Stratospheric airship long-period flight planning method, system, equipment and medium
Through the combination of hierarchical structure and deep neural network model, the efficiency and refinement of long-period flight planning of stratospheric airships is achieved, and the problem of difficulty in taking into account the refinement of planning results and the efficiency of long-term planning in the existing technology is solved.
Patent Information
- Application Number
- CN202510171725.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-17
- Publication Date
- 2025-06-06
AI Technical Summary
The existing flight planning methods are difficult to take into account the efficiency and refinement of long-term flight planning, especially in the ultra-long flight time of stratosphere airships, which are difficult to achieve accurate planning of flight trajectory and energy optimization.
Using a hierarchical structure flight planning method, by obtaining the system status and flight mission of the airship, first plan the top-level flight actions with a large time step, then carry out the underlying planning with a small time step, loop iteration until the end of the flight mission is reached, and the flight actions are evaluated through the deep neural network model to obtain the optimal planning.
The efficiency and refinement of long-period flight planning of stratospheric airships is achieved, and it can efficiently obtain feasible preliminary flight planning schemes based on environmental data such as wind farms, and improve the accuracy of flight planning results through fine planning.
Smart Images
Figure CN120106781A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of flight planning, and in particular relates to a method, system, equipment and medium for long-period flight planning of a stratospheric airship. Background Art
[0002] As a type of floating aircraft, the stratospheric airship relies on the buoyancy generated by a capsule filled with a large volume of light gas to maintain flight. It has the advantages of long flight time and high altitude, and can integrate multiple types of mission payloads to carry out applications such as remote sensing monitoring, communication support, and meteorological detection, and has broad application prospects.
[0003] The atmospheric flow in the altitude layer where stratospheric airships operate is dominated by horizontal winds whose speed and direction change with time. The horizontal wind has a significant effect on the wind pressure of the large volume of the airship capsule. At the same time, stratospheric airships generally rely on limited solar energy for power generation, and the drive motor drives the propeller to provide powered flight. Its flight speed is in the same order of magnitude as the stratospheric wind speed. Therefore, the actual flight trajectory achieved by the stratospheric airship is the result of the combined effect of the wind field and its own powered flight. The execution of stratospheric airship missions needs to take into account the impact of the wind field and plan a reasonable flight trajectory to achieve goals such as optimizing flight energy and maintaining precise position.
[0004] On the other hand, the continuous flight time of conventional aircraft is generally "hours" to "days", and appropriate mission windows can be selected to avoid various situations that are not suitable for flight. However, as an ultra-long-duration flight platform, the continuous operation time of stratospheric airships reaches "months" or even "years". Therefore, according to the characteristics of stratospheric airships, it is necessary to design a stratospheric airship flight planning method that can adapt to long-period flights, so as to provide the possibility for the efficient development of stratospheric airship applications. However, the existing flight planning methods mostly adopt a single-level structure with only one decision-making time step, which makes it difficult to take into account the refinement of planning results and the efficiency of long-term planning. Summary of the invention
[0005] The purpose of the present invention is to provide a method, system, device and medium for long-period flight planning of a stratospheric airship to solve the problems existing in the above-mentioned prior art.
[0006] To achieve the above object, the present invention provides a long-period flight planning method for a stratospheric airship, comprising:
[0007] Acquiring the system status and flight mission of the airship, wherein the system status includes the airship status, the mission status and the environment status;
[0008] Based on the starting point of the flight mission, the flight action is planned with a first time step to obtain a top-level planned action; based on the top-level planning result, the flight action is planned with a second time step to obtain a bottom-level planned action, and the second time step is executed cyclically until the time dimension of the first time step is reached; the end point of the bottom-level planning is used as a new starting point, and the top-level planning and bottom-level planning are iterated several times until the end point of the flight mission is reached;
[0009] The process of planning the flight action specifically includes: inputting the current system state of the airship into the flight action evaluation model for evaluation to obtain the optimal flight planning action; wherein the flight action evaluation model is constructed based on a deep neural network structure;
[0010] The results of previous bottom-level planning are spliced together to obtain the overall flight planning result.
[0011] Optionally, the first time step is greater than the second time step.
[0012] Optionally, the training process of the flight maneuver evaluation model specifically includes:
[0013] Construct an initial flight action evaluation model based on a deep neural network structure;
[0014] A simulation environment for stratospheric airship flight is constructed, in which flight planning decisions are executed, the loss function is calculated, and the model parameters are optimized by the gradient descent method to obtain a trained flight action evaluation model.
