling efficiency and success rate. For subsequent processing.

[0054] Further, the step of sorting the comprehensive optimization value to generate the target calling priority comprises: According to the updated features in the dynamic protocol feature map and the remaining load margin in the drone verification parameter, the comprehensive optimization value is time-weighted and corrected to obtain a corrected optimization value; Based on the corrected optimization value, the task urgency weight and the drone protocol adaptation stability score are fused to construct a hierarchical matching queue; Each element in the hierarchical matching queue is locally optimized and sorted to generate the target calling priority.

[0055] It should be noted that, first of all, the dynamic protocol feature map may generate new features due to the update of the unmanned aerial vehicle protocol, and the "remaining load margin" (reflecting the ability of the unmanned aerial vehicle to undertake new tasks) in the unmanned aerial vehicle verification parameter will also change dynamically with the task execution. In combination with these two dynamic factors, the time-effect weighting correction is performed on the comprehensive optimization value. Specifically, for example, the comprehensive optimization value of the unmanned aerial vehicle with high remaining load margin is multiplied by a correction coefficient of 1.1; the correction coefficient of the unmanned aerial vehicle with newly added high adaptation features in the map is increased to 1.2, and the corrected optimization value is obtained. Specifically, the corrected optimization value is more suitable for the real-time state of the unmanned aerial vehicle and the latest characteristics of the protocol, avoiding generating priorities based on outdated data.

[0056] Based on the corrected optimization value, the "task urgency weight" (such as the emergency rescue task urgency weight is 1.5, and the regular reconnaissance task is 1.0) and the "unmanned aerial vehicle protocol adaptation stability score" (such as the protocol adaptation score of a certain unmanned aerial vehicle for three consecutive tasks is all greater than or equal to 90 points, and the stability score is 1.1; if there is an adaptation fluctuation, the score is 0.9) are combined to construct a hierarchical matching queue: the first layer is the combination of high-urgency tasks and high-stability unmanned aerial vehicles, the second layer is the combination of regular-urgency tasks and medium-stability unmanned aerial vehicles, and the third layer is the combination of low-urgency tasks and basic-stability unmanned aerial vehicles. Specifically, the hierarchical matching ensures that high-priority tasks are matched with high-quality resources, avoiding resource waste.

[0057] The elements in each hierarchical matching queue are sorted locally (sorted by the corrected optimization value within the same layer), and the sorted results of each layer are integrated to generate a target call priority. Specifically, this "global hierarchical + local optimal" sorting method not only ensures the priority execution of emergency tasks, but also realizes the optimal allocation of resources within the same priority task, ensuring the efficiency and fairness of the dispatch.

[0058] Referring to Figure 2 The third embodiment of the present application provides: A multi-machine multi-task scheduling system for unmanned aerial vehicles, wherein the system comprises: A collection module configured to collect original protocol data of a plurality of types of unmanned aerial vehicles to be dispatched, and extract a plurality of core features contained in the original protocol data, so as to create a corresponding dynamic protocol feature map based on an unsupervised clustering algorithm according to the plurality of core features; A verification module configured to perform protocol feature consistency verification, instruction semantic rationality verification and unmanned aerial vehicle state feedback verification on the plurality of types of unmanned aerial vehicles simultaneously through the dynamic protocol feature map when receiving an original task allocation scheme; A construction module configured to collect verification results and unmanned aerial vehicle verification parameters, and construct a corresponding multi-objective optimization function according to the verification results and the unmanned aerial vehicle verification parameters; The calling module is configured to dynamically adjust original calling priorities in the original task allocation scheme by the multi-objective optimization function to generate target calling priorities, and complete scheduling of the multi-type unmanned aerial vehicle according to the target calling priorities, with the goal of maximizing task completion efficiency, minimizing data conversion delay and minimizing error rate.

