Emergency flight plan generation method, device, equipment, medium and program product

Abstract

The invention relates to the technical field of emergency communication, and provides an emergency flight plan generation method and device, equipment, a medium and a program product, and the method comprises the steps: predicting an emergency event in an emergency planning region, and obtaining the event description information of the emergency event; generating a multi-dimensional label of the emergency event according to the event description information; and inputting the multi-dimensional label and the event description information into a plan generation model, and generating an emergency flight plan of a low airspace in the emergency planning area by using the plan generation model. The plan generation model is obtained through iterative training based on a sample data set, and the sample data set is constructed based on historical emergency events in the emergency planning area and historical flight plans corresponding to the historical emergency events. The emergency flight plan can be generated online in real time through the plan generation model according to the prediction result of the emergency event, the efficiency and accuracy of generating the emergency flight plan are improved, and the requirement for frequent application of the low-airspace emergency flight plan can be met.

Description

Technical Field

[0001] The present application relates to the field of emergency communication technology, and in particular to an emergency flight plan generation method, device, equipment, medium and program product. Background Art

[0002] With the development of drone technology, the application of drones in the field of emergency rescue has become increasingly widespread. With the frequent, strong and widespread trend of emergency events such as natural disasters, emergency management units in various regions usually need to report emergency flight plans in low-altitude areas before emergency events occur, so as to apply for emergency flight permissions in low-altitude areas after emergency events occur. In the existing technology, the generation of emergency flight plans in low-altitude areas mainly relies on manual or expert decision-making, and the application for flight permissions is reported on an annual or semi-annual basis. This method is inefficient, time-consuming and labor-intensive, and prone to misjudgment, and it is difficult to meet the current demand for frequent applications for low-altitude emergency airspace. Summary of the invention

[0003] The present application provides an emergency flight plan generation method, device, equipment, medium and program product to address the defects in the prior art of relying on manual or expert decision-making to generate emergency flight plans, which are inefficient, time-consuming, labor-intensive, and prone to misjudgment, and are difficult to meet the current demand for frequent applications for low-altitude emergency airspace.

[0004] The present application provides a method for generating an emergency flight plan, comprising: Predicting emergency events within the emergency planning area and obtaining event description information of the emergency events; Generate a multi-dimensional label of the emergency event according to the event description information; the multi-dimensional label at least includes an event time label, an event type label and a three-break event label; Inputting the multi-dimensional label and the event description information into a plan generation model, and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; The plan generation model is obtained by iterative training based on a sample data set, and the sample data set is constructed based on historical emergency events in the emergency planning area and historical flight plans corresponding to the historical emergency events.

[0005] In one embodiment, generating the multi-dimensional label of the emergency event according to the event description information includes: Inputting the event description information into an encoder of a tag generator; the encoder includes a first word embedding layer and a long short-term memory layer, and the tag generator also includes a decoder, and the decoder includes a first fully connected layer and a first output layer; Using the first word embedding layer in the encoder, converting the event description information into a word vector, and extracting a feature vector from the word vector through the long short-term memory layer; The feature vector is input into the first fully connected layer of the decoder for decoding, and label classification prediction is performed on the decoded feature vector based on the first output layer of the decoder to obtain a multi-dimensional label of the emergency event.

[0006] In one embodiment, the plan generation model includes a multi-label input layer, a second word embedding layer, a large language model layer, a second fully connected layer, a discard layer, and a second output layer; the multi-dimensional label and the event description information are input into the plan generation model, and the plan generation model is used to generate an emergency flight plan for the low-altitude area within the emergency planning area, including: Inputting the multidimensional label and the event description information into a plan generation model, concatenating the multidimensional label and the event description information using the multi-label input layer of the plan generation model, and then inputting the concatenated labels into a second word embedding layer of the plan generation model; The concatenated multi-dimensional label and event description information are converted into a target word vector through the second word embedding layer; Using the large language model layer, the second fully connected layer and the discard layer of the planned generation model to perform feature extraction processing on the target word vector in sequence to obtain a target feature vector; The target feature vector is input into the second output layer of the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; the emergency flight plan includes a flight time period, a flight area and a flight mission.

[0007] In one embodiment, the inputting the multi-dimensional label and the event description information into a plan generation model, and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area, includes: Obtaining the time granularity for generating an emergency flight plan in the emergency planning area; The multi-dimensional label and the event description information are input into a plan generation model, and the plan generation model is used to generate an emergency flight plan for the low-altitude area within the emergency planning area at the time granularity.

