A near space unmanned aerial vehicle launching method and system

Abstract

The application relates to the field of unmanned aerial vehicle launching, and proposes a near-space unmanned aerial vehicle launching method and system, which comprises the following steps: constructing an environment prediction and state estimation model, setting initial parameters of an ant colony algorithm, taking a starting position and a target position of an unmanned aerial vehicle as initial inputs of an optimization problem, and optimizing a flight track of the unmanned aerial vehicle based on the ant colony algorithm; dynamically adjusting pheromone update rules and transition probabilities in the ant colony algorithm through an environment prediction result, performing multi-objective optimization calculation on flight safety, energy consumption, task completion time and economic cost, and outputting a flight track of the unmanned aerial vehicle and a launching scheme of a carrier vehicle; launching the unmanned aerial vehicle to a specified position in the near space based on the launching scheme, and analyzing real-time environment data obtained in the flight process by using the environment prediction and state estimation model, so that the flight track of the unmanned aerial vehicle can be adjusted in real time. The application improves launching precision and task adaptability of the near-space unmanned aerial vehicle in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of drone deployment, and more particularly to a method and system for deploying drones in near space. Background Technology

[0002] Near space refers to the airspace region at an altitude of approximately 20 to 100 kilometers above the Earth's surface, situated between the upper limit of traditional aviation flight and the lower limit of spaceflight. Due to its unique atmospheric characteristics and strategic location, this region has extremely high application value in fields such as military reconnaissance, communication relay, environmental monitoring, and emergency rescue. Near-space UAV deployment technology refers to a systematic engineering project that uses various carriers, such as rockets, balloons, and high-altitude aircraft, to transport specialized UAVs to a designated altitude in near space and release them to perform predetermined missions. Compared to traditional airspace UAVs, near-space UAVs have a wider ground coverage area, longer loiter time, and more covert operational capabilities.

[0003] In existing technologies, trajectory planning for near-space UAV deployments mainly employs traditional deterministic planning methods, such as rule-based fixed trajectory design, simple gradient descent optimization, or pre-defined deployment procedures. These methods have significant shortcomings in data processing and decision optimization, typically relying on static meteorological data and simplified environmental models. They lack the ability to accurately model and predict the complex and variable environment of near-space, resulting in significant environmental uncertainties during actual deployments. This leads to low success rates, poor mission adaptability, and inefficient resource utilization for near-space UAV deployments, particularly under complex meteorological conditions and in scenarios requiring high-precision deployment. Summary of the Invention

[0004] In view of this, the present invention proposes a method and system for deploying near-space unmanned aerial vehicles (UAVs) to address the problems of low success rate, poor mission adaptability, and low resource utilization efficiency in the actual deployment process of existing technologies due to the significant environmental uncertainty risks.

[0005] The technical solution of the present invention is implemented as follows: In a first aspect, the present invention provides a method for deploying near-space unmanned aerial vehicles, comprising the following steps: Historical flight data, near-space meteorological data, and real-time sensor data from vehicles and drones are acquired, and an environmental prediction and state estimation model is constructed based on a deep learning network. The environmental prediction and state estimation model outputs environmental prediction results. Based on the requirements of the drone deployment mission, the initial parameters of the ant colony algorithm are set, including the initial value of pheromone, the number of ants, and the candidate set of flight trajectories. The starting position and target position of the drone are used as the initial input of the optimization problem, and the drone flight trajectory is optimized based on the ant colony algorithm. By dynamically adjusting the pheromone update rules and transition probabilities in the ant colony algorithm based on environmental prediction results, multi-objective optimization calculations are performed on flight safety, energy consumption, mission completion time and economic cost, and the UAV flight trajectory and vehicle deployment scheme are output. Based on the deployment plan, the drone is deployed to a designated location in the near space, and during the flight, the acquired real-time environmental data is analyzed using environmental prediction and state estimation models to adjust the drone's flight trajectory in real time.

[0006] Based on the above technical solutions, preferably, the acquisition of historical flight data, near-space meteorological data, and real-time sensor data from the vehicle and the UAV, and the construction of an environment prediction and state estimation model based on a deep learning network, wherein the environment prediction and state estimation model outputs environment prediction results, including: Historical flight data, near-space meteorological data, and real-time sensor data of vehicles and drones are obtained from multi-source databases. Time alignment, noise reduction, normalization, and feature extraction are performed on the historical flight data, near-space meteorological data, and real-time sensor data of vehicles and drones to obtain a standardized training dataset. The original environment prediction and state estimation model is constructed based on a deep learning network that combines a bidirectional long short-term memory network with an attention mechanism. A hierarchical training strategy is adopted to train the model on the standardized training dataset to obtain an environment prediction and state estimation model for predicting near-space environment features and estimating the state of UAVs and vehicles.

[0007] Based on the above technical solutions, preferably, the initial parameters of the ant colony algorithm are set according to the requirements of the drone deployment task, including the initial pheromone value, the number of ants, and the candidate set of flight trajectories. The starting position and target position of the drone are used as the initial inputs to the optimization problem. The drone flight trajectory is optimized based on the ant colony algorithm, including: Based on the requirements of the drone delivery mission, including operating altitude, mission priority, weather conditions and energy limitations, the ant colony algorithm parameter matrix is ​​determined, and the initial value of pheromone, volatility coefficient, number of ants and number of iterations are dynamically configured. The initial parameters of the ant colony algorithm are adjusted according to the priority of task requirements, and a multi-level task priority evaluation system is established, which quantifies and assigns weights to four dimensions: safety, timeliness, energy efficiency, and economy. The initial value of the pheromone is set based on the historical deployment experience database, the volatility coefficient is dynamically adjusted according to the rate of change of the nearby spatial environment, the number of ants is positively correlated with the task complexity and accuracy requirements, and the number of iterations is determined by the dynamic convergence judgment criterion.