[0059] Further, the collection module is specifically configured to: divide a plurality of the core features into a plurality of feature subsets according to protocol types of unmanned aerial vehicles, and adopt a heterogeneous feature adaptation coding algorithm based on protocol attributes to map the core features of different feature subsets to a feature space of a uniform dimension to generate a corresponding standard core feature set; adopt a hierarchical density clustering algorithm to perform first layer clustering on instruction format features in the standard core feature set to obtain a plurality of basic feature clusters, and calculate a cooperative triggering frequency of the core features inside each basic feature cluster, and perform second layer density clustering based on the cooperative triggering frequency to obtain a plurality of protocol feature clusters; construct an initial protocol feature map based on the plurality of protocol feature clusters, and perform dynamic updating processing on the initial protocol feature map to generate the dynamic protocol feature map.

[0060] Further, the collection module is specifically configured to: extract a functional semantic label of the core features in each protocol feature cluster, and calculate a semantic similarity and a triggering time sequence correlation degree between different protocol feature clusters based on an instruction interaction time sequence relationship in a multi-task scenario of unmanned aerial vehicles to correspondingly create the initial protocol feature map; collect newly added protocol data corresponding to the unmanned aerial vehicle generated in the process of executing a task, and perform local clustering on the newly added protocol data to generate a temporary feature cluster; calculate a semantic compatibility and a time sequence matching degree between the temporary feature cluster and each map node in the initial protocol feature map, and adjust a map topology structure of the initial protocol feature map based on an incremental updating algorithm according to the semantic compatibility and the time sequence matching degree to generate the dynamic protocol feature map.

[0061] Further, the construction module is specifically configured to: extract a feature matching degree, a semantic confidence and a state adaptation rate in the verification result, calculate a corresponding coupling coefficient, and generate an adapted optimization target; statistically calculate a fluctuation variance of the unmanned aerial vehicle inspection parameter based on a sliding time window, add an adaptive compensation term to the optimization target according to the fluctuation variance, and construct a historical scheduling data feedback model to extract an association rule of historical verification results and scheduling effects; The initial optimization function is adaptively adjusted according to the adaptive compensation term and the correlation law to generate the multi-target optimization function.

[0062] Further, the construction module is specifically used for: According to the nonlinear correlation coefficient in the correlation law, a plurality of target balance effective samples are extracted from historical scheduling data, a contribution weight of each optimization target is quantified by a Gaussian kernel function mapping, a corresponding weight distribution matrix is generated, and adaptive weights are assigned to target terms of the initial optimization function; According to the fluctuation variance grading of the unmanned aerial vehicle verification parameters, a differentiated compensation coefficient is matched for different fluctuation levels, and the adaptive compensation term is converted into a targeted constraint condition; The adaptive weights, the differentiated compensation coefficient, and the targeted constraint condition are embedded in the initial optimization function, a non-dominated sorting genetic algorithm is used for multi-round iteration solving to generate the multi-target optimization function.

[0063] Further, the calling module is specifically used for: Based on the distribution of the analysis error rate in the verification result and the difference in the task processing capability of the unmanned aerial vehicle verification parameters, an entropy weight-analytic hierarchy process is used to dynamically allocate optimization weights of task completion efficiency, data conversion delay, and analysis error rate; According to the optimization weights and the dynamic protocol feature map, a corresponding task-unmanned aerial vehicle bidirectional adaptation matrix is constructed, and a comprehensive optimization value of each element in the task-unmanned aerial vehicle bidirectional adaptation matrix is calculated through the multi-target optimization function; The comprehensive optimization values are sorted to generate the target calling priority.

[0064] Further, the calling module is specifically used for: In combination with the updated features in the dynamic protocol feature map and the remaining load margin in the unmanned aerial vehicle verification parameters, the comprehensive optimization values are time-effectively weighted and corrected to obtain modified optimization values; Based on the modified optimization values, a layered matching queue is constructed by fusing a task urgency weight and an unmanned aerial vehicle protocol adaptation stability score. Each element in the layered matching queue is locally optimally sorted to generate the target calling priority.

[0065] The fourth embodiment of the application provides a computer, including a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the unmanned aerial vehicle multi-machine multi-task scheduling method as described above when executing the computer program.

[0066] The fifth embodiment of the present application provides a readable storage medium, which stores a computer program, wherein the program is executed by a processor to realize the unmanned aerial vehicle multi-machine multi-task scheduling method as described above.