[0008] In one embodiment, before inputting the multi-dimensional label and the event description information into a plan generation model and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area, the process further includes: Obtain information on historical emergency events within the emergency planning area; Extracting historical description information and multi-dimensional information of the historical emergency event from the news information; the multi-dimensional information includes event time, event type and three-break event identifier; the three-break event identifier is used to indicate whether the historical emergency event is a three-break event; Acquire a historical emergency flight plan of the historical emergency event, use the multidimensional information as a multidimensional label of the historical description information, use the historical emergency flight plan as a prediction result of the historical description information, and construct a sample data set; The preset plan generation model is iteratively trained using the sample data set to obtain a pre-trained plan generation model.

[0009] In one embodiment, the iterative training of a preset plan generation model using the sample data set to obtain a pre-trained plan generation model includes: Based on a preset number of iterations, the sample data set is input into a preset plan generation model to perform iterative training on the plan generation model; In each round of iteration, the preset multi-class cross entropy is used as the loss function, and the preset gradient descent optimization algorithm is adopted to adjust the model parameters of the plan generation model to optimize the function value of the loss function, and when the function value of the loss function is optimal, the pre-trained plan generation model is obtained.

[0010] The present application also provides an emergency flight plan generation device, comprising the following modules: An event prediction module is used to predict emergency events within the emergency planning area and obtain event description information of the emergency events; A label generation module, used to generate a multi-dimensional label of the emergency event according to the event description information; the multi-dimensional label at least includes an event time label, an event type label and a three-break event label; A plan generation module, used for inputting the multi-dimensional label and the event description information into a plan generation model, and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; The plan generation model is obtained by iterative training based on a sample data set, and the sample data set is constructed based on historical emergency events in the emergency planning area and historical flight plans corresponding to the historical emergency events.

[0011] The present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, an emergency flight plan generation method as described in any one of the above is implemented.

[0012] The present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the emergency flight plan generation method as described in any one of the above is implemented.

[0013] The present application also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the emergency flight plan generation method as described in any one of the above is implemented.

[0014] The emergency flight plan generation method, device, equipment, medium and program product provided by the present application predicts emergency events and generates multi-dimensional labels based on the description information of the predicted emergency events. The plan generation model can generate the corresponding low-altitude emergency flight plan based on the multi-dimensional labels and description information of the emergency events, thereby realizing the automatic generation of the emergency flight plan. Through the plan generation model, the emergency flight plan for the low-altitude area within the emergency planning area can be generated online in real time based on the prediction results of the emergency events, which improves the efficiency and accuracy of generating emergency flight plans, saves time and effort, and can realize the automatic submission and application of low-altitude emergency flight permissions based on the emergency flight plan, which is conducive to meeting the needs of frequent applications for low-altitude emergency flight plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are 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.

[0016] Figure 1 It is a flow chart of the emergency flight plan generation method provided in an embodiment of the present application.

[0017] Figure 2 It is a schematic diagram of the generation process of the multi-dimensional label provided in the embodiment of the present application.

[0018] Figure 3 This is one of the schematic diagrams of the generation process of the emergency flight plan provided in the embodiment of the present application.

[0019] Figure 4 This is the second schematic diagram of the emergency flight plan generation process provided in the embodiment of the present application.

[0020] Figure 5 It is a schematic diagram of the training process of the plan generation model provided in the embodiment of the present application.

[0021] Figure 6 It is a structural schematic diagram of the emergency flight plan generating device provided by the present invention.

[0022] Figure 7 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0024] Figure 1 is a flow chart of the method for generating an emergency flight plan provided in an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps: Step 100, predicting emergency events within the emergency planning area and obtaining event description information of the emergency events; Step 200, generating a multi-dimensional label of the emergency event according to the event description information; the multi-dimensional label at least includes an event time label, an event type label and a three-break event label; Step 300, inputting the multi-dimensional label and the event description information into a plan generation model, and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; The plan generation model is obtained by iterative training based on a sample data set, and the sample data set is constructed based on historical emergency events in the emergency planning area and historical flight plans corresponding to the historical emergency events.

[0025] Predict emergency events in the emergency planning area, obtain event description information of the emergency events, and predict emergency events in a manner that includes but is not limited to using a prediction model to predict, and mining the occurrence patterns of emergency events based on historical emergency events, and predicting according to the occurrence patterns of emergency events. Wherein, when using a prediction model to predict emergency events, the prediction model can be obtained by constructing a sample data set based on historical emergency events that have occurred in the emergency planning area, and iteratively training a preset basic prediction model.