[0008] Based on the above technical solutions, preferably, the adjustment formula for the initial parameters is as follows: ; ; ; in, For the first k Initial pheromone values ​​for task-like tasks As the baseline pheromone value, It is an exponential function. This is a task priority vector. For the work height factor, As an energy constraint factor, , , These are the weighting coefficients for task priority, operation altitude, and energy constraints, respectively. for t The pheromone evaporation coefficient at any given time. and These are the maximum and minimum limits for pheromone evaporation coefficients, respectively. For the gradient of environmental change, For environmental sensitivity parameters, For the first k The number of ants for each task. As a metric for task complexity, For the accuracy requirement coefficient, and These are the first adjustment parameter and the second adjustment parameter, respectively. This is the floor function.

[0009] Based on the above technical solutions, preferably, the step of dynamically adjusting the pheromone update rules and transition probabilities in the ant colony algorithm based on environmental prediction results, performing multi-objective optimization calculations on flight safety, energy consumption, mission completion time, and economic costs, and outputting the UAV flight trajectory and vehicle deployment scheme includes: Using a pre-built set of candidate flight trajectories, multi-objective fitness calculations are performed on each candidate trajectory based on flight safety, energy consumption, mission completion time, and economic cost of the UAV delivery mission, thereby generating a preliminary set of excellent candidate trajectories. Based on the fitness evaluation results of candidate trajectories, and combined with real-time environmental prediction data, the pheromone update rules and trajectory transition probabilities in the ant colony algorithm are dynamically adjusted to complete the iterative optimization of candidate trajectories and output the UAV flight trajectory and the delivery plan of the carrier.

[0010] Based on the above technical solutions, preferably, the formula for calculating the multi-objective fitness is: ; in, For multi-objective fitness functions, Candidate trajectories, , , and These are evaluation functions for flight safety, mission completion time, energy consumption, and economic cost, respectively. , , and These are the dynamically adjusted weighting coefficients for the safety, time, energy, and cost evaluation functions, respectively. As a safety volatility factor, n This represents the current iteration number. T This represents the total number of iterations. The formula for calculating the pheromone update rule is: ; in, for The updated value, From trajectory nodes i To trajectory node j pheromone concentration value, The pheromone evaporation coefficient, To increase the proportion of pheromones, The fitness evaluation value of the candidate trajectory. To prevent zero constant.

[0011] Based on the above technical solutions, preferably, the step of deploying the drone to a designated location in near space based on the deployment scheme, and analyzing the acquired real-time environmental data using an environmental prediction and state estimation model during flight to adjust the drone's flight trajectory in real time, includes: Based on the drone's flight trajectory and deployment plan, control the transport vehicle to reach the specified deployment conditions, activate the deployment mechanism to release the drone to the designated location in the near space, and perform initial attitude stabilization control after the drone separates. After the UAV is deployed to a designated location in near space, it continuously acquires and processes data from the UAV's onboard sensors. It then uses environmental prediction and state estimation models to analyze near-space weather conditions in real time and dynamically adjusts flight control parameters and trajectory planning to ensure that the UAV reaches the mission target area.

[0012] Secondly, the present invention also provides a near-space unmanned aerial vehicle (UAV) deployment system, the system comprising: The data acquisition and modeling module is used to acquire historical flight data, near-space meteorological data, and real-time sensor data of vehicles and drones, and to build an environment prediction and state estimation model based on a deep learning network. The environment prediction and state estimation model outputs the environment prediction results. The ant colony algorithm initialization module is used to set the initial parameters of the ant colony algorithm according to the requirements of the drone deployment task, including the initial value of pheromone, the number of ants and the candidate set of flight trajectories, and to use the starting position and target position of the drone as the initial input of the optimization problem, and to optimize the drone flight trajectory based on the ant colony algorithm. The trajectory optimization and iteration module is used to dynamically adjust the pheromone update rules and transition probabilities in the ant colony algorithm based on environmental prediction results. It performs multi-objective optimization calculations on flight safety, energy consumption, mission completion time and economic cost, and outputs the UAV flight trajectory and the delivery plan of the carrier vehicle. The deployment execution and adjustment module is used to deploy the drone to a designated location in the near space based on the deployment plan, and to analyze the acquired real-time environmental data using environmental prediction and state estimation models during flight in order to adjust the drone's flight trajectory in real time.

[0013] Thirdly, the present invention also provides an electronic device, comprising: at least one processor, at least one memory, a communication interface, and a bus; The processor, memory, and communication interface communicate with each other through the bus. The memory stores program instructions that can be executed by the processor. The processor calls the program instructions to implement the steps of a near-space UAV deployment method.

[0014] Fourthly, the present invention also provides a computer-readable storage medium storing computer instructions that enable a computer to perform steps such as a near-space unmanned aerial vehicle (UAV) deployment method.

