Unmanned aerial vehicle autonomy assessment method, system and electronic equipment based on knowledge graph
Through the drone autonomy evaluation method based on knowledge graph, a multi-level evaluation system is built, and the autonomous score is calculated using training samples and task execution logs, the problem of subjectivity of drone autonomy evaluation is solved, and the detection accuracy and Performance optimization effect.
Patent Information
- Application Number
- CN202311735004.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-15
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2043-12-15
AI Technical Summary
In the prior art, the drone autonomy evaluation is highly subjective, which leads to a large deviation from the actual situation, making it difficult to achieve quantitative evaluation and accurate detection.
Adopting the drone autonomy evaluation method based on knowledge graph, by constructing a multi-level autonomy index evaluation system, using training samples to calculate node weights and index values, and combining with task execution logs to update the drone autonomy scores. Quantitative evaluation and optimization.
The accuracy of drone performance detection is improved, the quantitative evaluation and performance optimization of drone autonomy is realized, and subjective deviation is reduced.
Smart Images

Figure CN117688838B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of drone technology, and in particular to a knowledge graph-based drone autonomy assessment method, system, and electronic equipment. Background Art
[0002] The performance of a drone includes flight performance, mission payload, environmental adaptability, communication distance, positioning accuracy, concealment and autonomy. Among them, drone autonomy refers to the ability of a drone to make its own decisions and perform tasks without direct intervention from humans or other entities.
[0003] In the related art, users usually rate the autonomy of drones, and this rating mainly depends on the user's qualitative evaluation of the drone's mission completion. Since the drone autonomy scores given by users are highly subjective, the drone autonomy scores deviate greatly from the actual situation. Therefore, how to quantitatively evaluate the autonomy of drones and improve the accuracy of drone performance testing is a technical problem that technicians in this field currently need to solve. Summary of the invention
[0004] The purpose of this application is to provide a UAV autonomy evaluation method based on a knowledge graph, a UAV autonomy evaluation system based on a knowledge graph, and an electronic device, which can quantitatively evaluate the autonomy of the UAV and improve the accuracy of UAV performance detection.
[0005] In order to solve the above technical problems, the present application provides a method for evaluating the autonomy of a UAV based on a knowledge graph, which includes:
[0006] Constructing an autonomy index evaluation system for the UAV; wherein the autonomy index evaluation system includes multiple levels of nodes, the highest level node in the autonomy index evaluation system is the autonomous capability of the UAV, and the lowest level node in the autonomy index evaluation system is the basic factor of autonomy;
[0007] Acquire training samples from the data set; wherein the training samples include sample weights and sample index values of the basic factors of autonomy;
[0008] Based on the knowledge graph, the weight and index value of each node in the autonomy index evaluation system are calculated layer by layer using the sample weight and the sample index value, and the index value corresponding to the highest-level node in the autonomy index evaluation system is set as the autonomy prediction score of the drone;
[0009] Calculating a loss function value according to the autonomy prediction score and the autonomy reference score corresponding to the training sample, so as to iteratively update the knowledge graph using the loss function value;
[0010] If a performance test request is received, determining a corresponding target task according to the performance test request;
[0011] Allocating a corresponding actual weight to each of the autonomous basic factors according to the model of the drone and the task content of the target task; wherein the sum of the actual weights of all the autonomous basic factors belonging to the same parent node is 1;
[0012] Controlling the UAV to execute the target task to obtain a task execution log, and determining actual index values of all the basic factors of autonomy according to the task execution log;
[0013] The weight and index value of each node in the autonomy index evaluation system are calculated layer by layer based on the knowledge graph using the actual weight and the actual index value, and the index value corresponding to the highest-level node in the autonomy index evaluation system is set as the actual autonomy score of the UAV, so as to optimize the performance of the UAV according to the actual autonomy score.
[0014] Optionally, build an autonomy indicator evaluation system for drones, including:
[0015] Determining a plurality of corresponding autonomy evaluation dimensions according to the model of the drone;
[0016] Selecting a current autonomy evaluation dimension from all the autonomy evaluation dimensions;
[0017] Querying multiple influencing factors of the current autonomy evaluation dimension; wherein the influencing factors include direct influencing factors and indirect influencing factors;
[0018] The organizational structure chart corresponding to the current autonomy evaluation dimension is constructed based on the correlation between all the influencing factors;
[0019] Determining whether the organizational structure chart corresponding to all the autonomy evaluation dimensions is obtained;
[0020] If so, constructing the autonomy indicator evaluation system according to the organizational structure chart corresponding to all the autonomy evaluation dimensions;
[0021] If not, the process proceeds to the step of selecting the current autonomy evaluation dimension from all the autonomy evaluation dimensions.
[0022] Optionally, constructing the autonomy indicator evaluation system according to the organizational structure chart corresponding to all the autonomy evaluation dimensions includes:
[0023] Integrate the organizational structure charts corresponding to all the autonomy evaluation dimensions to obtain an overall organizational structure chart;
[0024] Determine whether there is a dispute node in the overall organizational structure diagram; wherein the dispute node is a single child node in the overall organizational structure diagram that has at least two parent nodes;
[0025] If yes, then set the parent node with the highest correlation with the dispute node as the true parent node, and set all parent nodes of the dispute node except the true parent node as false parent nodes;
[0026] The parent-child node relationship between the false parent node and the dispute node in the overall organizational structure diagram is removed to obtain the autonomy indicator evaluation system.
[0027] Optionally, determining actual index values of all the basic factors of autonomy according to the task execution log includes:
[0028] Obtaining the indicator value calculation rules corresponding to each of the autonomy basic factors in the autonomy indicator evaluation system;
[0029] querying the task execution log for log content matching each of the autonomy basic factors;
[0030] The index value calculation rule is used to calculate the log content corresponding to the same autonomy basic factor to obtain the actual index values corresponding to all the autonomy basic factors.
[0031] Optionally, assigning a corresponding actual weight to each of the autonomy basic factors according to the model of the drone and the task content of the target task includes:
[0032] Constructing a weight query table according to all the autonomy basic factors in the autonomy index evaluation system, and adding the model of the drone and the task content of the target task to the weight query table;
[0033] Sending the weight query table to a terminal device so that the terminal device returns weight configuration information; wherein the weight configuration information includes a correspondence between the autonomy basic factors and the actual weights;
[0034] A corresponding actual weight is allocated to each of the autonomous basic factors according to the weight configuration information.
[0035] Optionally, allocating a corresponding actual weight to each of the autonomous basic factors according to the weight configuration information includes:
[0036] Input the weight configuration information into a weight prediction model, and assign a corresponding actual weight to each of the autonomy basic factors according to the weight prediction result of the weight prediction model; wherein the weight prediction model is a model trained based on a deep learning framework, and the weight prediction model includes an input layer, a hidden layer, and an output layer, wherein the hidden layer is used to extract and map features of the weight configuration information transmitted from the input layer, and the output layer is used to convert the processing result of the hidden layer into the weight prediction result;
[0037] Accordingly, after setting the index value corresponding to the highest level node in the autonomy index evaluation system as the actual autonomy score of the drone, the method further includes:
[0038] Determining whether the actual autonomy score is within a preset score range;
[0039] If not, the parameters of the weight prediction model are updated according to the degree of deviation between the actual autonomy score and the preset score range, so as to use the updated weight prediction model to re-assign the corresponding actual weight to each of the basic autonomy factors.