[0026] Optionally, the prediction of emergency events is based on the environmental information in the emergency planning area. The environmental information in the emergency planning area is monitored, environmental features are extracted, and the environmental features are matched with historical emergency events to determine whether an emergency event of the same type as the historical emergency event will occur. The collected environmental information includes meteorological data.

[0027] Furthermore, the event description information of an emergency event is key information used to describe the emergency event, which includes at least the time of occurrence of the emergency event, the event type of the emergency event, and whether the emergency event is a three-outage event, etc. The three-outage events are "circuit outage, power outage, and network outage" events.

[0028] According to the predicted event description information of the emergency event, a multi-dimensional label of the emergency event is generated, which includes an event time label, an event type label and a three-break event label. Among them, the event time label is used to characterize the occurrence time of the emergency event, and is generated based on the time information of the emergency event extracted from the event description information; the event type label is used to characterize the event type of the emergency event, and is generated based on the event type extracted from the event description information; the three-break event label is used to characterize whether the emergency event is a three-break event, and the three-break event label is generated based on the three-break event identifier extracted from the event description information.

[0029] The generated multi-dimensional labels and event description information of the emergency events are input into the plan generation model, and the plan generation model is used to generate an emergency flight plan for the low-altitude airspace within the emergency planning area. The emergency flight plan at least includes the flight time period, flight area and flight mission, etc.

[0030] In this embodiment, by predicting emergency events and generating multidimensional labels based on the description information of the predicted emergency events, the plan generation model can be used to generate the corresponding low-altitude emergency flight plan based on the multidimensional labels and description information of the emergency events, thereby realizing the automatic generation of the emergency flight plan. Through the plan generation model, the emergency flight plan for the low-altitude area within the emergency planning area can be generated online in real time based on the prediction results of the emergency events, which improves the efficiency and accuracy of generating the emergency flight plan, saves time and effort, and can realize the automatic submission and application of the low-altitude emergency flight authority based on the emergency flight plan, which is conducive to meeting the demand for frequent application of low-altitude emergency flight plans.

[0031] Furthermore, the multi-dimensional label of the emergency event is generated based on a label generator, and the label generator is pre-trained based on historical emergency events. Step 200 may also include: Step 201, inputting the event description information into an encoder of a tag generator; the encoder includes a first word embedding layer and a long short-term memory layer, the tag generator also includes a decoder, and the decoder includes a first fully connected layer and a first output layer; Step 202, using the first word embedding layer in the encoder, converting the event description information into a word vector, and extracting a feature vector from the word vector through the long short-term memory layer; Step 203: input the feature vector into the first fully connected layer of the decoder for decoding, and perform label classification prediction on the decoded feature vector based on the first output layer of the decoder to obtain a multi-dimensional label of the emergency event.

[0032] The tag generator is built based on the long short-term memory network (LSTM), and mainly includes two parts: an encoder and a decoder, wherein the encoder further includes a first word embedding layer and a long short-term memory layer, and the decoder includes a first fully connected layer and a first output layer. The tag generator further includes an input layer, and optionally, the long short-term memory layer may include one or more layers, and the first fully connected layer in the decoder may also include one or more layers.

[0033] Reference Figure 2 The label generation process based on the label generator shown in the figure first inputs the event description information of the emergency event into the encoder of the label generator through the input layer of the label generator, converts the event description information into a word vector using the first word embedding layer of the encoder, and extracts the feature vector of the emergency event from the word vector through the long short-term memory layer.

[0034] Then, the extracted feature vector of the emergency event is input into the decoder for decoding, specifically, it is decoded through the first fully connected layer in the decoder. Based on the first output layer of the decoder, the decoded feature vector is subjected to label classification prediction to obtain a multi-dimensional label of the emergency event.

[0035] Optionally, the input layer is also used to index the event description information of the emergency event. The input in the label generator is the indexed event description text, and the length of each index sequence is the preset length value emergency_length. In the encoder, the time description text is first transformed by the first word embedding layer, and each word in the event description text is converted into a corresponding vector. The data dimension of the first word embedding layer input is represented as emergency_vocab_size. The output of the first embedding layer is set to convert the word into a spatial vector of a preset dimension, such as 128 dimensions. The input sequence length is emergency_length, and the output data of this layer is in the form of (None, emergency_length, 128). The function of the first word embedding layer is to perform vector mapping on the input description text, and convert the index of each word in the description text into a fixed vector of 128 dimensions. Then, the feature vector of the event description text of the emergency event is extracted through multiple LSTM layers, wherein each layer in the LSTM layer contains a first preset number of LSTM neurons, and the activation function of each layer is ReLU.