[0015] The near-space drone deployment method and system of the present invention have the following advantages over the prior art: (1) By constructing an environmental prediction and state estimation model to improve environmental adaptability, and combining the mechanism of dynamically adjusting the pheromone update rules and transition probability of the ant colony algorithm, and using real-time environmental data feedback to correct the trajectory, the multi-objective collaborative optimization of UAV deployment flight safety, energy efficiency, mission timeliness and economic cost is realized, enabling UAV deployment to have accurate environmental perception, dynamic path optimization and online adjustment capabilities, and greatly improving the deployment accuracy, safety and reliability and mission adaptability of near-space UAVs in complex environments; (2) By introducing a dynamically adjusted ant colony algorithm parameter matrix, and combining key constraints of near-space deployment such as task priority, operation altitude, energy limitation and weather conditions, the initial value of pheromone, volatility coefficient, number of ants and number of iterations are dynamically set, and a comprehensive task priority evaluation system and parameter self-correction mechanism are established. This enables personalized optimization configuration for different deployment task requirements, and realizes rapid convergence of deployment path planning and efficient optimization of global optimal solution, thereby significantly improving the trajectory optimization accuracy, deployment task adaptability and resource utilization efficiency of UAVs in complex near-space environments. (3) By performing multi-target fitness calculation on the flight trajectory of the UAV, the four key indicators of flight safety, mission completion time, energy consumption and economic cost are comprehensively evaluated. Dynamic adjustment weight coefficients and safety fluctuation factors are introduced to enable trajectory optimization to adapt to mission requirements and environmental changes, thereby generating a more efficient, safe and economical deployment plan. Combined with real-time environmental prediction data, the pheromone update rules and trajectory transfer probability in the ant colony algorithm are iteratively optimized and dynamically correlated with the fitness of candidate trajectories, thus realizing the rapid convergence of the global optimal solution of trajectory planning. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a near-space drone deployment method according to the present invention; Figure 2 This is a structural diagram of a near-space unmanned aerial vehicle (UAV) delivery system according to the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 This invention provides a method for deploying near-space unmanned aerial vehicles (UAVs), comprising the following steps: Historical flight data, near-space meteorological data, and real-time sensor data from vehicles and drones are acquired, and an environmental prediction and state estimation model is constructed based on a deep learning network. The environmental prediction and state estimation model outputs environmental prediction results. Based on the requirements of the drone deployment mission, the initial parameters of the ant colony algorithm are set, including the initial value of pheromone, the number of ants, and the candidate set of flight trajectories. The starting position and target position of the drone are used as the initial input of the optimization problem, and the drone flight trajectory is optimized based on the ant colony algorithm. By dynamically adjusting the pheromone update rules and transition probabilities in the ant colony algorithm based on environmental prediction results, multi-objective optimization calculations are performed on flight safety, energy consumption, mission completion time and economic cost, and the UAV flight trajectory and vehicle deployment scheme are output. Based on the deployment plan, the drone is deployed to a designated location in the near space, and during the flight, the acquired real-time environmental data is analyzed using environmental prediction and state estimation models to adjust the drone's flight trajectory in real time.

[0020] Specifically, this embodiment improves environmental adaptability by constructing an environmental prediction and state estimation model, and combines a mechanism for dynamically adjusting the pheromone update rules and transition probabilities of the ant colony algorithm. At the same time, it uses real-time environmental data feedback to correct the trajectory, achieving multi-objective collaborative optimization of UAV deployment flight safety, energy efficiency, mission timeliness, and economic cost. This enables UAV deployment to have accurate environmental perception, dynamic path optimization, and online adjustment capabilities, significantly improving the deployment accuracy, safety, reliability, and mission adaptability of near-space UAVs in complex environments.

[0021] The process involves acquiring historical flight data, near-space meteorological data, and real-time sensor data from vehicles and drones, and constructing an environmental prediction and state estimation model based on a deep learning network. The environmental prediction and state estimation model outputs environmental prediction results, including: Historical flight data, near-space meteorological data, and real-time sensor data from vehicles and drones are obtained from multi-source databases. Time alignment, noise reduction, normalization, and feature extraction are performed on the historical flight data, near-space meteorological data, and real-time sensor data from vehicles and drones to obtain a standardized training dataset.

[0022] In one specific embodiment, a time series smoothing and outlier detection algorithm is used. Historical flight data and near-space meteorological data are smoothed by sliding window convolutional filtering. At the same time, Mahalanobis distance is applied to calculate outliers and outliers are replaced with local interpolation results. For feature extraction of multi-source data, principal component analysis is used to reduce data dimensionality. Normalization is used to map different types of environmental data to a unified feature space to improve the accuracy of environmental prediction.

[0023] The original environment prediction and state estimation model is constructed based on a deep learning network that combines a bidirectional long short-term memory network with an attention mechanism. A hierarchical training strategy is adopted to train the model on the standardized training dataset to obtain an environment prediction and state estimation model for predicting near-space environment features and estimating the state of UAVs and vehicles.