[0040] Optionally, sending the weight query table to a terminal device includes:
[0041] Sending the weight query table to the plurality of terminal devices;
[0042] Accordingly, allocating a corresponding actual weight to each of the autonomous basic factors according to the weight configuration information includes:
[0043] All the weight configuration information is weighted according to the confidence of each terminal device, and a corresponding actual weight is allocated to each of the autonomous basic factors according to the weighted calculation result.
[0044] Optionally, after the terminal device returns the weight configuration information, the method further includes:
[0045] Determining whether the weight configuration information includes actual weights of all the autonomous basic factors;
[0046] If not, the autonomous basic factors that are not configured with actual weights in the weight configuration information are set as irrelevant factors, and the nodes corresponding to the irrelevant factors in the autonomous indicator evaluation system are removed to update the autonomous indicator evaluation system.
[0047] This application also provides a UAV autonomy assessment system based on knowledge graph, which includes:
[0048] An evaluation system construction module is used to construct an autonomy index evaluation system for the UAV; wherein the autonomy index evaluation system includes multiple levels of nodes, the highest level node in the autonomy index evaluation system is the autonomous capability of the UAV, and the lowest level node in the autonomy index evaluation system is the basic factor of autonomy;
[0049] A training module, used to obtain training samples from a data set; wherein the training samples include sample weights and sample index values of the autonomy basic factors; further used to calculate the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph using the sample weights and the sample index values, and set the index value corresponding to the highest-level node in the autonomy index evaluation system as the autonomy prediction score of the drone; further used to calculate a loss function value based on the autonomy prediction score and the autonomy reference score corresponding to the training samples, so as to iteratively update the knowledge graph using the loss function value;
[0050] A task determination module, configured to determine a corresponding target task according to a performance detection request if a performance detection request is received;
[0051] A weight allocation module, used to allocate a corresponding actual weight to each of the autonomous basic factors according to the model of the drone and the task content of the target task; wherein the sum of the actual weights of all the autonomous basic factors belonging to the same parent node is 1;
[0052] an index value determination module, used for controlling the UAV to execute the target task to obtain a task execution log, and determining the actual index values of all the autonomous basic factors according to the task execution log;
[0053] The autonomy evaluation module is used to calculate the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph using the actual weight and the actual index value, and set the index value corresponding to the highest-level node in the autonomy index evaluation system as the actual autonomy score of the drone, so as to optimize the performance of the drone according to the actual autonomy score.
[0054] The present application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, the steps of executing the above-mentioned knowledge graph-based drone autonomy assessment method are implemented.
[0055] The present application provides a method for evaluating the autonomy of a drone based on a knowledge graph, comprising: constructing an autonomy index evaluation system for the drone; wherein the autonomy index evaluation system comprises nodes at multiple levels, the highest-level node in the autonomy index evaluation system is the autonomous capability of the drone, and the lowest-level node in the autonomy index evaluation system is the basic factor of autonomy; obtaining training samples from a data set; wherein the training samples comprise sample weights and sample index values of the basic factor of autonomy; calculating the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph using the sample weights and the sample index values, and setting the index value corresponding to the highest-level node in the autonomy index evaluation system as the autonomy prediction score of the drone; calculating the loss function value according to the autonomy prediction score and the autonomy reference score corresponding to the training sample, so as to utilize the The loss function value is iteratively updated in the knowledge graph; if a performance test request is received, the corresponding target task is determined according to the performance test request; the corresponding actual weight is assigned to each of the autonomous basic factors according to the model of the UAV and the task content of the target task; wherein the sum of the actual weights of all the autonomous basic factors belonging to the same parent node is 1; the UAV is controlled to execute the target task to obtain a task execution log, and the actual index values of all the autonomous basic factors are determined according to the task execution log; based on the knowledge graph, the weight and index value of each node in the autonomous index evaluation system are calculated layer by layer using the actual weight and the actual index value, and the index value corresponding to the highest-level node in the autonomous index evaluation system is set as the actual autonomy score of the UAV, so as to optimize the performance of the UAV according to the actual autonomy score.
[0056] The present application constructs an autonomy index rating system for drones. There are multiple levels of nodes in the autonomy index rating system. The highest-level node in the autonomy index rating system is the autonomous capability of the drone, and the remaining nodes are factors that directly or indirectly affect the autonomous capability of the drone. The most basic factor affecting the autonomous capability of the drone in the autonomy index rating system is the basic autonomy factor corresponding to the lowest-level node. The present application obtains training samples from a data set, and uses the training samples to calculate the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph, and finally obtains the autonomy prediction score of the drone. The present application calculates the loss function value based on the autonomy prediction score of the drone and the autonomy reference score corresponding to the training sample, and then uses the loss function value to iteratively update the knowledge in the knowledge graph, so that the weight relationship and index value relationship between adjacent level nodes in the autonomy index evaluation system are correctly recorded in the knowledge graph. After receiving the performance test request, the present application assigns a corresponding actual weight to each of the autonomous basic factors according to the model of the drone and the task content of the target task, determines the actual index values of all autonomous basic factors according to the task execution log when the drone performs the target task, and calculates the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph using the actual weight and the actual index value, thereby obtaining the actual autonomy score of the drone, so as to optimize the performance of the drone according to the actual autonomy score. The present application determines the actual autonomy score of the drone based on the autonomous index rating system of the drone and the knowledge graph obtained by iterative training, realizes the quantitative evaluation of the autonomy of the drone, and improves the accuracy of the drone performance test. The present application also provides a drone autonomy evaluation system and an electronic device based on a knowledge graph, which have the above-mentioned beneficial effects and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS
[0057] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0058] Figure 1 A flowchart of a method for evaluating the autonomy of a drone based on a knowledge graph provided in an embodiment of the present application;
[0059] Figure 2 A schematic diagram of an autonomous indicator evaluation system provided in an embodiment of the present application;
[0060] Figure 3 A flowchart of a method for constructing an autonomous indicator evaluation system provided in an embodiment of the present application;
[0061] Figure 4 A schematic diagram of the structure of an autonomous indicator evaluation system in a practical application provided by an embodiment of the present application;
[0062] Figure 5 A schematic diagram of the structure of a knowledge graph-based drone autonomy assessment system provided in an embodiment of the present application. DETAILED DESCRIPTION
[0063] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0064] See below Figure 1 , Figure 1 A flowchart of a method for evaluating the autonomy of a drone based on a knowledge graph provided in an embodiment of the present application.