[0036] Furthermore, the first fully connected layer in the decoder includes a second preset number of neurons, the activation function is also set to ReLU, the output layer is also composed of a fully connected layer (Dense), the number of neurons is the same as the number of label dimensions of the multidimensional label to be generated, the activation function is sigmod, and the output layer is used to output the label prediction result. Exemplarily, the first preset number is 64 and the second preset number is 128.

[0037] In the pre-training stage of the label generator, based on the preset number of iterations, historical emergency events and their multidimensional labels are used as training samples. The multidimensional labels generated by the label generator are compared with the true labels of the historical emergency events. The error between the two is calculated based on the two-class logarithmic loss function. With the goal of minimizing the error, the gradient descent optimization algorithm is used to adjust the parameters of the label generator, thereby optimizing the loss function and finding the optimal parameters that minimize the loss function.

[0038] In one embodiment, the plan generation model is constructed based on a large language model, and the plan generation model specifically includes a multi-label input layer, a second word embedding layer, a large language model layer, a second fully connected layer, a discard layer, and a second output layer, wherein the large language model layer can be constructed using a general large language model (LLM).

[0039] Based on this, step 300 may also include: Step 301, inputting the multi-dimensional label and the event description information into a plan generation model, concatenating the multi-dimensional label and the event description information using the multi-label input layer of the plan generation model, and then inputting the concatenated labels into a second word embedding layer of the plan generation model; Step 302, converting the concatenated multi-dimensional label and event description information into a target word vector through the second word embedding layer; Step 303, using the large language model layer, the second fully connected layer and the discard layer of the planned generation model to perform feature extraction processing on the target word vector in sequence to obtain a target feature vector; Step 304, input the target feature vector into the second output layer of the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; the emergency flight plan includes a flight time period, a flight area, and a flight mission.

[0040] When using the plan generation model to generate an emergency flight plan for low-altitude airspace within the emergency planning area, first, the event description information and the generated multidimensional labels are input into the plan generation model, and the multidimensional labels and event description information are spliced ​​using the multi-label input layer of the plan generation model, and the spliced ​​multidimensional labels and event description information are input into the second word embedding layer of the plan generation model. Then, the spliced ​​multidimensional labels and event description information are converted into target word vectors using the second word embedding layer, and the target word vectors are feature extracted once using the large language model layer, the second fully connected layer, and the discard layer of the plan generation model to obtain the target feature vector. Finally, the target feature vector is input into the second output layer of the plan generation model to generate an emergency flight plan for low-altitude airspace within the emergency planning area, which includes a flight time period, a flight area, and a flight mission.

[0041] In one embodiment, referring to Figure 3 The generation process of the emergency flight plan shown in the figure, the plan generation model also includes an emergency event extractor, an emergency event classifier, a three-break event filter and an emergency event collector. Among them, the emergency event extractor is used to extract the event time label from the multi-dimensional label of the emergency event, the emergency event classifier is used to classify the emergency event according to the event type label in the multi-dimensional label of the emergency event, and extract the event type of the emergency event, the three-break event filter is used to filter out the three-break events in the emergency event according to the three-break event label, and the emergency event collector is used to collect the event description information of the emergency event.

[0042] Furthermore, the event time label extracted by the emergency event time extractor, the event type label extracted by the emergency event classifier, the three-break event label extracted by the three-break event filter and the time description information extracted by the emergency event collector are input into the second word embedding layer through the multi-label input layer for word mapping, and the event time label, event type label, three-break event label and event description information are respectively converted into target word vectors, and then the target word vectors are feature extracted in turn through the pre-trained large language model, the second fully connected layer, the discard layer and the output layer, and finally an emergency flight plan for the low-altitude area within the emergency planning area is generated.

[0043] In another embodiment, referring to Figure 4The emergency flight plan generation process shown in the figure first predicts the emergency events in any given emergency planning area to obtain the event description information of the emergency event, and then generates the event time label, event type label and three-break event label of the emergency event based on the event description information of the emergency event based on the pre-trained label generator. Furthermore, the generated event time label, event type label, three-break event label and time description information are input into the plan generation model to automatically generate the emergency flight plan for the low-altitude area in the emergency planning area. Finally, the generated emergency flight plan is sent to the low-altitude emergency control platform to apply for the emergency flight authority of the low-altitude area corresponding to the emergency flight plan.

[0044] Furthermore, the generated emergency flight plan may be of different time granularities. Based on this, step 300 may further include: Step 310, obtaining the time granularity for generating the emergency flight plan in the emergency planning area; Step 320: input the multi-dimensional label and the event description information into a plan generation model, and use the plan generation model to generate an emergency flight plan for the low airspace within the emergency planning area at the time granularity.