[0024] In one specific embodiment, the environment prediction and state estimation model adopts a dual-path network architecture. The environment prediction path consists of stacked bidirectional long short-term memory network layers to capture the temporal features of the near-space environment. The state estimation path consists of a multilayer perceptron and a self-attention mechanism to estimate the real-time state of the UAV and the vehicle. The two paths achieve information fusion through a shared attention layer and improve network stability through residual connections. The training process adopts batch normalization and dynamic learning rate strategies to improve the model's generalization ability under complex weather conditions. The calculation formula for the time series smoothing and outlier detection algorithm is as follows: ; in, for t The raw data at that moment for t The original data at time -1 for t Time-smoothed data, for t- Smoothed data at time 1 As the first dynamic smoothing factor, It is the second dynamic smoothing factor; The output of the environmental prediction and state estimation model is: ; ; ; in, The output of the environmental prediction and state estimation model includes environmental prediction results and state estimates. For dynamic fusion coefficients, As an interaction enhancement factor, for The environment prediction vector at time t, for t The state estimation vector at time t, For activation function, For environmental prediction weight matrix, This represents the forward hidden state of a bidirectional long short-term memory network. This represents the backward hidden state of a bidirectional long short-term memory network. for t The input feature vector at time t, This is a vector concatenation symbol. This is the environmental path attention weight matrix. This is the environmental prediction bias vector. For the multilayer perceptron mapping function, This is the output of the state path self-attention mechanism. The symbol for element-wise product. For feature extraction function, For residual connection, The weight matrix is ​​for state estimation. This is the state estimation bias vector.

[0025] Specifically, this embodiment introduces a deep learning network to construct an environmental prediction and state estimation model. It efficiently fuses historical flight data, near-space meteorological data, and real-time sensor data, outputting high-precision environmental prediction results. Compared to traditional rule-based or static model-based prediction methods, deep learning networks possess stronger nonlinear learning capabilities and data adaptability. They can accurately capture the dynamic changes in complex near-space meteorological conditions and the state characteristics of the vehicle and the UAV itself, thereby significantly improving the prediction accuracy of environmental disturbances and state changes. This provides more reliable basic data support for parameter optimization and trajectory planning in ant colony algorithms, while also enhancing the environmental adaptability of the UAV during deployment, improving deployment safety and mission execution success rate.

[0026] The process involves setting initial parameters for the ant colony algorithm based on the drone deployment task requirements, including initial pheromone values, ant count, and a candidate flight trajectory set. The drone's starting and target positions are used as initial inputs to the optimization problem. The drone's flight trajectory is then optimized using the ant colony algorithm, including: Based on the requirements of the drone delivery mission, including operating altitude, mission priority, weather conditions and energy limitations, the ant colony algorithm parameter matrix is ​​determined, and the initial value of pheromone, volatility coefficient, number of ants and number of iterations are dynamically configured. The initial parameters of the ant colony algorithm are adjusted according to the priority of task requirements, and a multi-level task priority evaluation system is established, which quantifies and weights four dimensions: safety, timeliness, energy efficiency, and economy. The initial pheromone value is set based on a historical deployment experience database, the volatile coefficient is dynamically adjusted according to the rate of change of the nearby spatial environment, the number of ants is positively correlated with the task complexity and accuracy requirements, and the number of iterations is determined by a dynamic convergence criterion. At the same time, a parameter self-correction mechanism is applied to fine-tune the parameter matrix online based on the initial iteration results, thereby improving the adaptability of parameter configuration.

[0027] Specifically, this embodiment introduces a dynamically adjusted ant colony algorithm parameter matrix, combined with key constraints of near-space deployment such as task priority, operating altitude, energy limitations, and weather conditions, dynamically sets the initial value of pheromones, volatility coefficient, number of ants, and number of iterations, and establishes a comprehensive task priority evaluation system and parameter self-correction mechanism. It can achieve personalized optimization configuration for different deployment task requirements, realize rapid convergence of deployment path planning and efficient optimization of the global optimal solution, thereby significantly improving the trajectory optimization accuracy, deployment task adaptability, and resource utilization efficiency of UAVs in complex near-space environments.

[0028] The formula for adjusting the initial parameters is: ; ; ; in, For the first k Initial pheromone values ​​for task-like tasks As the baseline pheromone value, It is an exponential function. This is a task priority vector. For the work height factor, As an energy constraint factor, , , These are the weighting coefficients for task priority, operation altitude, and energy constraints, respectively. for t The pheromone evaporation coefficient at any given time. and These represent the maximum and minimum limits of pheromone evaporation coefficients, respectively. For the gradient of environmental change, For environmental sensitivity parameters, For the first k The number of ants for each task. As a metric for task complexity, For the accuracy requirement coefficient, and These are the first adjustment parameter and the second adjustment parameter, respectively. This is the floor function.

[0029] Specifically, this embodiment uses an exponential function structure to nonlinearly map three key factors—task priority, task height, and energy constraints—achieving a precise match between pheromone intensity and task requirements. This allows high-priority tasks, tasks with specific heights, or tasks with strict energy constraints to obtain higher initial pheromone concentrations, thereby guiding the ants (i.e., the search algorithm) to explore in a direction that meets the key task requirements from the early stages of path search, greatly improving the convergence speed and solution quality of the algorithm.

[0030] This embodiment compares the volatility coefficient with the environmental change gradient. Dynamic correlation: When environmental changes are drastic (large gradient value), the volatility coefficient tends to... More information is retained to maintain algorithm stability; when the environment is relatively stable (gradient value is small), the volatile coefficient tends to be... This accelerates pheromone evaporation to avoid the algorithm getting trapped in local optima, enabling it to maintain high adaptability in complex and variable environments in the near space, converge quickly when the environment is stable, and maintain sufficient search diversity when the environment fluctuates.

[0031] This embodiment establishes a quantitative relationship between the number of ants and the complexity and accuracy requirements of the task. By using an up-rounding function, it ensures that the calculation results conform to the discrete characteristics of the number of ants, ensuring that more computing resources (more ants) can be allocated to complex tasks and high-precision requirements, thereby increasing the breadth and depth of path search and ultimately improving the quality of the solution.