[0065] Specific steps may include:
[0066] S101: Build an autonomy indicator evaluation system for drones;
[0067] Among them, this embodiment can be applied to electronic devices equipped with a knowledge graph to obtain the model of the drone whose autonomy is to be evaluated, and construct a corresponding autonomy index evaluation system based on the correspondence between the drone model and the factors affecting the autonomy.
[0068] The autonomy index evaluation system includes multiple levels of nodes, the highest level node in the autonomy index evaluation system is the autonomous capability of the drone, and the lowest level node in the autonomy index evaluation system is the basic factor of autonomy. Specifically, there are multiple levels in the autonomy index evaluation system, and each level has a corresponding number of nodes. The number of the highest level node in the autonomy index evaluation system is 1. In the autonomy index evaluation system, the higher the level, the fewer nodes it contains. The lowest level node in the autonomy index evaluation system is the basic factor of autonomy, that is, the smallest factor affecting the autonomy evaluation.
[0069] The autonomy index evaluation system is a tree diagram for evaluating the factors affecting the autonomy of drones. Figure 2 , Figure 2This is a schematic diagram of an autonomy index evaluation system provided in an embodiment of the present application, where P1 represents the highest-level node drone autonomy, the nodes corresponding to P21, P22, and P23 represent factors that directly affect the drone autonomy evaluation, and the remaining nodes are factors that indirectly affect the drone autonomy evaluation. The nodes corresponding to the lowest-level nodes Pn1, Pn2, Pn3..., Pnm are basic factors of autonomy.
[0070] S102: Obtain training samples from the data set;
[0071] Among them, this embodiment can be pre-set with a data set, in which there are index values corresponding to the basic factors of autonomy recorded when the drone performs historical tasks, that is, sample index values; the data set also contains weights of the basic factors of autonomy used to evaluate the autonomy scores of the drone when performing historical tasks, that is, sample weights. The training samples obtained from the data set include the sample weights and sample index values of the basic factors of autonomy. The above data set can also include a reference score of autonomy when the drone performs historical tasks, and the reference score of autonomy can be an autonomy score calculated by combining the analytic hierarchy process and fuzzy comprehensive evaluation.
[0072] S103: Calculate the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph using the sample weight and the sample index value, and set the index value corresponding to the highest-level node in the autonomy index evaluation system as the autonomy prediction score of the drone;
[0073] On the basis of obtaining the training samples, this embodiment can obtain a knowledge graph, in which the weight calculation relationship and the indicator value calculation relationship of adjacent level nodes in the autonomy indicator evaluation system are recorded.
[0074] Taking the highest level as level 1 and the number of levels of the autonomous index evaluation system increasing from top to bottom as an example, the knowledge graph records the knowledge of calculating the weight and index value of the i-1th level node according to the weight and index value of the i-th level node. This embodiment can calculate the weight and index value of each node in the autonomous index evaluation system layer by layer based on the knowledge graph using the sample weight and the sample index value.
[0075] Specifically, this embodiment can calculate the weight and index value of the n-1th level node based on the knowledge graph using the weights and index values of the basic factors of autonomy of the n-1th level node (i.e., the lowest level node); this embodiment can also calculate the weight and index value of the n-2th level node based on the knowledge graph using the weights and index values of the n-1th level node, and so on, to achieve the calculation of the weight and index value of each node in the autonomy index evaluation system layer by layer from bottom to top based on the knowledge graph.
[0076] After obtaining the weights and index values of all nodes, this embodiment can set the index value corresponding to the highest-level node in the autonomy index evaluation system as the autonomy prediction score of the drone, that is, the autonomy score of the drone predicted using the autonomy index evaluation system and the knowledge graph.
[0077] S104: Calculating a loss function value according to the autonomy prediction score and the autonomy reference score corresponding to the training sample, so as to iteratively update the knowledge graph using the loss function value;
[0078] After obtaining the autonomy prediction score, the present embodiment can calculate the loss function value according to the autonomy prediction score (i.e., the predicted value) and the autonomy reference score (actual value) corresponding to the training sample, and update the parameters of the knowledge graph based on the loss function value, so that the autonomy prediction score calculated based on the knowledge graph tends to be consistent with the autonomy reference score.
[0079] It should be noted that there may be multiple training samples in the data set, and the present embodiment may repeatedly perform the operation steps of S102 to S104 to implement iterative update of the knowledge graph. When the number of iterations or the loss function value meets the preset conditions, it is determined that the knowledge graph training is completed, and the related operations of S105 to S108 may be performed.
[0080] Specifically, after comparing and analyzing the autonomy prediction score with the autonomy reference score to obtain the loss function value, the sample weights and sample index values of the autonomy basic factors obtained from the knowledge graph data set and the autonomy reference score of the drone's historical missions can be updated. Update the relative weight values of the factors within each level, that is, more accurately determine the relative importance of the factors within the level to obtain more accurate index values; update the relative weight values of the lower-level indicators to the upper-level indicators, that is, clarify the relative importance of the lower-level indicators to the upper-level indicators to obtain more accurate index values. The loss function obtained through the autonomy prediction score and the autonomy reference score forms a closed-loop feedback on the sample weights and sample index values in the knowledge graph, and continuously iterates and updates to form the optimal quantitative evaluation calculation method.
[0081] S105: If a performance test request is received, determining a corresponding target task according to the performance test request;
[0082] After receiving the performance detection request, the performance detection request can be parsed to determine the corresponding target task, which is the task that the drone needs to perform.
[0083] S106: assigning a corresponding actual weight to each of the autonomy basic factors according to the model of the drone and the task content of the target task;
[0084] This embodiment can pre-set a correspondence table of drone models, mission content, and weights of autonomous basic factors. After obtaining the drone model and the mission content of the target mission, this embodiment can query the corresponding weight based on the correspondence table, that is, the actual weight corresponding to each of the autonomous basic factors. The above process needs to allocate the actual weight according to the constraint condition, and the above constraint condition is: the sum of the actual weights of all the autonomous basic factors belonging to the same parent node is 1.
[0085] S107: Controlling the UAV to execute the target task to obtain a task execution log, and determining actual index values of all the autonomy basic factors according to the task execution log;
[0086] After determining the target task corresponding to the performance detection request, the UAV can be controlled to perform the target task, and the task execution log during the execution of the target task can be obtained. The task execution log may include the UAV's operating parameters, environmental information, and user-defined information. Based on the above tasks, the actual index values of all the basic factors of autonomy can be determined.
[0087] As a feasible implementation method, the actual index value of the basic factor of autonomy can be determined in the following way: obtaining the index value calculation rules corresponding to each of the basic factors of autonomy in the autonomy index evaluation system; querying the log content matching each of the basic factors of autonomy in the task execution log; using the index value calculation rules to calculate the log content corresponding to the same basic factor of autonomy, and obtaining the actual index values corresponding to all the basic factors of autonomy.
[0088] S108: Based on the knowledge graph, the weight and index value of each node in the autonomy index evaluation system are calculated layer by layer using the actual weight and the actual index value, and the index value corresponding to the highest-level node in the autonomy index evaluation system is set as the actual autonomy score of the UAV, so as to optimize the performance of the UAV according to the actual autonomy score.