[0045] When using the plan generation model to generate an emergency flight plan for the low airspace within the emergency planning area, the time granularity for generating the emergency flight plan in the emergency planning area is first obtained, the event description information and the generated multidimensional labels are input into the plan generation model, and the plan generation model is used to generate an emergency flight plan for the low airspace within the emergency planning area at the corresponding time granularity.

[0046] Optionally, the time granularity is configurable, and the time granularity may be monthly, quarterly, or annual. According to the configured time granularity, an emergency flight plan at the configured time granularity is generated in the manner shown in steps 301 to 304 above.

[0047] Optionally, before using the plan generation model to generate an emergency flight plan based on the multi-dimensional labels and the event description information of the emergency event, it is also necessary to pre-train the plan generation model based on historical emergency events and their corresponding historical emergency flight plans. Therefore, before step 300, it may also include: Step 501, obtaining information about historical emergency events in the emergency planning area; Step 502, extracting historical description information and multi-dimensional information of the historical emergency event from the news information; the multi-dimensional information includes event time, event type and three-break event identifier; the three-break event identifier is used to indicate whether the historical emergency event is a three-break event; Step 503, obtaining the historical emergency flight plan of the historical emergency event, using the multidimensional information as the multidimensional label of the historical description information, using the historical emergency flight plan as the prediction result of the historical description information, and constructing a text data set; Step 504, iteratively train the preset plan generation model using the text data set to obtain a pre-trained plan generation model.

[0048] First, we obtain the information of historical emergency events in the emergency planning area, extract the key information of historical emergency events, and obtain the historical description information and multi-dimensional information of historical emergency events. Among them, the multi-dimensional information of historical emergency events includes event time, event type and three-break event identification. The three-break event identification is specifically marked as "yes" or "no" to characterize whether the historical emergency event is a three-break event.

[0049] Furthermore, historical emergency flight plans corresponding to historical emergency events are obtained, the extracted multidimensional information is used as the multidimensional label of the historical description information, and the historical emergency flight plans are used as the prediction results of the historical description information to construct a text dataset. The constructed text dataset is used to iteratively train the preset plan generation model to obtain a pre-trained plan generation model.

[0050] In one embodiment, information about historical emergency events and low-altitude emergency flight plans of historical emergency events marked manually or by experts are obtained from a low-altitude emergency management and control platform, and multi-dimensional information of historical emergency events and description information of historical emergency events are extracted from the information, as well as a total data set of emergency flight plan itineraries of historical emergency events.

[0051] Optionally, before constructing the sample dataset, preprocessing operations such as text cleaning and text serialization may be performed on the total dataset.

[0052] When constructing the sample data set, the extracted multidimensional information is used as the multidimensional label of the historical emergency events. Specifically, the event time of the historical emergency events is used as the time label to form the event time label set of the historical emergency events: ,in, The word vector of the time word corresponding to the time label of the Nth emergency event. The time word can be a word at different time granularities, such as monthly, quarterly, and annual. The event type of the historical emergency event is used as the event type label to form the event type label set of the historical emergency event: , The word vector of the event type word corresponding to the event type label of the Tth emergency event, such as typhoon, flood, and earthquake. The three-break event identifiers of historical emergency events are used as the three-break event labels to form the three-break event label set of historical emergency events: , The word vector of the three-break event identifier corresponding to the three-break event label of the Mth emergency event, and the label result corresponding to each three-break event identifier includes two types: "yes" or "no". Furthermore, the description information of historical emergency events is used to form a description text set: , Represents the word vector of the Wth word in the description text of the emergency event. The historical emergency flight plans of historical emergency events can form a plan set: , The word vector representing the Lth word in the historical emergency flight plan can be a flight time period, a flight area, or a flight mission. Based on multi-label datasets such as the event time label set, event type label set, and three-break event label set of historical emergency events, combined with the description text set of historical emergency events and the historical emergency flight plan set, a sample dataset for training the plan generation model is constructed.

[0053] Optionally, perform text cleaning on the sample dataset and serialize the text, including unifying letter case, converting uppercase letters to lowercase letters, removing all punctuation marks, and then perform word segmentation on the text. At the same time, tokenize each word obtained by the word segmentation so that each piece of text is converted into a piece of index numbers, and fill zeros for sequences that do not reach the preset maximum text length.

[0054] The longest length Multi-tag_length of the multi-label dataset is taken as its unified index sequence length, and the dictionary size of the multi-label dataset is set to Multi-tag_vocab_size; the longest length emergency_length of the description text set of emergency events is taken as its unified index sequence length, and the dictionary size of the description text set of emergency events is set to emergency_vocab_size; the longest length plan_length of the low-altitude emergency flight plan set is taken as its unified index sequence length, and the dictionary size of the low-altitude emergency flight plan set is set to plan_vocab_size.