[0032] The method dynamically adjusts the pheromone update rules and transition probabilities in the ant colony algorithm based on environmental prediction results, performs multi-objective optimization calculations on flight safety, energy consumption, mission completion time, and economic costs, and outputs the UAV flight trajectory and vehicle deployment scheme, including: Using a pre-built set of candidate flight trajectories, multi-objective fitness calculations are performed on each candidate trajectory based on flight safety, energy consumption, mission completion time, and economic cost of the UAV delivery mission, thereby generating a preliminary set of excellent candidate trajectories. Based on the fitness evaluation results of candidate trajectories, and combined with real-time environmental prediction data, the pheromone update rules and trajectory transition probabilities in the ant colony algorithm are dynamically adjusted to complete the iterative optimization of candidate trajectories and output the UAV flight trajectory and the delivery plan of the carrier.

[0033] Specifically, this embodiment calculates the multi-target fitness of the UAV flight trajectory, comprehensively evaluates four key indicators: flight safety, mission completion time, energy consumption, and economic cost, and introduces dynamically adjusted weight coefficients and safety fluctuation factors to enable trajectory optimization to adaptively adjust according to mission requirements and environmental changes, thereby generating a more efficient, safe, and economical deployment plan. Combined with real-time environmental prediction data, the pheromone update rules and trajectory transition probabilities in the ant colony algorithm are iteratively optimized and dynamically correlated with the fitness of candidate trajectories, achieving rapid convergence of the global optimal solution for trajectory planning.

[0034] The formula for calculating the multi-objective fitness is: ; in, For multi-objective fitness functions, Candidate trajectories, , , and These are evaluation functions for flight safety, mission completion time, energy consumption, and economic cost, respectively. , , and These are the dynamically adjusted weighting coefficients for the safety, time, energy, and cost evaluation functions, respectively. As a safety volatility factor, n This represents the current iteration number. T This represents the total number of iterations. The formula for calculating the pheromone update rule is: ; in, for The updated value, From trajectory nodes i To trajectory node j pheromone concentration value, The pheromone evaporation coefficient, To increase the proportion of pheromones, The fitness evaluation value of the candidate trajectory. To prevent zero constant.

[0035] Specifically, this embodiment comprehensively evaluates the flight path through multi-objective fitness calculation, dynamically balancing multiple objectives such as safety, mission completion time, energy consumption, and economic cost. In particular, it incorporates a safety fluctuation factor. This further strengthens the priority of flight safety in trajectory planning and ensures the reliable operation of UAVs in complex environments.

[0036] The dynamically adjustable pheromone update formula improves the convergence efficiency of the ant colony algorithm in global path optimization. By incorporating key parameters such as environmental change rate and fitness score into the volatile coefficient and update rule, the algorithm's adaptability in complex near-space environments is enhanced, avoiding getting trapped in local optima, while taking into account both real-time performance and globality.

[0037] Overall, this embodiment, through the synergistic effect of multi-target fitness evaluation and dynamic pheromone adjustment mechanism, can quickly generate optimal UAV deployment schemes that balance safety, efficiency, and economy. While improving deployment success rate and path accuracy, it achieves adaptive optimization for complex mission requirements. This provides a more flexible, efficient, and reliable deployment solution for near-space UAV mission execution.

[0038] The method involves deploying the drone to a designated location in near space based on a deployment plan, and analyzing the acquired real-time environmental data using environmental prediction and state estimation models during flight to adjust the drone's flight trajectory in real time. This includes: Based on the drone's flight trajectory and deployment plan, control the transport vehicle to reach the specified deployment conditions (including altitude, speed, attitude, and time window), activate the deployment mechanism to release the drone to the designated location in the near space, and perform initial attitude stabilization control after the drone separates. In one specific embodiment, the control vehicle accurately calculates the optimal delivery time window and delivery point location based on weather conditions, mission urgency, and target arrival accuracy requirements, and employs a multi-level safety release mechanism to execute the UAV delivery. During the delivery process, an adaptive separation impact mitigation mechanism is introduced, including an elastic separation system and an aerodynamic deceleration device, to reduce interference at the moment of delivery and protect the UAV equipment. After the UAV separates from the vehicle, a cascaded attitude controller achieves rapid stabilization of the initial attitude. This controller comprehensively considers factors such as the characteristics of the thin atmosphere in near space, dynamic pressure changes, and weakened aerodynamic forces, and adopts a predictive-correction-compensation control strategy to ensure that the UAV can reach a stable state and execute the predetermined trajectory flight within 3 seconds after separation.

[0039] After the UAV is deployed to a designated location in near space, it continuously acquires and processes data from the UAV's onboard sensors. It then uses environmental prediction and state estimation models to analyze near-space meteorological conditions (such as wind shear, temperature gradient, and air pressure changes) in real time, dynamically adjusting flight control parameters and trajectory planning to ensure that the UAV reaches the mission target area safely and efficiently.

[0040] In one specific embodiment, a hierarchical perception processing architecture is employed to acquire and fuse multi-sensor data, including output information from inertial measurement units, meteorological sensors, altimeters, and vision systems. The fused sensor data is then input into an environmental prediction and state estimation model to generate a local dynamic map of the near-space environment and a UAV state assessment report. Based on the dynamic environmental map and state assessment report, a feasible flight area containing a safety corridor is constructed, and local trajectory optimization is performed within this area. When sudden environmental changes or potential hazards are detected, emergency avoidance strategies are automatically triggered, including rapid turns, deceleration to avoid obstacles, or temporary hovering maneuvers. Simultaneously, a continuous data link is maintained with ground or satellite communication systems to receive mission command updates and wide-area environmental information, further improving the reliability of trajectory adjustments and mission adaptability.