[0089] Among them, after obtaining the actual weight and actual index value corresponding to the target task performed by the drone, the weight and index value of each node in the autonomy index evaluation system can be calculated layer by layer based on the trained knowledge graph using the actual weight and the actual index value, and then the index value corresponding to the highest-level node in the autonomy index evaluation system is set as the actual autonomy score of the drone. The actual autonomy score describes the ability of the drone to self-determine and perform tasks when performing the target task. After obtaining the actual autonomy score, this embodiment can optimize the performance of the drone according to the actual autonomy score. This embodiment can pre-set the corresponding relationship between the autonomy score interval and the performance optimization strategy. This embodiment can determine the target performance optimization strategy in the autonomy score interval where the actual autonomy score is located, and then perform corresponding operations according to the target performance optimization strategy. Specifically, this embodiment can optimize the software of the drone according to the actual autonomy score, such as adjusting the control strategy; this embodiment can also optimize the hardware of the drone according to the actual autonomy score, such as installing auxiliary flight equipment.
[0090] This embodiment constructs an autonomy index rating system for drones. There are multiple levels of nodes in the autonomy index rating system. The highest-level node in the autonomy index rating system is the autonomous capability of the drone, and the remaining nodes are factors that directly or indirectly affect the autonomous capability of the drone. The most basic factor affecting the autonomous capability of the drone in the autonomy index rating system is the basic autonomy factor corresponding to the lowest-level node. This embodiment obtains training samples from a data set, and uses the training samples to calculate the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph, and finally obtains the autonomy prediction score of the drone. This embodiment calculates the loss function value based on the autonomy prediction score of the drone and the autonomy reference score corresponding to the training sample, and then uses the loss function value to iteratively update the knowledge in the knowledge graph, so that the weight relationship and index value relationship between adjacent level nodes in the autonomy index evaluation system are correctly recorded in the knowledge graph. After receiving the performance test request, this embodiment assigns a corresponding actual weight to each of the autonomous basic factors according to the model of the drone and the task content of the target task, determines the actual index values of all autonomous basic factors according to the task execution log when the drone performs the target task, and calculates the weight and index value of each node in the autonomous index evaluation system layer by layer based on the knowledge graph using the actual weight and the actual index value, thereby obtaining the actual autonomy score of the drone, so as to optimize the performance of the drone according to the actual autonomy score. This embodiment determines the actual autonomy score of the drone based on the autonomous index rating system of the drone and the knowledge graph obtained by iterative training, realizes the quantitative evaluation of the autonomy of the drone, and improves the accuracy of the performance test of the drone.
[0091] See also Figure 3 , Figure 3 A flowchart of a method for constructing an autonomous indicator evaluation system provided in an embodiment of the present application specifically includes the following steps:
[0092] S301: Determine a plurality of corresponding autonomy evaluation dimensions according to the model of the drone.
[0093] Among them, this embodiment can pre-set the correspondence between the drone's signal and the autonomy evaluation dimension, and then determine multiple autonomy evaluation dimensions to be selected in this embodiment based on the correspondence. The above autonomy evaluation dimensions can include task complexity dimension, environmental complexity dimension, human-machine interaction degree dimension, motor working parameter change dimension, etc.
[0094] S302: Selecting a current autonomy evaluation dimension from all the autonomy evaluation dimensions.
[0095] Each time S302 is executed, an autonomy evaluation dimension for which an organizational chart has not yet been generated may be selected from all autonomy evaluation dimensions.
[0096] S303: Query multiple influencing factors of the current autonomy evaluation dimension;
[0097] Among them, this embodiment can obtain the knowledge graph corresponding to the current autonomy evaluation dimension, and query multiple influencing factors of the current autonomy evaluation dimension from the knowledge graph, and the above-mentioned influencing factors include direct influencing factors and indirect influencing factors.
[0098] S304: constructing an organizational structure chart corresponding to the current autonomy evaluation dimension according to the correlation relationship between all the influencing factors;
[0099] In this embodiment, an organizational structure chart corresponding to the current autonomy evaluation dimension can be constructed according to the correlation between the influencing factors, and each influencing factor corresponds to a node in the organizational structure chart.
[0100] S305: Determine whether the organizational structure charts corresponding to all autonomy evaluation dimensions are obtained; if so, proceed to S306; if not, proceed to S302;
[0101] S306: Constructing the autonomy indicator evaluation system according to the organizational structure chart corresponding to all the autonomy evaluation dimensions.
[0102] After obtaining the organizational chart corresponding to all autonomy evaluation dimensions, the highest-level nodes of all organizational charts can be hung under the nodes corresponding to the autonomous capabilities of the drone to obtain an autonomy indicator evaluation system.
[0103] As a further introduction to the above embodiment, if there are multiple parent nodes corresponding to the same child node, the same factor will be calculated repeatedly during the autonomy evaluation, affecting the accuracy of the autonomy evaluation. In view of the above problem, this embodiment can construct the autonomy index evaluation system in the following manner to ensure that a child node has only one parent node. The specific process includes:
[0104] Integrate the organizational charts corresponding to all the autonomy evaluation dimensions to obtain an overall organizational chart; determine whether there is a dispute node in the overall organizational chart; wherein the dispute node is a single child node with at least two parent nodes in the overall organizational chart; if there is a dispute node, set the parent node with the highest degree of association with the dispute node as a true parent node, and set all parent nodes of the dispute node except the true parent node as false parent nodes; remove the parent-child node relationship between the false parent node and the dispute node in the overall organizational chart to obtain the autonomy indicator evaluation system.
[0105] As for Figure 1 Further introduction to the corresponding embodiment, this embodiment can allocate corresponding actual weights to the autonomy basic factors in the following manner:
[0106] A weight query table is constructed according to all the autonomous basic factors in the autonomous index evaluation system, and the model of the drone and the task content of the target task are added to the weight query table; the weight query table is sent to the terminal device so that the terminal device returns the weight configuration information; and the corresponding actual weight is assigned to each of the autonomous basic factors according to the weight configuration information. The above-mentioned weight query table may include all autonomous basic factors, the model of the drone and the target task, so that the terminal device can set the weight corresponding to each autonomous basic factor based on the above and obtain the weight configuration information. The above-mentioned weight configuration information includes the corresponding relationship between the autonomous basic factors and the actual weights.
[0107] As a feasible implementation method, this embodiment can input the weight configuration information into a weight prediction model, and assign a corresponding actual weight to each of the autonomous basic factors according to the weight prediction results of the weight prediction model. The weight prediction model is a model trained based on a deep learning framework, and the weight prediction model includes an input layer, a hidden layer, and an output layer. The hidden layer is used to extract and map the weight configuration information passed in by the input layer, and the output layer is used to convert the processing results of the hidden layer into the weight prediction results.