[0055] The sample data set is divided into a training set and a test set. For example, 90% of the sample data set is divided into a training set, and 10% of the sample data set is divided into a test set. The training set is used to train the plan generation model, and the test set is used to test the model performance of the trained plan generation model.

[0056] Optionally, the multi-dimensional labels of historical emergency events in the sample dataset can also be generated using a label generator. Figure 5The model training process shown extracts historical description information of historical emergency events from the news information of historical emergency events, and then extracts the event time of the historical emergency events from the historical description information through a label generator, generates the event time label of the historical emergency events, extracts the event type of the historical emergency events from the historical description information, generates the event type label of the historical emergency events, extracts the three-break event identifiers of the historical emergency events from the historical description information, and generates the three-break event labels of the historical emergency events.

[0057] Furthermore, the event time label, event type label, three-break event label and historical description information of historical emergency events are preprocessed and a sample data set is constructed. The pre-set plan generation model is trained using the constructed sample data set to obtain a pre-trained plan generation model, through which an emergency flight plan can be generated online in real time based on the prediction results of emergency events in the emergency planning area.

[0058] Optionally, step 504 may further include: Step 601, based on a preset number of iterations, input the sample data set into a preset plan generation model, and perform iterative training on the plan generation model; Step 602, in each round of iteration, using the preset multi-class cross entropy as the loss function, adopting the preset gradient descent optimization algorithm, adjusting the model parameters of the plan generation model to optimize the function value of the loss function, and obtaining the pre-trained plan generation model when the function value of the loss function is optimal.

[0059] Based on a preset number of iterations, for example, 1500 (epochs=1500), the constructed sample data set is input into the preset plan generation model to iteratively train the plan generation model. In each round of iteration, the preset multi-class cross entropy is used as the loss function, and the preset gradient descent optimization algorithm is used to adjust the model parameters of the plan generation model to optimize the function value of the loss function. When the function value of the loss function reaches the optimal value, the pre-trained plan generation model is obtained.

[0060] Optionally, on the basis of the general large language model, a multi-label input layer, a word embedding layer, a fully connected layer, a discard layer and an output layer are added to construct a plan generation model, thereby improving the performance of the large language model in special task scenarios. During the model training process, the weight of the large language model is used as the initial weight of the model training, and only the fully connected layer, the discard layer and the output layer in the plan generation model are trained. That is, in each round of iteration, the preset multi-class cross entropy is used as the loss function, and the preset gradient descent optimization algorithm is used to adjust the weights of the fully connected layer, the discard layer and the output layer in the plan generation model to optimize the function value of the loss function. Among them, the gradient descent optimization algorithm is used to improve the learning speed of the traditional gradient descent. Through gradient descent, the model can find the optimal weight value that minimizes the loss function.

[0061] In this embodiment, the plan generation model can generate an emergency flight plan for the low-altitude airspace online in real time based on the prediction results of emergency events in the emergency planning area. It is accurate and efficient and can meet the demand for frequent applications for low-altitude emergency airspace.

[0062] In addition, based on the large language model, the plan generation model adds a multi-label input layer, word embedding layer, fully connected layer, discard layer and output layer, so that the large language model can better adapt to tasks in special scenarios, and can more accurately and efficiently generate emergency flight plans for low-altitude emergency airspace including flight time periods, flight areas and flight missions, so as to better ensure the smooth implementation of emergency rescue tasks.

[0063] The emergency flight plan generation device provided in an embodiment of the present application is described below. The emergency flight plan generation device described below and the emergency flight plan generation method described above can be referenced to each other.

[0064] Reference Figure 6 The emergency flight plan generation device provided in the embodiment of the present application includes: An event prediction module 10 is used to predict emergency events within the emergency planning area and obtain event description information of the emergency events; The label generation module 20 is used to generate a multi-dimensional label of the emergency event according to the event description information; the multi-dimensional label at least includes an event time label, an event type label and a three-break event label; A plan generation module 30, for inputting the multi-dimensional label and the event description information into a plan generation model, and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; The plan generation model is obtained by iterative training based on a sample data set, and the sample data set is constructed based on historical emergency events in the emergency planning area and historical flight plans corresponding to the historical emergency events.