[0041] Specifically, this embodiment establishes a systematic and multi-layered near-space UAV deployment control method, optimizing the entire deployment process from release preparation and separation execution to subsequent flight. By accurately calculating the deployment time window and location, combined with a multi-level safety release mechanism and an adaptive separation impact mitigation mechanism, this embodiment significantly improves deployment accuracy and reliability. Simultaneously, it introduces a cascaded attitude controller to achieve rapid attitude stabilization within 3 seconds after separation, effectively addressing the unique challenges of the thin atmosphere in near-space.

[0042] In the post-deployment phase, a hierarchical perception and processing architecture and multi-sensor fusion technology are used to construct a dynamic environmental map, enabling local trajectory optimization and emergency avoidance capabilities based on safety corridors. By maintaining data links with ground or satellite systems, it is ensured that the UAV can adapt to complex weather changes, dynamically adjust flight parameters, and maintain efficient mission execution and safety assurance even in extreme environments.

[0043] In one specific embodiment, the launch vehicle is a rocket, serving as the primary launch platform for the drone. The rocket carries the drone, and according to the launch time, launch altitude, attitude, and other parameters output by the trajectory optimization module, the drone is transported to a designated altitude in near space. Once the rocket reaches the preset launch point, the launch execution and adjustment module controls the rocket's launch bay to open or uses a dedicated release mechanism to safely release the drone into the target airspace.

[0044] During rocket launch, this embodiment further considers the significant environmental disturbances during high-speed rocket flight. To address this, flexible vibration reduction and adaptive separation impact mitigation devices are employed to ensure the structural integrity of the UAV at the moment of rocket separation. Release also includes precise measurement and control of the UAV's initial attitude and velocity. Combining environmental prediction and state estimation models, the instantaneous parameters of the UAV separation (such as separation velocity and attitude angle) are adjusted in real time based on the actual rocket flight status and external atmospheric environment to ensure the UAV smoothly enters the autonomous flight and trajectory stabilization phase.

[0045] After the rocket release, the UAV immediately initiated autonomous attitude adjustment and trajectory takeover mode. Through multi-sensor data fusion and environmental perception, it achieved rapid stabilization within three seconds of separation, smoothly connecting to subsequent autonomous flight missions, effectively adapting to the dual challenges of high-speed rocket deployment and the changing environment of near space.

[0046] This embodiment covers the entire process optimization of using rockets as near-space UAV delivery platforms, including key technical aspects such as transportation, separation, initial control, and environmental adaptation. It improves the safety, delivery accuracy, and system reliability of UAV delivery using rockets, adapting to the needs of various high-precision and highly complex near-space missions.

[0047] In this embodiment, the rocket-launched drone deployment scheme is closely integrated with the deployment execution and adjustment module, and is implemented through the following steps: 1. Rocket launch preparation and trajectory planning: Based on the environmental prediction results output by the environmental prediction and state estimation model, and combined with the UAV flight trajectory and launch plan optimized by the ant colony algorithm, the optimal launch window, ascent trajectory and launch point location of the rocket are calculated.

[0048] Specifically, based on the multi-objective fitness function Determine the optimal trajectory parameters for rocket launch and ascent phases, with rocket deployment safety as a weighted factor. It accounts for a relatively high proportion.

[0049] 2. Integrated Rocket and UAV Design: To adapt to the rocket's high-speed flight and the extreme environment of near-space, the UAV section is equipped with a dedicated vibration reduction and heat insulation system, employing a special composite material structure to resist high-speed airflow and thermal stress. Simultaneously, a dedicated delivery bay is designed on the rocket's top, carrying a two-way data link and a high-precision inertial measurement unit, transmitting real-time rocket status data to the delivery execution and adjustment module for continuous optimization of delivery timing.

[0050] 3. Environmental Perception and Data Fusion During Rocket Launch Phase: During the rocket's ascent phase, onboard sensors continuously collect real-time meteorological data from near-space, which is then processed and analyzed using the aforementioned environmental prediction and state estimation model. When the rocket reaches a designated altitude range (typically 20-100 km in near-space), the release execution and adjustment module performs real-time comparison and analysis between the environmental data and the preset trajectory to calculate the optimal release point. The environmental perception data processing employs a network structure combining the aforementioned bidirectional long short-term memory network and an attention mechanism to adapt to the rapidly changing environmental data characteristics during the rocket's high-speed flight.

[0051] 4. Precise Rocket Launch Execution: When the rocket reaches the optimal launch altitude and position determined through ant colony optimization, the launch execution and adjustment module, based on real-time environmental condition assessment, sends a release command to the rocket launch system. The rocket launch system includes:

[0052] A multi-level safety release mechanism, with redundant design, ensures reliable unlocking and separation in near-space environments; The intelligent delivery cabin uses a hydraulic buffer device or a pneumatic deceleration control system to adjust the separation speed and attitude angle according to environmental changes. The separation shock mitigation system reduces the impact load on the UAV during the moment of separation when the rocket is at high speed.