[0108] Correspondingly, after setting the index value corresponding to the highest-level node in the autonomy index evaluation system as the actual autonomy score of the drone, it is also possible to determine whether the actual autonomy score is within the preset score range; if not, it means that there is an error in the weight allocation. At this time, the parameters of the weight prediction model can be updated according to the degree of deviation between the actual autonomy score and the preset score range, so as to use the updated weight prediction model to re-allocate the corresponding actual weight to each of the basic autonomy factors.
[0109] As a feasible implementation, this embodiment can send the weight query table to multiple terminal devices so that each terminal device returns corresponding weight configuration information. In order to improve the accuracy of the actual weight, this embodiment can set a corresponding confidence level for each terminal device, and then perform weighted calculation on all the weight configuration information according to the confidence level of each terminal device, and assign a corresponding actual weight to each of the autonomous basic factors according to the weighted calculation result.
[0110] As a feasible implementation, after the terminal device returns the weight configuration information, it can also be determined whether the weight configuration information includes the actual weights of all the autonomous basic factors; if not, the autonomous basic factors that are not configured with actual weights in the weight configuration information are set as irrelevant factors, and the nodes corresponding to the irrelevant factors in the autonomous index evaluation system are removed to update the autonomous index evaluation system. The above method can realize the automatic update of the autonomous index evaluation system and improve the accuracy of drone performance detection.
[0111] The following is an example of an actual application to illustrate the structure of the above autonomous indicator evaluation system. Figure 4 , Figure 4 This is a structural diagram of an autonomy indicator evaluation system in a practical application provided by an embodiment of the present application. The autonomy indicator evaluation system includes a target layer, a capability layer, an indicator layer and a factor layer.
[0112] The nodes in the target layer are: UAV autonomy capabilities.
[0113] The nodes in the capability layer are: task complexity, environmental complexity, and degree of human-computer interaction.
[0114] The nodes in the indicator layer are: tasks and decisions, planning and analysis, coordination and collaboration, situational awareness, electromagnetic interference, meteorology, recognition performance, interaction, learning ability and operators.
[0115] The nodes in the factor layer are: task constraints, task types, decision structures, dynamic programming, performance analysis, cost analysis, heterogeneous system collaboration, multi-system collaboration, environmental perception, working status perception, natural electromagnetic radiation, artificial electromagnetic radiation, other electromagnetic interference, day and night, wind, rain, haze, dust, target tracking, obstacle inspection and avoidance, terrain classification, settlement classification, dynamic collision avoidance perception, interaction frequency, interaction level, behavior learning, learning based on feature classifiers, multi-system obstacle avoidance behavior learning, advanced assistance behavior learning, skill level, and work intensity.
[0116] The analysis of the autonomous index of UAV mainly analyzes the influencing factors in the execution of UAV tasks, gives multiple indicators that affect the autonomy of the UAV system, and forms a multi-level autonomy evaluation index system. This embodiment, in accordance with the theoretical requirements of the hierarchical analysis method, takes the autonomous capability of UAV as the evaluation target, and constructs an autonomy index evaluation system from the three dimensions of mission complexity, environmental complexity and human-machine interaction.
[0117] The mission complexity indicates the complexity level of the mission that the drone can handle. The higher the mission complexity, the higher the ability of the drone system to handle complex missions. This embodiment defines the mission complexity by dividing it into four aspects: mission and decision-making, planning and analysis, coordination and collaboration, and situational awareness.
[0118] The description of tasks and decisions is as follows: Tasks can be defined from task constraints and task types. Task constraints refer to the time and space requirements, safety levels, and knowledge reserve levels required to complete the task. Task types can range from high-level, high-complexity, and high-adversarial tasks to low-level basic tasks. The decision-making structure mainly refers to the drone collecting data on the current situation based on sensors in emergency situations and outputting key information through the data fusion module, comparing it with the pre-set models and rules to make correct decisions. Based on the decisions made, the drone system will perform corresponding operations and feedback the execution results to the control center in real time to further optimize the strategy.
[0119] The autonomous capability of drones requires that drones be able to independently plan and re-plan in emergency situations. They need to have capabilities such as knowledge base, mission understanding and work intelligence, and be able to complete dynamic planning, performance analysis and cost analysis based on mission requirements.
[0120] UAV collaboration can be divided into multi-system collaboration and heterogeneous system collaboration. Multi-system collaboration means that multiple UAV systems can form formations, search, reconnaissance and tasks, and improve work efficiency through task division; heterogeneous system collaboration means that UAV systems can complete system collaboration with ground unmanned systems and underwater unmanned systems through data links and ground stations, forming a three-dimensional working environment in multiple aspects such as land, sea, air, space, and electromagnetic.
[0121] Situation awareness can be divided into environmental awareness and working status awareness according to different elements of the situation.
[0122] Environmental complexity indicates the degree of environmental complexity that the drone can adapt to. The higher the environmental complexity, the stronger the ability of the drone system to adapt to complex working environments and the stronger its autonomous ability. This embodiment defines environmental complexity by dividing it into three aspects: electromagnetic interference, meteorology, and recognition performance.
[0123] The electromagnetic interference to drones can be divided into natural electromagnetic radiation, artificial electromagnetic radiation and other electromagnetic radiation. Once the data link of the drone system is subject to high-intensity and multi-means electromagnetic interference, the connection between the ground control station and the drone will be cut off, the drone's automatic flight program will be destroyed, or other equipment will release active interference to deceive the system uplink, causing the drone to receive wrong instructions, which will have a significant impact on the execution of the mission.
[0124] The evaluation of UAV autonomous capabilities is based on meteorological conditions, mainly considering the impact of factors such as day and night, wind, clouds, rain and snow on visibility, flight conditions, etc. Other meteorological conditions such as climate and season are known prior information and are not included in the relevant indicators.
[0125] Identification performance can be divided into target tracking, obstacle inspection and avoidance, terrain, settlement classification and dynamic collision avoidance perception according to the different elements of the drone system's perception of the environment. Target tracking means that the drone can automatically identify the target and its surrounding environment, track the target in real time, and assess the risk of executing the mission; obstacle inspection and avoidance mainly face obstacles that the drone may face when moving at high speed (such as flying in the woods). The higher the speed, the more difficult it is to detect and avoid; terrain classification mainly faces environments such as mountains and plains, involving vegetation, slopes, rivers and other elements. The drone needs to be able to perceive various terrain obstacles in real time, determine the flyable area and re-plan in real time. Dynamic collision avoidance perception means that when multiple drone systems work together, they can estimate the spatial relationship between themselves and obstacles, maintain different routes from other drone systems and avoid obstacles.
[0126] Human-machine interaction indicates the degree of dependence of the UAV system on humans. The higher the degree of human-machine interaction, the higher the ability of the UAV system to work independently and the lower the dependence on the operator. This embodiment divides the degree of human-machine interaction into three aspects: interaction, learning ability and operator for measurement.