[0065] In one embodiment, the label generation module 20 is further configured to: Inputting the event description information into an encoder of a tag generator; the encoder includes a first word embedding layer and a long short-term memory layer, and the tag generator also includes a decoder, and the decoder includes a first fully connected layer and a first output layer; Using the first word embedding layer in the encoder, converting the event description information into a word vector, and extracting a feature vector from the word vector through the long short-term memory layer; The feature vector is input into the first fully connected layer of the decoder for decoding, and label classification prediction is performed on the decoded feature vector based on the first output layer of the decoder to obtain a multi-dimensional label of the emergency event.

[0066] In one embodiment, the plan generation model includes a multi-label input layer, a second word embedding layer, a large language model layer, a second fully connected layer, a discard layer, and a second output layer; the plan generation module 30 is also used to: Inputting the multidimensional label and the event description information into a plan generation model, concatenating the multidimensional label and the event description information using the multi-label input layer of the plan generation model, and then inputting the concatenated labels into a second word embedding layer of the plan generation model; The concatenated multi-dimensional label and event description information are converted into a target word vector through the second word embedding layer; Using the large language model layer, the second fully connected layer and the discard layer of the planned generation model to perform feature extraction processing on the target word vector in sequence to obtain a target feature vector; The target feature vector is input into the second output layer of the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; the emergency flight plan includes a flight time period, a flight area and a flight mission.

[0067] In one embodiment, the plan generation module 30 is further used to: Obtaining the time granularity for generating an emergency flight plan in the emergency planning area; The multi-dimensional label and the event description information are input into a plan generation model, and the plan generation model is used to generate an emergency flight plan for the low-altitude area within the emergency planning area at the time granularity.

[0068] In one embodiment, the emergency flight plan generation device further includes a model pre-training module for: Obtain information on historical emergency events within the emergency planning area; Extracting historical description information and multi-dimensional information of the historical emergency event from the news information; the multi-dimensional information includes event time, event type and three-break event identifier; the three-break event identifier is used to indicate whether the historical emergency event is a three-break event; Acquire a historical emergency flight plan of the historical emergency event, use the multidimensional information as a multidimensional label of the historical description information, use the historical emergency flight plan as a prediction result of the historical description information, and construct a sample data set; The preset plan generation model is iteratively trained using the sample data set to obtain a pre-trained plan generation model.

[0069] In one embodiment, the model pre-training module is further used to: Based on a preset number of iterations, the sample data set is input into a preset plan generation model to perform iterative training on the plan generation model; In each round of iteration, the preset multi-class cross entropy is used as the loss function, and the preset gradient descent optimization algorithm is adopted to adjust the model parameters of the plan generation model to optimize the function value of the loss function, and when the function value of the loss function is optimal, the pre-trained plan generation model is obtained.

[0070] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, the electronic device may include: a processor 710, a communications interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communications interface 720 and the memory 730 communicate with each other via the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the emergency flight plan generation method, which includes: Predicting emergency events within the emergency planning area and obtaining event description information of the emergency events; Generate a multi-dimensional label of the emergency event according to the event description information; the multi-dimensional label at least includes an event time label, an event type label and a three-break event label; Inputting the multi-dimensional label and the event description information into a plan generation model, and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; The plan generation model is obtained by iterative training based on a sample data set, and the sample data set is constructed based on historical emergency events in the emergency planning area and historical flight plans corresponding to the historical emergency events.

[0071] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0072] On the other hand, an embodiment of the present application further provides a computer program product, the computer program product comprising a computer program, the computer program may be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer can execute the emergency flight plan generation method provided by the above methods, the method comprising: Predicting emergency events within the emergency planning area and obtaining event description information of the emergency events; Generate a multi-dimensional label of the emergency event according to the event description information; the multi-dimensional label at least includes an event time label, an event type label and a three-break event label; Inputting the multi-dimensional label and the event description information into a plan generation model, and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; The plan generation model is obtained by iterative training based on a sample data set, and the sample data set is constructed based on historical emergency events in the emergency planning area and historical flight plans corresponding to the historical emergency events.

[0073] In another aspect, an embodiment of the present application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the emergency flight plan generation method provided by the above methods is implemented, and the method includes: Predicting emergency events within the emergency planning area and obtaining event description information of the emergency events; Generate a multi-dimensional label of the emergency event according to the event description information; the multi-dimensional label at least includes an event time label, an event type label and a three-break event label; Inputting the multi-dimensional label and the event description information into a plan generation model, and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; The plan generation model is obtained by iterative training based on a sample data set, and the sample data set is constructed based on historical emergency events in the emergency planning area and historical flight plans corresponding to the historical emergency events.