[0015] Optionally, executing the flight planning decision in the simulation environment specifically includes:
[0016] Input the current system state into the initial flight action evaluation model, output the evaluation value of each predicted action, select the predicted action corresponding to the maximum evaluation value as the flight execution action, and obtain the reward after executing the action and the system state at the next moment;
[0017] Calculate the loss function based on the system state at the current moment, the predicted action corresponding to the maximum evaluation value, the reward, and the system state at the next moment;
[0018] The model is optimized by reducing the loss function through the gradient descent method to obtain the trained flight action evaluation model.
[0019] Optionally, the calculation process of the evaluation value specifically includes:
[0020]
[0021] In the formula, Q* is the system state s at the current moment of the model t Take action a tThe optimal evaluation value of t is the reward for the current decision, s t+1 To evaluate the input of the model at the next moment, a t+1 Output the action decision of the unmanned airship at the next moment. For all possible s t+1 expectations.
[0022] A stratospheric airship long-period flight planning system, comprising:
[0023] A data acquisition module, used to obtain the system status and flight mission of the airship, wherein the system status includes the airship status, mission status and environment status;
[0024] A flight planning module is used to plan the flight action with a first time step according to the starting point of the flight mission to obtain a top-level planning action; based on the top-level planning result, plan the flight action with a second time step to obtain a bottom-level planning action, and execute the second time step in a loop until the time dimension of the first time step is reached; take the end point of the bottom-level planning as a new starting point, and iterate the top-level planning and bottom-level planning several times until the end point of the flight mission is reached;
[0025] The process of planning the flight action specifically includes: inputting the current system state of the airship into the flight action evaluation model for evaluation to obtain the optimal flight planning action; wherein the flight action evaluation model is constructed based on a deep neural network structure;
[0026] The results of previous bottom-level planning are spliced together to obtain the overall flight planning result.
[0027] An electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the long-period flight planning method for a stratospheric airship.
[0028] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for long-period flight planning of a stratospheric airship is implemented.
[0029] The technical effects of the present invention are:
[0030] The present invention adopts a hierarchical structure and can combine environmental data such as wind fields. First, a preliminary flight planning scheme is planned with a large time step to efficiently obtain a feasible preliminary flight planning scheme; then, based on the preliminary scheme, a detailed flight planning scheme is obtained with a small time step; finally, through splicing, a flight planning scheme with a long period and a small time step is obtained. The entire planning process takes into account both efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0032] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0033] Figure 1 This is a flight planning flowchart in an embodiment of the present invention. DETAILED DESCRIPTION
[0034] Various exemplary embodiments of the present invention will now be described in detail. This detailed description should not be considered as limiting the present invention, but should be understood as a more detailed description of certain aspects, features, and embodiments of the present invention.
[0035] It should be understood that the terms described in the present invention are only for describing special embodiments and are not intended to limit the present invention. In addition, for the numerical range in the present invention, it should be understood that each intermediate value between the upper and lower limits of the scope is also specifically disclosed. Each smaller range between the intermediate value in any stated value or stated range and any other stated value or intermediate value in the described range is also included in the present invention. The upper and lower limits of these smaller ranges can be independently included or excluded in the scope.
[0036] It will be apparent to those skilled in the art that various modifications and variations may be made to the specific embodiments of the present invention description without departing from the scope or spirit of the present invention. Other embodiments derived from the present invention description will be apparent to those skilled in the art. The present application description and examples are exemplary only.
[0037] The words “include,” “including,” “have,” “contain,” etc. used in this article are open-ended terms, meaning including but not limited to.
[0038] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0039] like Figure 1As shown, in this embodiment, a long-period flight planning method for a stratospheric airship is provided, including: obtaining the system state and flight mission of the airship, wherein the system state includes the airship state, the mission state and the environment state; planning the flight action with a first time step based on the starting point of the flight mission to obtain the top-level planning action; planning the flight action with a second time step based on the top-level planning result to obtain the bottom-level planning action, and cyclically executing the second time step until the time dimension of the first time step is reached; the first time step is generally several to dozens of times longer than the second time step; taking the end point of the bottom-level planning as a new starting point, iterating the top-level planning and the bottom-level planning for several times until the end point of the flight mission is reached; the process of planning the flight action specifically includes: inputting the current system state of the airship into the flight action evaluation model for evaluation to obtain the optimal flight planning action; wherein the flight action evaluation model is constructed based on a deep neural network structure; splicing the results of the previous bottom-level planning to obtain the overall flight planning result.
[0040] The specific implementation steps of this embodiment include:
[0041] Step 1: Clarify the starting and ending points of the flight mission, discretize the continuous flight process, represent the flight decision with the position S to be reached at each moment, and regard the flight planning process as two nested sequential decision processes, defined as the top-level planning process and the bottom-level planning process respectively.