[0067] In summary, the unmanned aerial vehicle multi-machine multi-task scheduling method and system provided by the above embodiments can quickly and effectively complete the scheduling of multiple unmanned aerial vehicles, thereby improving the scheduling efficiency.

[0068] It should be noted that the above-mentioned modules can be functional modules or program modules, which can be implemented by software or hardware. For the modules implemented by hardware, the above-mentioned modules can be located in the same processor, or the above-mentioned modules can be located in different processors in any combination.

[0069] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered a list of executable instructions for implementing logic functions, and can be specifically embodied in any computer-readable medium for use by an instruction execution system, apparatus, or device, such as a computer-based system, a processor-based system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions, or in conjunction with the instruction execution system, apparatus, or device. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.

[0070] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection having one or more wires (electrical devices), a portable computer diskette (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium can even be paper or other suitable medium on which the program is printed, because the program can be electronically obtained, for example, by optical scanning of the paper or other medium, followed by electronic conversion, interpretation, or processing, or in other suitable manner, to be stored in the computer memory.

[0071] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the embodiments described above, various steps or methods can be implemented, for example, in software or firmware that is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following techniques can be used to implement the hardware used to implement the described functions: discrete logic circuitry having logic gates for implementing logic functions upon data signals, application specific integrated circuits having logic gates for implementing the logic functions on data signals, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0072] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0073] The above-described embodiments only express several implementation manners of the present application, which are described in a more specific and detailed manner, but cannot be understood as limiting the scope of the present application. It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present application, which are all within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for multi-drone multi-task scheduling, characterized in that, The method comprises: Collecting native protocol data of multiple types of unmanned aerial vehicles to be dispatched, and extracting a plurality of core features contained in the native protocol data, to create a corresponding dynamic protocol feature map based on an unsupervised clustering algorithm according to the plurality of core features; When receiving an original task allocation scheme, performing protocol feature consistency verification, instruction semantic rationality verification and unmanned aerial vehicle state feedback verification on the multiple types of unmanned aerial vehicles simultaneously through the dynamic protocol feature map; Collecting verification results and unmanned aerial vehicle verification parameters, and constructing a corresponding multi-objective optimization function according to the verification results and the unmanned aerial vehicle verification parameters; With the goal of maximizing task completion efficiency, minimizing data conversion delay and minimizing analysis error rate, dynamically adjusting the original call priority in the original task allocation scheme through the multi-objective optimization function to generate a target call priority, and completing the dispatch of the multiple types of unmanned aerial vehicles according to the target call priority.

2. The method of claim 1, wherein, The step of creating a corresponding dynamic protocol feature map based on an unsupervised clustering algorithm according to a plurality of core features comprises: Divide a plurality of core features into a plurality of feature subsets according to the protocol type of the unmanned aerial vehicle, and use a heterogeneous feature adaptation coding algorithm based on protocol attributes to map the core features of different feature subsets to a unified dimension feature space to generate a corresponding standard core feature set; Use a hierarchical density clustering algorithm to perform first layer clustering on the instruction format features in the standard core feature set to obtain a plurality of basic feature clusters, and calculate the cooperative trigger frequency of the core features inside each basic feature cluster, and perform second layer density clustering based on the cooperative trigger frequency to obtain a plurality of protocol feature clusters; Based on a plurality of protocol feature clusters, an initial protocol feature map is constructed, and the initial protocol feature map is dynamically updated to generate the dynamic protocol feature map.

3. The method of claim 2, wherein, The step of constructing an initial protocol feature map based on a plurality of protocol feature clusters, and dynamically updating the initial protocol feature map to generate the dynamic protocol feature map comprises: Extract the functional semantic labels of the core features in each protocol feature cluster, and calculate the semantic similarity and trigger timing correlation between different protocol feature clusters based on the instruction interaction timing relationship in the unmanned aerial vehicle multi-task scenario, to correspondingly create the initial protocol feature map; Collecting new protocol data corresponding to the process of the unmanned aerial vehicle executing tasks, and locally clustering the new protocol data to generate temporary feature clusters; Calculate the semantic compatibility and timing matching degree between the temporary feature clusters and each graph node in the initial protocol feature map, and adjust the graph topology structure of the initial protocol feature map based on an incremental update algorithm according to the semantic compatibility and the timing matching degree to generate the dynamic protocol feature map.