[0127] UAV interaction can be divided into interaction frequency and interaction level. Interaction frequency refers to the ratio of the interaction time between the UAV and the operator to the total mission time of the UAV system. The longer the operator's control time, the greater the interaction frequency, the lower the degree of human-machine interaction, and the poorer the autonomous capability of the UAV system. The interaction level refers to the task level of the operator's interaction with the UAV system, which can be measured by multiple factors such as the communication and interaction method between the operator and the UAV, and the intelligence level of the UAV system. The higher the task level, the higher the degree of human-machine interaction and the stronger the autonomous capability of the UAV system.
[0128] Learning ability refers to a class of programs that improve performance through experience. Currently, machine learning can be divided into behavioral learning, feature classifier-based learning, multi-system obstacle avoidance learning, and advanced collaborative behavioral learning. The first two types of learning can be considered basic learning, and the latter two types of learning can be considered reinforcement learning. Behavioral learning is a machine learning technology in which artificial intelligence software or systems use algorithms and models to automatically absorb data and identify and learn human behavior patterns and trends; based on feature classifiers, different samples (such as images, sounds, etc.) are classified into a fixed number of categories. It uses machine learning algorithms to learn features from samples and then distinguish different categories based on these features; the core of multi-system obstacle avoidance learning is to allow drones to detect environments and obstacles through multiple different sensors and systems. Using machine learning algorithms, drones can learn and obtain more information about their surroundings in this data, so as to avoid possible obstacles more predictively and in real time; advanced collaborative behavioral learning is an advanced technology in the unmanned field. Drones can predict and execute control behaviors and dynamic patterns through their machine learning algorithms and models to improve safety and accuracy.
[0129] The operator is mainly concerned about his skill level and work intensity. The skill level can be comprehensively measured by multiple aspects such as the level of manipulation technology, awareness and coordination ability, flight control ability, programming and software operation ability, safety awareness and emergency response ability. These aspects include the operator's proficiency in the drone controller, remote control and ground station, the ability to configure and debug the flight control system, the ability to deal with system failures and driving errors, the programming and software operation ability of intelligent functions, and good safety awareness and emergency response ability. By evaluating these skill levels, the operator's operating ability, safety management and problem-solving ability can be measured; work intensity can be measured by comprehensive factors such as task volume, work environment factors, work pressure, working hours and psychological factors. The operator needs to often work continuously for a long time, constantly observe, adjust the parameters of the flight control system, etc., and at the same time bear pressure from multiple aspects such as task difficulty and completion quality. Therefore, it is necessary to have a certain physical quality, psychological tolerance and coping ability to ensure that the flight mission can be completed efficiently, stably and safely. If the work intensity is low, it means that the drone system can complete most tasks autonomously and has strong autonomy.
[0130] This embodiment establishes an autonomy evaluation index system for drones. The index system decomposes the autonomy index into three primary indicators: task complexity, environmental complexity, and human-computer interaction level, and further decomposes it downward into several indicator factors, extending layer by layer. It includes establishing an environmental complexity evaluation system to calculate the environmental complexity of the drone system during the execution of the task; establishing a task complexity evaluation system to calculate the complexity of the drone system in executing the task; establishing a human-computer interaction level evaluation system to calculate the human-computer interaction level of the drone system. This embodiment can establish a drone autonomy capability evaluation model (i.e., a knowledge graph), input the environmental complexity, task complexity, and human-computer interaction level into the drone autonomy capability evaluation model, and obtain the actual autonomy score. The drone's operating mode can be dynamically adjusted according to the actual autonomy score to change the operator's control authority.
[0131] See also Figure 5 , Figure 5 A schematic diagram of the structure of a UAV autonomy assessment system based on a knowledge graph provided in an embodiment of the present application includes:
[0132] An evaluation system construction module 501 is used to construct an autonomy index evaluation system for the UAV; wherein the autonomy index evaluation system includes multiple levels of nodes, the highest level node in the autonomy index evaluation system is the autonomous capability of the UAV, and the lowest level node in the autonomy index evaluation system is the basic factor of autonomy;
[0133] A training module 502 is used to obtain training samples from a data set; wherein the training samples include sample weights and sample index values of the autonomy basic factors; further used to calculate the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph using the sample weights and the sample index values, and set the index value corresponding to the highest-level node in the autonomy index evaluation system as the autonomy prediction score of the drone; further used to calculate a loss function value based on the autonomy prediction score and the autonomy reference score corresponding to the training samples, so as to iteratively update the knowledge graph using the loss function value;
[0134] The task determination module 503 is used to determine the corresponding target task according to the performance detection request if a performance detection request is received;
[0135] The weight allocation module 504 is used to allocate a corresponding actual weight to each of the autonomous basic factors according to the model of the drone and the task content of the target task; wherein the sum of the actual weights of all the autonomous basic factors belonging to the same parent node is 1;
[0136] An index value determination module 505 is used to control the UAV to execute the target task to obtain a task execution log, and determine the actual index values of all the autonomy basic factors according to the task execution log;
[0137] The autonomy evaluation module 506 is used to calculate the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph using the actual weight and the actual index value, and set the index value corresponding to the highest-level node in the autonomy index evaluation system as the actual autonomy score of the drone, so as to optimize the performance of the drone according to the actual autonomy score.
[0138] This embodiment constructs an autonomy index rating system for drones. There are multiple levels of nodes in the autonomy index rating system. The highest-level node in the autonomy index rating system is the autonomous capability of the drone, and the remaining nodes are factors that directly or indirectly affect the autonomous capability of the drone. The most basic factor affecting the autonomous capability of the drone in the autonomy index rating system is the basic autonomy factor corresponding to the lowest-level node. This embodiment obtains training samples from a data set, and uses the training samples to calculate the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph, and finally obtains the autonomy prediction score of the drone. This embodiment calculates the loss function value based on the autonomy prediction score of the drone and the autonomy reference score corresponding to the training sample, and then uses the loss function value to iteratively update the knowledge in the knowledge graph, so that the weight relationship and index value relationship between adjacent level nodes in the autonomy index evaluation system are correctly recorded in the knowledge graph. After receiving the performance test request, this embodiment assigns a corresponding actual weight to each of the autonomous basic factors according to the model of the drone and the task content of the target task, determines the actual index values of all autonomous basic factors according to the task execution log when the drone performs the target task, and calculates the weight and index value of each node in the autonomous index evaluation system layer by layer based on the knowledge graph using the actual weight and the actual index value, thereby obtaining the actual autonomy score of the drone, so as to optimize the performance of the drone according to the actual autonomy score. This embodiment determines the actual autonomy score of the drone based on the autonomous index rating system of the drone and the knowledge graph obtained by iterative training, realizes the quantitative evaluation of the autonomy of the drone, and improves the accuracy of the performance test of the drone.
[0139] Furthermore, the process of the evaluation system construction module 501 constructing an autonomy index evaluation system for the UAV includes: determining the corresponding multiple autonomy evaluation dimensions according to the model of the UAV; selecting the current autonomy evaluation dimension from all the autonomy evaluation dimensions; querying multiple influencing factors of the current autonomy evaluation dimension; wherein the influencing factors include direct influencing factors and indirect influencing factors; constructing an organizational structure chart corresponding to the current autonomy evaluation dimension according to the correlation relationship between all the influencing factors; judging whether the organizational structure chart corresponding to all the autonomy evaluation dimensions is obtained; if so, constructing the autonomy index evaluation system according to the organizational structure chart corresponding to all the autonomy evaluation dimensions; if not, entering the step of selecting the current autonomy evaluation dimension from all the autonomy evaluation dimensions.