[0074] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0075] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for generating an emergency flight plan, characterized in that: include: Predicting emergency events within the emergency planning area and obtaining event description information of the emergency events; Generate a multi-dimensional label for the emergency event according to the event description information; The multi-dimensional label at least includes an event time label, an event type label and a three-break event label; Inputting the multi-dimensional label and the event description information into a plan generation model, and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; The plan generation model is obtained by iterative training based on a sample data set, and the sample data set is constructed based on historical emergency events in the emergency planning area and historical flight plans corresponding to the historical emergency events.

2. The method for generating an emergency flight plan according to claim 1, characterized in that: Generating the multi-dimensional label of the emergency event according to the event description information includes: Inputting the event description information into an encoder of a tag generator; the encoder includes a first word embedding layer and a long short-term memory layer, and the tag generator also includes a decoder, and the decoder includes a first fully connected layer and a first output layer; Using the first word embedding layer in the encoder, converting the event description information into a word vector, and extracting a feature vector from the word vector through the long short-term memory layer; The feature vector is input into the first fully connected layer of the decoder for decoding, and label classification prediction is performed on the decoded feature vector based on the first output layer of the decoder to obtain a multi-dimensional label of the emergency event.

3. The method for generating an emergency flight plan according to claim 1, characterized in that: The plan generation model includes a multi-label input layer, a second word embedding layer, a large language model layer, a second fully connected layer, a discard layer, and a second output layer; the multi-dimensional label and the event description information are input into the plan generation model, and the plan generation model is used to generate an emergency flight plan for the low-altitude area within the emergency planning area, including: Inputting the multidimensional label and the event description information into a plan generation model, concatenating the multidimensional label and the event description information using the multi-label input layer of the plan generation model, and then inputting the concatenated labels into a second word embedding layer of the plan generation model; The concatenated multi-dimensional label and event description information are converted into a target word vector through the second word embedding layer; Using the large language model layer, the second fully connected layer and the discard layer of the planned generation model to perform feature extraction processing on the target word vector in sequence to obtain a target feature vector; The target feature vector is input into the second output layer of the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; the emergency flight plan includes a flight time period, a flight area and a flight mission.

4. The method for generating an emergency flight plan according to claim 1, characterized in that: The step of inputting the multi-dimensional label and the event description information into a plan generation model and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area includes: Obtaining the time granularity for generating an emergency flight plan in the emergency planning area; The multi-dimensional label and the event description information are input into a plan generation model, and the plan generation model is used to generate an emergency flight plan for the low-altitude area within the emergency planning area at the time granularity.

5. The method for generating an emergency flight plan according to claim 1, characterized in that: Before the multi-dimensional label and the event description information are input into the plan generation model and the emergency flight plan for the low-altitude area in the emergency planning area is generated by using the plan generation model, the method further includes: Obtain information on historical emergency events within the emergency planning area; Extracting historical description information and multi-dimensional information of the historical emergency event from the news information; the multi-dimensional information includes event time, event type and three-break event identifier; the three-break event identifier is used to indicate whether the historical emergency event is a three-break event; Acquire a historical emergency flight plan of the historical emergency event, use the multidimensional information as a multidimensional label of the historical description information, use the historical emergency flight plan as a prediction result of the historical description information, and construct a sample data set; The preset plan generation model is iteratively trained using the sample data set to obtain a pre-trained plan generation model.

6. The method for generating an emergency flight plan according to claim 5, characterized in that: The iterative training of the preset plan generation model using the sample data set to obtain a pre-trained plan generation model includes: Based on a preset number of iterations, the sample data set is input into a preset plan generation model, and the plan generation model is iteratively trained; In each round of iteration, the preset multi-class cross entropy is used as the loss function, and the preset gradient descent optimization algorithm is adopted to adjust the model parameters of the plan generation model to optimize the function value of the loss function, and when the function value of the loss function is optimal, the pre-trained plan generation model is obtained.

7. An emergency flight plan generating device, characterized in that: include: An event prediction module is used to predict emergency events within the emergency planning area and obtain event description information of the emergency events; A label generation module, used to generate a multi-dimensional label of the emergency event according to the event description information; The multi-dimensional label at least includes an event time label, an event type label and a three-break event label; A plan generation module, used for inputting the multi-dimensional label and the event description information into a plan generation model, and using the plan generation model to generate an emergency flight plan for the low-altitude area within the emergency planning area; The plan generation model is obtained by iterative training based on a sample data set, and the sample data set is constructed based on historical emergency events in the emergency planning area and historical flight plans corresponding to the historical emergency events.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the emergency flight plan generation method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the emergency flight plan generation method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the emergency flight plan generation method according to any one of claims 1 to 6 is implemented.

Images