4. The method of claim 1, wherein, The step of constructing a corresponding multi-objective optimization function according to the verification results and the unmanned aerial vehicle verification parameters comprises: Extract the feature matching degree, semantic confidence and state adaptation rate in the verification results, calculate the corresponding coupling coefficient, and generate an adaptive optimization target; statistical fluctuation variance of the unmanned aerial vehicle inspection parameters based on a sliding time window, adding an adaptive compensation term to the optimization target according to the fluctuation variance, and constructing a historical scheduling data feedback model to extract the correlation between historical calibration results and scheduling effects; adaptively adjusting an initial optimization function according to the adaptive compensation term and the correlation to generate the multi-objective optimization function.

5. The method of claim 4, wherein, The step of adaptively adjusting an initial optimization function according to the adaptive compensation term and the correlation to generate the multi-objective optimization function includes: extracting multi-objective balance effective samples from historical scheduling data according to nonlinear correlation coefficients in the correlation, mapping and quantifying the contribution weight of each optimization target by a Gaussian kernel function, generating a corresponding weight distribution matrix, and giving the target term of the initial optimization function an adaptive weight; According to the fluctuation variance of the unmanned aerial vehicle calibration parameters, the differential compensation coefficients are matched for different fluctuation levels, and the adaptive compensation term is converted into a targeted constraint condition; embed the adaptive weight, the differential compensation coefficient and the targeted constraint condition into the initial optimization function, and use a non-dominated sorting genetic algorithm for multi-round iteration to generate the multi-objective optimization function.

6. The method of claim 1, wherein, The step of dynamically adjusting the original call priority in the original task allocation scheme by the multi-objective optimization function to generate a target call priority includes: Based on the distribution of the analysis error rate in the calibration result and the difference in the task processing capability of the unmanned aerial vehicle calibration parameters, the optimization weights of the task completion efficiency, the data conversion delay and the analysis error rate are dynamically allocated by using the entropy weight-hierarchical analysis method; According to the optimization weight and the dynamic protocol feature map, a corresponding task-unmanned aerial vehicle bidirectional adaptation matrix is constructed, and a comprehensive optimization value of each element in the task-unmanned aerial vehicle bidirectional adaptation matrix is calculated by the multi-objective optimization function; sort the comprehensive optimization values to generate the target call priority.

7. The method of claim 6, wherein, The step of sorting the comprehensive optimization values to generate the target call priority includes: Combine the update features in the dynamic protocol feature map and the remaining load margin in the unmanned aerial vehicle calibration parameters to time-weight correct the comprehensive optimization values to obtain a corrected optimization value; Based on the corrected optimization value, the task urgency weight and the unmanned aerial vehicle protocol adaptation stability score are fused to construct a hierarchical matching queue; perform local optimal sorting on each element in the hierarchical matching queue to generate the target call priority.

8. A multi-drone multi-task scheduling system for unmanned aerial vehicles, characterized in that, The system includes: The acquisition module is configured to acquire original protocol data of a plurality of types of unmanned aerial vehicles to be scheduled, and extract a plurality of core features contained in the original protocol data, and create a corresponding dynamic protocol feature map based on an unsupervised clustering algorithm according to the plurality of core features; The calibration module is configured to perform protocol feature consistency calibration, instruction semantic rationality calibration and unmanned aerial vehicle state feedback calibration on the plurality of types of unmanned aerial vehicles simultaneously when receiving an original task allocation scheme. A construction module is configured to collect a verification result and a UAV verification parameter, and construct a corresponding multi-objective optimization function according to the verification result and the UAV verification parameter; A calling module is configured to maximize task completion efficiency, minimize data conversion delay, and minimize error rate, dynamically adjust an original calling priority in the original task allocation scheme through the multi-objective optimization function to generate a target calling priority, and complete scheduling of the multi-type UAVs according to the target calling priority.

9. A computer comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the UAV multi-aircraft multi-task scheduling method of any one of claims 1 to 7.

10. A readable storage medium, having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the UAV multi-aircraft multi-task scheduling method of any one of claims 1 to 7.

Images