[0042] Step 2: Carry out top-level planning. The top-level planning is for the entire long-term planning task, starting from the flight mission starting point (denoted as S 0 ) and arrive at the end of the task (denoted as S N ), with a larger time step, plan the flight action, first get the top-level flight planning decision S 1 .
[0043] Model application: Use a deep neural network model to build a flight evaluation intelligent decision-making model, input the current state into the model, and obtain the top-level planning action.
[0044] Step 3: Based on the top-level planning results, carry out the bottom-level flight planning with a smaller time step to obtain the 0 To S 1 The n-step detailed flight planning result is recorded as 0 ,S 0 1 ,S 0 2 …,S 0 n >.
[0045] Model application: The underlying flight planning also uses a deep neural network model to input the current state into the flight evaluation intelligent decision-making model to obtain the corresponding underlying planning actions.
[0046] Step 4: The end point S of the bottom-level planning in the previous step 0 n Instead of S 1 , carry out top-level planning, and obtain S 2 , and then carry out the bottom layer S 1 To S 2 Repeat steps 2 and 3, and so on, to get the entire flight planning result 0 ,S 0 1 ,S 0 2 ,…,S 0 n ,S 1 1 ,S 1 2 ,…,S 1 n ,…,S N-2 n ,S N-1 1 ,S N-1 2 ,…,S N-1 n >.
[0047] The sequential decision process in step 1 can be defined as a Markov decision process (MDP), that is, at a certain time t, based on the current system state s t , decide the action a that the current airship should perform t , when performing action a t After that, at time t+1, the system state changes to s t+1 , and based on s t+1 , based on the task constraints, we get the reward value r t Among them, the system state s is composed of the environmental state (including environmental wind speed, wind direction, local time, etc.), the airship state (including the remaining energy of the airship, the airship speed, the current coordinates, etc.) and the mission state (including the coordinates of the target point, etc.), and all state parameters are converted into 1-dimensional vectors; a is based on the dynamic model of the airship, and a feasible airship flight action is selected, and its elements include the flight speed and flight direction of the airship, etc.; r is an indicator for evaluating the quality of the airship flight action. The collective evaluation content includes whether the energy consumption of the airship is optimal and whether the flight action has a positive effect on the implementation of the mission.
[0048] The top-level planning in step 2 plans flight actions with a time step of 1 to 6 hours (adjustable), wherein the environmental state required for decision-making can be extracted from weather forecast data, and the airship state can be obtained through simulation based on airship characteristics or actual measurement by sensors.
[0049] The bottom-level planning in step 3 plans the flight action with a time step of 1 to 10 minutes (adjustable).
[0050] This embodiment takes into account the characteristics of long-duration flights of stratospheric airships. Existing methods cannot take into account both the refinement of planning results and the efficiency of long-term planning. Specifically, taking a three-day flight planning as an example, if the time step is 5 minutes, the total number of planning steps is 864, resulting in an overly long decision chain and inefficient planning results. If the step length is increased, although the decision chain can be shortened to improve planning efficiency, the decision interval is too large, and extensive planning will reduce the accuracy of flight planning results.
[0051] The specific process of flight action evaluation in the flight evaluation intelligent decision-making model: the quality of the flight action is evaluated by calculating the Q value, and the calculation method is as follows:
[0052]
[0053] Among them, γ is the attenuation factor, r t is the reward for the current decision, s t is the input of the intelligent decision-making model at the current moment, a t is the unmanned airship action decision output at the current moment, s t+1 is the input of the intelligent decision-making model at the next moment, a t+1 Output the action decision of the unmanned airship at the next moment. For all possible s t+1 expectations.
[0054] Flight planning: The flight planning action selection is selected through the deep neural network model, and the current state s t Input the deep neural network model to obtain the Q value of each possible action, and decide to select the action with the largest Q value to obtain the optimal flight plan.
[0055] Model construction: The deep neural network contains three fully connected layers, where the input layer dimension is the state s dimension, the output layer dimension is the action a dimension, and the activation function is the Relu function.
[0056] Model training: Model training is achieved through deep reinforcement learning methods. The model performs flight planning decisions in a simulation environment and calculates the loss function. The gradient descent method is used to reduce the loss function and optimize the model. The loss function calculation formula is as follows:
[0057]
[0058] The present invention adopts a hierarchical structure and can combine environmental data such as wind fields. First, a preliminary flight planning scheme is planned with a large time step to efficiently obtain a feasible preliminary flight planning scheme; then, based on the preliminary scheme, a detailed flight planning scheme is obtained with a small time step; finally, through splicing, a flight planning scheme with a long period and a small time step is obtained. The entire planning process takes into account both efficiency and accuracy.