[0140] Furthermore, the process of the evaluation system construction module 501 constructing the autonomy indicator evaluation system according to the organizational structure charts corresponding to all the autonomy evaluation dimensions includes: integrating the organizational structure charts corresponding to all the autonomy evaluation dimensions to obtain an overall organizational structure chart; judging whether there is a dispute node in the overall organizational structure chart; wherein the dispute node is a single child node with at least two parent nodes in the overall organizational structure chart; if so, setting the parent node with the highest degree of correlation with the dispute node as a true parent node, and setting all parent nodes of the dispute node except the true parent node as false parent nodes; removing the parent-child node relationship between the false parent node and the dispute node in the overall organizational structure chart to obtain the autonomy indicator evaluation system.
[0141] Furthermore, the process of the indicator value determination module 505 determining the actual indicator values of all the autonomous basic factors according to the task execution log includes: obtaining the indicator value calculation rules corresponding to each of the autonomous basic factors in the autonomous indicator evaluation system; querying the log content matching each of the autonomous basic factors in the task execution log; and calculating the log content corresponding to the same autonomous basic factor using the indicator value calculation rules to obtain the actual indicator values corresponding to all the autonomous basic factors.
[0142] Furthermore, the process in which the weight allocation module 504 allocates a corresponding actual weight to each of the autonomous basic factors according to the model of the drone and the task content of the target task includes: constructing a weight query table according to all the autonomous basic factors in the autonomous index evaluation system, and adding the model of the drone and the task content of the target task in the weight query table; sending the weight query table to the terminal device so that the terminal device returns weight configuration information; wherein the weight configuration information includes the correspondence between the autonomous basic factors and the actual weights; and allocating a corresponding actual weight to each of the autonomous basic factors according to the weight configuration information.
[0143] Furthermore, the process in which the weight allocation module 504 allocates a corresponding actual weight to each of the autonomous basic factors according to the weight configuration information includes: inputting the weight configuration information into a weight prediction model, and allocating a corresponding actual weight to each of the autonomous basic factors according to the weight prediction result of the weight prediction model; wherein the weight prediction model is a model obtained by training based on a deep learning framework, and the weight prediction model includes an input layer, a hidden layer and an output layer, wherein the hidden layer is used to extract and map features of the weight configuration information transmitted from the input layer, and the output layer is used to convert the processing result of the hidden layer into the weight prediction result;
[0144] Correspondingly, it also includes:
[0145] The weight updating module is used to set the indicator value corresponding to the highest-level node in the autonomy indicator evaluation system as the actual autonomy score of the UAV, and then determine whether the actual autonomy score is within a preset score range; if not, update the parameters of the weight prediction model according to the degree of deviation between the actual autonomy score and the preset score range, so as to use the updated weight prediction model to re-assign the corresponding actual weight to each of the basic autonomy factors.
[0146] Furthermore, the process of the weight allocation module 504 sending the weight query table to the terminal device includes: sending the weight query table to a plurality of the terminal devices;
[0147] Correspondingly, the process in which the weight allocation module 504 allocates a corresponding actual weight to each of the autonomous basic factors according to the weight configuration information includes: performing weighted calculation on all the weight configuration information according to the confidence of each of the terminal devices, and allocating a corresponding actual weight to each of the autonomous basic factors according to the weighted calculation result.
[0148] Furthermore, it also includes:
[0149] The system adjustment module is used to determine whether the weight configuration information includes the actual weights of all the autonomous basic factors; if not, the autonomous basic factors that are not configured with actual weights in the weight configuration information are set as irrelevant factors, and the nodes corresponding to the irrelevant factors in the autonomous indicator evaluation system are removed to update the autonomous indicator evaluation system.
[0150] Since the embodiments of the system part correspond to the embodiments of the method part, please refer to the description of the embodiments of the method part for the embodiments of the system part, which will not be repeated here.
[0151] The present application also provides a storage medium on which a computer program is stored, and when the computer program is executed, the steps provided in the above embodiment can be implemented. The storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.
[0152] The present application also provides an electronic device, which may include a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps provided in the above embodiment may be implemented. Of course, the electronic device may also include various network interfaces, power supplies and other components.
[0153] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of this application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of this application.
[0154] It should also be noted that, in this specification, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.
Claims
1. A method for evaluating the autonomy of unmanned aerial vehicles based on knowledge graph, characterized in that: include: Constructing an autonomy index evaluation system for the UAV; wherein the autonomy index evaluation system includes multiple levels of nodes, the highest level node in the autonomy index evaluation system is the autonomous capability of the UAV, and the lowest level node in the autonomy index evaluation system is the basic factor of autonomy; Acquire training samples from the data set; wherein the training samples include sample weights and sample index values of the basic factors of autonomy; Based on the knowledge graph, the weight and index value of each node in the autonomy index evaluation system are calculated layer by layer using the sample weight and the sample index value, and the index value corresponding to the highest-level node in the autonomy index evaluation system is set as the autonomy prediction score of the drone; Calculating a loss function value according to the autonomy prediction score and the autonomy reference score corresponding to the training sample, so as to iteratively update the knowledge graph using the loss function value; If a performance test request is received, determining a corresponding target task according to the performance test request; Allocating a corresponding actual weight to each of the autonomous basic factors according to the model of the drone and the task content of the target task; wherein the sum of the actual weights of all the autonomous basic factors belonging to the same parent node is 1; Controlling the UAV to execute the target task to obtain a task execution log, and determining actual index values of all the basic factors of autonomy according to the task execution log; Based on the knowledge graph, the weight and index value of each node in the autonomy index evaluation system are calculated layer by layer using the actual weight and the actual index value, and the index value corresponding to the highest-level node in the autonomy index evaluation system is set as the actual autonomy score of the drone, so as to optimize the performance of the drone according to the actual autonomy score; Among them, an autonomous index evaluation system is built for drones, including: Determining a plurality of corresponding autonomy evaluation dimensions according to the model of the drone; Selecting a current autonomy evaluation dimension from all the autonomy evaluation dimensions; Querying multiple influencing factors of the current autonomy evaluation dimension; wherein the influencing factors include direct influencing factors and indirect influencing factors; The organizational structure chart corresponding to the current autonomy evaluation dimension is constructed based on the correlation between all the influencing factors; Determining whether the organizational structure chart corresponding to all the autonomy evaluation dimensions is obtained; If so, constructing the autonomy indicator evaluation system according to the organizational structure chart corresponding to all the autonomy evaluation dimensions; If not, entering the step of selecting the current autonomy evaluation dimension from all the autonomy evaluation dimensions; The autonomy indicator evaluation system is constructed according to the organizational structure chart corresponding to all the autonomy evaluation dimensions, including: Integrate the organizational structure charts corresponding to all the autonomy evaluation dimensions to obtain an overall organizational structure chart; Determine whether there is a dispute node in the overall organizational structure diagram; wherein the dispute node is a single child node in the overall organizational structure diagram that has at least two parent nodes; If yes, then set the parent node with the highest correlation with the dispute node as the true parent node, and set all parent nodes of the dispute node except the true parent node as false parent nodes; The parent-child node relationship between the false parent node and the dispute node in the overall organizational structure diagram is removed to obtain the autonomy indicator evaluation system.