[0053] 5. Post-rocket separation UAV attitude control: After separating from the rocket, the UAV enters the near-space autonomous flight phase. At this time, the system performs the following key steps: The cascaded attitude controller was activated to handle the initial disturbances caused by rocket release and the characteristics of the thin atmosphere in near space. An environmental prediction and state estimation model is used to process sensor data in real time. Based on the predictive-correction-compensation control strategy, the aerodynamic control surfaces or propulsion system of the UAV are quickly adjusted to ensure that a stable state is reached within 3 seconds after separation.

[0054] 6. Trajectory Connection and Mission Execution: After rocket release, the UAV executes subsequent flight missions based on the flight trajectory generated by the ant colony algorithm and combined with real-time environmental perception data. During this stage, the UAV continuously optimizes its flight path and dynamically adjusts control parameters according to the multi-target fitness calculation and pheromone update rules to adapt to changes in the near-space environment and ensure safe and efficient arrival at the mission target area.

[0055] Please see Figure 2 The present invention also provides a near-space unmanned aerial vehicle (UAV) delivery system, the system comprising: The data acquisition and modeling module is used to acquire historical flight data, near-space meteorological data, and real-time sensor data of vehicles and drones, and to build an environment prediction and state estimation model based on a deep learning network. The environment prediction and state estimation model outputs the environment prediction results. The ant colony algorithm initialization module is used to set the initial parameters of the ant colony algorithm according to the requirements of the drone deployment task, including the initial value of pheromone, the number of ants and the candidate set of flight trajectories, and to use the starting position and target position of the drone as the initial input of the optimization problem, and to optimize the drone flight trajectory based on the ant colony algorithm. The trajectory optimization and iteration module is used to dynamically adjust the pheromone update rules and transition probabilities in the ant colony algorithm based on environmental prediction results. It performs multi-objective optimization calculations on flight safety, energy consumption, mission completion time and economic cost, and outputs the UAV flight trajectory and the delivery plan of the carrier vehicle. The deployment execution and adjustment module is used to deploy the drone to a designated location in the near space based on the deployment plan, and to analyze the acquired real-time environmental data using environmental prediction and state estimation models during flight in order to adjust the drone's flight trajectory in real time.

[0056] Specifically, this embodiment of a near-space UAV delivery system implements end-to-end management from data acquisition, environmental modeling, trajectory optimization to precise delivery. It relies on a data acquisition and modeling module to achieve accurate prediction of environmental characteristics, an ant colony algorithm initialization and trajectory optimization iteration module to achieve global optimization of multi-target flight paths, and a delivery execution and adjustment module to ensure precise and reliable UAV release and flight control. The advantages of this near-space UAV delivery system are: real-time environmental modeling using deep learning networks improves task adaptability; and adaptive parameter adjustment of the ant colony algorithm solves the trajectory optimization problem under complex conditions.

[0057] The present invention also discloses an electronic device, comprising: at least one processor, at least one memory communication interface and bus; wherein the processor, memory and communication interface communicate with each other through the bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement a near-space unmanned aerial vehicle (UAV) deployment method.

[0058] This invention also discloses a computer-readable storage medium that stores computer instructions that cause the computer to implement all or part of the steps of the near-space UAV deployment method described in this embodiment of the invention. The storage medium includes various media capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0059] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for deploying near-space unmanned aerial vehicles (UAVs), characterized in that, The drone is deployed using a carrier vehicle, including the following steps: Historical flight data, near-space meteorological data, and real-time sensor data from vehicles and drones are acquired, and an environmental prediction and state estimation model is constructed based on a deep learning network. The environmental prediction and state estimation model outputs environmental prediction results. Based on the requirements of the drone deployment mission, the initial parameters of the ant colony algorithm are set, including the initial value of pheromone, the number of ants, and the candidate set of flight trajectories. The starting position and target position of the drone are used as the initial input of the optimization problem, and the drone flight trajectory is optimized based on the ant colony algorithm. By dynamically adjusting the pheromone update rules and transition probabilities in the ant colony algorithm based on environmental prediction results, multi-objective optimization calculations are performed on flight safety, energy consumption, mission completion time and economic cost, and the UAV flight trajectory and vehicle deployment scheme are output. Based on the deployment plan, the drone is deployed to a designated location in the near space, and during the flight, the acquired real-time environmental data is analyzed using environmental prediction and state estimation models to adjust the drone's flight trajectory in real time.

2. The near-space UAV deployment method as described in claim 1, characterized in that, The process involves acquiring historical flight data, near-space meteorological data, and real-time sensor data from vehicles and drones, and constructing an environmental prediction and state estimation model based on a deep learning network. The environmental prediction and state estimation model outputs environmental prediction results, including: Historical flight data, near-space meteorological data, and real-time sensor data of vehicles and drones are obtained from multi-source databases. Time alignment, noise reduction, normalization, and feature extraction are performed on the historical flight data, near-space meteorological data, and real-time sensor data of vehicles and drones to obtain a standardized training dataset. The original environment prediction and state estimation model is constructed based on a deep learning network that combines a bidirectional long short-term memory network with an attention mechanism. A hierarchical training strategy is adopted to train the model on the standardized training dataset to obtain an environment prediction and state estimation model for predicting near-space environment features and estimating the state of UAVs and vehicles.