[0059] A stratospheric airship long-period flight planning system, comprising:
[0060] A data acquisition module, used to obtain the system status and flight mission of the airship, wherein the system status includes the airship status, mission status and environment status;
[0061] A flight planning module is used to plan the flight action with a first time step according to the starting point of the flight mission to obtain a top-level planning action; based on the top-level planning result, plan the flight action with a second time step to obtain a bottom-level planning action, and execute the second time step in a loop until the time dimension of the first time step is reached; take the end point of the bottom-level planning as a new starting point, and iterate the top-level planning and the bottom-level planning until the end point of the flight mission is reached;
[0062] The process of planning the flight action specifically includes: inputting the current system state of the airship into the flight action evaluation model for evaluation and classification to obtain the optimal flight planning action; wherein the flight action evaluation model is constructed based on a deep neural network structure;
[0063] The results of previous bottom-level planning are spliced together to obtain the overall flight planning result.
[0064] An electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the long-period flight planning method for a stratospheric airship.
[0065] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for long-period flight planning of a stratospheric airship is implemented.
[0066] The above is only a preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. A long-period flight planning method for a stratospheric airship, characterized in that: include: Acquiring the system status and flight mission of the airship, wherein the system status includes the airship status, the mission status and the environment status; Planning a flight action based on a starting point of the flight mission with a first time step to obtain a top-level planned action; Based on the top-level planning result, the flight action is planned with the second time step, the bottom-level planning action is obtained, and the second time step is executed repeatedly until the time dimension of the first time step is reached; Take the end point of the bottom-level planning as the new starting point, and iterate the top-level planning and bottom-level planning several times until the end point of the flight mission is reached; The process of planning the flight action specifically includes: inputting the current system state of the airship into the flight action evaluation model for evaluation to obtain the optimal flight planning action; wherein the flight action evaluation model is constructed based on a deep neural network structure; The results of previous bottom-level planning are spliced together to obtain the overall flight planning result.
2. A stratospheric airship long-period flight planning method according to claim 1, characterized in that: The first time step is greater than the second time step.
3. A stratospheric airship long-period flight planning method according to claim 1, characterized in that: The training process of the flight action evaluation model specifically includes: Construct an initial flight action evaluation model based on a deep neural network structure; A simulation environment for stratospheric airship flight is constructed, in which flight planning decisions are executed, the loss function is calculated, and the model parameters are optimized by the gradient descent method to obtain a trained flight action evaluation model.
4. A stratospheric airship long-period flight planning method according to claim 3, characterized in that: The executing flight planning decision in the simulation environment specifically includes: Input the current system state into the initial flight action evaluation model, output the evaluation value of each predicted action, select the predicted action corresponding to the maximum evaluation value as the flight execution action, and obtain the reward after executing the action and the system state at the next moment; Calculate the loss function based on the system state at the current moment, the predicted action corresponding to the maximum evaluation value, the reward, and the system state at the next moment; The model is optimized by reducing the loss function through the gradient descent method to obtain the trained flight action evaluation model.
5. A stratospheric airship long-period flight planning method according to claim 4, characterized in that: The calculation process of the evaluation value specifically includes: In the formula, Q* is the system state s at the current moment of the model t Take action a t The optimal evaluation value of t is the reward for the current decision, s t+1 To evaluate the input of the model at the next moment, a t+1 Output the action decision of the unmanned airship at the next moment. For all possible s t+1 expectations.
6. A stratospheric airship long-period flight planning system, characterized in that: include: A data acquisition module, used to obtain the system status and flight mission of the airship, wherein the system status includes the airship status, mission status and environment status; A flight planning module, used for planning a flight action at a first time step according to a starting point of the flight mission to obtain a top-level planned action; Based on the top-level planning result, the flight action is planned with the second time step, the bottom-level planning action is obtained, and the second time step is executed repeatedly until the time dimension of the first time step is reached; Take the end point of the bottom-level planning as the new starting point, and iterate the top-level planning and bottom-level planning several times until the end point of the flight mission is reached; The process of planning the flight action specifically includes: inputting the current system state of the airship into the flight action evaluation model for evaluation to obtain the optimal flight planning action; wherein the flight action evaluation model is constructed based on a deep neural network structure; The results of previous bottom-level planning are spliced together to obtain the overall flight planning result.
7. An electronic device, characterized in that: The invention comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a stratospheric airship long-period flight planning method according to any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that: The computer program is stored therein, and when the computer program is executed by a processor, a long-period flight planning method for a stratospheric airship as claimed in any one of claims 1 to 5 is implemented.