2. The method for evaluating the autonomy of unmanned aerial vehicles based on knowledge graph according to claim 1 is characterized in that: Determine the actual index values of all the basic factors of autonomy according to the task execution log, including: Obtaining the indicator value calculation rules corresponding to each of the autonomy basic factors in the autonomy indicator evaluation system; querying the task execution log for log content matching each of the autonomy basic factors; The index value calculation rule is used to calculate the log content corresponding to the same autonomy basic factor to obtain the actual index values corresponding to all the autonomy basic factors.
3. The method for evaluating the autonomy of unmanned aerial vehicles based on knowledge graph according to claim 1 is characterized in that: According to the model of the UAV and the task content of the target task, each of the basic factors of autonomy is assigned a corresponding actual weight, including: Constructing a weight query table according to all the autonomy basic factors in the autonomy index evaluation system, and adding the model of the drone and the task content of the target task to the weight query table; Sending the weight query table to a terminal device so that the terminal device returns weight configuration information; wherein the weight configuration information includes a correspondence between the autonomy basic factors and the actual weights; A corresponding actual weight is allocated to each of the autonomous basic factors according to the weight configuration information.
4. The method for evaluating the autonomy of unmanned aerial vehicles based on knowledge graph according to claim 3 is characterized in that: Allocating a corresponding actual weight to each of the autonomous basic factors according to the weight configuration information includes: Input the weight configuration information into a weight prediction model, and assign a corresponding actual weight to each of the autonomy basic factors according to the weight prediction result of the weight prediction model; wherein the weight prediction model is a model trained based on a deep learning framework, and the weight prediction model includes an input layer, a hidden layer, and an output layer, wherein the hidden layer is used to extract and map features of the weight configuration information transmitted from the input layer, and the output layer is used to convert the processing result of the hidden layer into the weight prediction result; Accordingly, after setting the index value corresponding to the highest level node in the autonomy index evaluation system as the actual autonomy score of the drone, the method further includes: Determining whether the actual autonomy score is within a preset score range; If not, the parameters of the weight prediction model are updated according to the degree of deviation between the actual autonomy score and the preset score range, so as to use the updated weight prediction model to re-assign the corresponding actual weight to each of the basic autonomy factors.
5. The method for evaluating the autonomy of unmanned aerial vehicles based on knowledge graph according to claim 3 is characterized in that: Sending the weight query table to the terminal device includes: Sending the weight query table to the plurality of terminal devices; Accordingly, allocating a corresponding actual weight to each of the autonomous basic factors according to the weight configuration information includes: All the weight configuration information is weighted according to the confidence of each terminal device, and a corresponding actual weight is allocated to each of the autonomous basic factors according to the weighted calculation result.
6. The method for evaluating the autonomy of unmanned aerial vehicles based on knowledge graph according to claim 3 is characterized in that: After the terminal device returns the weight configuration information, the method further includes: Determining whether the weight configuration information includes actual weights of all the autonomous basic factors; If not, the autonomous basic factors that are not configured with actual weights in the weight configuration information are set as irrelevant factors, and the nodes corresponding to the irrelevant factors in the autonomous indicator evaluation system are removed to update the autonomous indicator evaluation system.
7. A knowledge graph-based drone autonomy assessment system, characterized in that: include: An evaluation system construction module is used to construct an autonomy index evaluation system for the UAV; wherein the autonomy index evaluation system includes multiple levels of nodes, the highest level node in the autonomy index evaluation system is the autonomous capability of the UAV, and the lowest level node in the autonomy index evaluation system is the basic factor of autonomy; A training module, used to obtain training samples from a data set; wherein the training samples include sample weights and sample index values of the autonomy basic factors; further used to calculate the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph using the sample weights and the sample index values, and set the index value corresponding to the highest-level node in the autonomy index evaluation system as the autonomy prediction score of the drone; further used to calculate a loss function value based on the autonomy prediction score and the autonomy reference score corresponding to the training samples, so as to iteratively update the knowledge graph using the loss function value; A task determination module, configured to determine a corresponding target task according to a performance detection request if a performance detection request is received; A weight allocation module, used to allocate a corresponding actual weight to each of the autonomous basic factors according to the model of the drone and the task content of the target task; wherein the sum of the actual weights of all the autonomous basic factors belonging to the same parent node is 1; an index value determination module, used for controlling the UAV to execute the target task to obtain a task execution log, and determining the actual index values of all the autonomous basic factors according to the task execution log; An autonomy evaluation module, configured to calculate the weight and index value of each node in the autonomy index evaluation system layer by layer based on the knowledge graph using the actual weight and the actual index value, and set the index value corresponding to the highest-level node in the autonomy index evaluation system as the actual autonomy score of the drone, so as to optimize the performance of the drone according to the actual autonomy score; Among them, the process of the evaluation system construction module constructing an autonomy index evaluation system for the UAV includes: determining the corresponding multiple autonomy evaluation dimensions according to the model of the UAV; selecting the current autonomy evaluation dimension from all the autonomy evaluation dimensions; querying multiple influencing factors of the current autonomy evaluation dimension; wherein the influencing factors include direct influencing factors and indirect influencing factors; constructing an organizational structure chart corresponding to the current autonomy evaluation dimension according to the correlation relationship between all the influencing factors; judging whether the organizational structure chart corresponding to all the autonomy evaluation dimensions is obtained; if so, constructing the autonomy index evaluation system according to the organizational structure chart corresponding to all the autonomy evaluation dimensions; if not, entering the step of selecting the current autonomy evaluation dimension from all the autonomy evaluation dimensions; Among them, the process of the evaluation system construction module constructing the autonomy index evaluation system according to the organizational structure charts corresponding to all the autonomy evaluation dimensions includes: integrating the organizational structure charts corresponding to all the autonomy evaluation dimensions to obtain an overall organizational structure chart; judging whether there is a dispute node in the overall organizational structure chart; wherein the dispute node is a single child node with at least two parent nodes in the overall organizational structure chart; if so, setting the parent node with the highest degree of correlation with the dispute node as a true parent node, and setting all parent nodes of the dispute node except the true parent node as false parent nodes; removing the parent-child node relationship between the false parent node and the dispute node in the overall organizational structure chart to obtain the autonomy index evaluation system.
8. An electronic device, characterized in that: It includes a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps of the drone autonomy assessment method based on the knowledge graph are implemented as described in any one of claims 1 to 6.