3. The near-space UAV deployment method as described in claim 1, characterized in that, The process involves setting initial parameters for the ant colony algorithm based on the drone deployment task requirements, including initial pheromone values, ant count, and a candidate flight trajectory set. The drone's starting and target positions are used as initial inputs to the optimization problem. The drone's flight trajectory is then optimized using the ant colony algorithm, including: Based on the requirements of the drone delivery mission, including operating altitude, mission priority, weather conditions and energy limitations, the ant colony algorithm parameter matrix is ​​determined, and the initial value of pheromone, volatility coefficient, number of ants and number of iterations are dynamically configured. The initial parameters of the ant colony algorithm are adjusted according to the priority of task requirements, and a multi-level task priority evaluation system is established, which quantifies and assigns weights to four dimensions: safety, timeliness, energy efficiency, and economy. The initial value of the pheromone is set based on the historical deployment experience database, the volatility coefficient is dynamically adjusted according to the rate of change of the nearby spatial environment, the number of ants is positively correlated with the task complexity and accuracy requirements, and the number of iterations is determined by the dynamic convergence judgment criterion.

4. The near-space UAV deployment method as described in claim 3, characterized in that, The formula for adjusting the initial parameters is: ； ； ； in, For the first k Initial pheromone values ​​for task-like tasks As the baseline pheromone value, It is an exponential function. This is a task priority vector. For the work height factor, As an energy constraint factor, , , These are the weighting coefficients for task priority, operation altitude, and energy constraints, respectively. for t The pheromone evaporation coefficient at any given time. and These represent the maximum and minimum limits of pheromone evaporation coefficients, respectively. For the gradient of environmental change, For environmental sensitivity parameters, For the first k The number of ants for each task. As a metric for task complexity, For the accuracy requirement coefficient, and These are the first adjustment parameter and the second adjustment parameter, respectively. This is the floor function.

5. The near-space UAV deployment method as described in claim 1, characterized in that, The method dynamically adjusts the pheromone update rules and transition probabilities in the ant colony algorithm based on environmental prediction results, performs multi-objective optimization calculations on flight safety, energy consumption, mission completion time, and economic costs, and outputs the UAV flight trajectory and vehicle deployment scheme, including: Using a pre-built set of candidate flight trajectories, multi-objective fitness calculations are performed on each candidate trajectory based on flight safety, energy consumption, mission completion time, and economic cost of the UAV delivery mission, thereby generating a preliminary set of excellent candidate trajectories. Based on the fitness evaluation results of candidate trajectories, and combined with real-time environmental prediction data, the pheromone update rules and trajectory transition probabilities in the ant colony algorithm are dynamically adjusted to complete the iterative optimization of candidate trajectories and output the UAV flight trajectory and the delivery plan of the carrier.

6. The near-space UAV deployment method as described in claim 5, characterized in that, The formula for calculating the multi-objective fitness is: ； in, For multi-objective fitness functions, Candidate trajectories, , , and These are evaluation functions for flight safety, mission completion time, energy consumption, and economic cost, respectively. , , and These are the dynamically adjusted weighting coefficients for the safety, time, energy, and cost evaluation functions, respectively. As a safety volatility factor, n This represents the current iteration number. T This represents the total number of iterations. The formula for calculating the pheromone update rule is: ； in, for The updated value, To start from trajectory nodes i To trajectory node j pheromone concentration value, The pheromone evaporation coefficient, To increase the proportion of pheromones, The fitness evaluation value of the candidate trajectory. To prevent zero constant.

7. The near-space UAV deployment method as described in claim 1, characterized in that, The method involves deploying the drone to a designated location in near space based on a deployment plan, and analyzing the acquired real-time environmental data using environmental prediction and state estimation models during flight to adjust the drone's flight trajectory in real time. This includes: Based on the drone's flight trajectory and deployment plan, control the transport vehicle to reach the specified deployment conditions, activate the deployment mechanism to release the drone to the designated location in the near space, and perform initial attitude stabilization control after the drone separates. After the UAV is deployed to a designated location in near space, it continuously acquires and processes data from the UAV's onboard sensors. It then uses environmental prediction and state estimation models to analyze near-space weather conditions in real time and dynamically adjusts flight control parameters and trajectory planning to ensure that the UAV reaches the mission target area.

8. A near-space unmanned aerial vehicle (UAV) delivery system, used to execute a near-space UAV delivery method as described in any one of claims 1-7, characterized in that, The system includes: The data acquisition and modeling module is used to acquire historical flight data, near-space meteorological data, and real-time sensor data of vehicles and drones, and to build an environment prediction and state estimation model based on a deep learning network. The environment prediction and state estimation model outputs the environment prediction results. The ant colony algorithm initialization module is used to set the initial parameters of the ant colony algorithm according to the requirements of the drone deployment task, including the initial value of pheromone, the number of ants and the candidate set of flight trajectories, and to use the starting position and target position of the drone as the initial input of the optimization problem, and to optimize the drone flight trajectory based on the ant colony algorithm. The trajectory optimization and iteration module is used to dynamically adjust the pheromone update rules and transition probabilities in the ant colony algorithm based on environmental prediction results. It performs multi-objective optimization calculations on flight safety, energy consumption, mission completion time and economic cost, and outputs the UAV flight trajectory and the delivery plan of the carrier vehicle. The deployment execution and adjustment module is used to deploy the drone to a designated location in the near space based on the deployment plan, and to analyze the acquired real-time environmental data using environmental prediction and state estimation models during flight in order to adjust the drone's flight trajectory in real time.

9. An electronic device, characterized in that, include: At least one processor, at least one memory, a communication interface, and a bus; The processor, memory, and communication interface communicate with each other through the bus. The memory stores program instructions that can be executed by the processor. The processor calls the program instructions to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause the computer to perform the method as described in any one of claims 1 to 7.

Images