Extreme climate self-adaptive flight control system for unmanned aerial vehicle in complex terrain

Abstract

The invention discloses an extreme climate self-adaptive flight control system for an unmanned aerial vehicle in a complex terrain, and relates to the technical field of unmanned aerial vehicle self-adaptive flight control. The extreme climate adaptive flight control system for the unmanned aerial vehicle in the complex terrain comprises a flight data acquisition and preprocessing module, a climate prediction and risk assessment module, an optimal path and task sequence adjustment module, a real-time flight control and path optimization module and a safety emergency response and recording module. Wherein the flight data acquisition and preprocessing module is used for acquiring flight data; the climate prediction and risk assessment module is used for constructing a neural network model to predict climate; the optimal path and task sequence adjusting module is used for performing path and task security evaluation; the real-time flight control and path optimization module is used for converting a path and a task sequence into an aircraft instruction; and the safety emergency response and recording module is used for executing and recording the emergency strategy. The problem that the unmanned aerial vehicle is difficult to adjust flight in real time due to continuous change of climate states in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of self-adaptive flight control for unmanned aerial vehicles (UAVs), and in particular to an extreme climate self-adaptive flight control system for UAVs in complex terrain. Background Art

[0002] In recent years, drones have played an increasingly important role in power inspections, geological surveys, forest fire prevention, disaster monitoring, and emergency rescue. Especially in complex terrain, drones offer significant advantages over traditional manual methods, including speed, efficiency, and cost. However, in practical applications, drones often face challenges from various uncertainties, especially when complex terrain and extreme weather conditions intersect, placing higher demands on flight control systems.

[0003] Existing technologies still have significant shortcomings in climate forecasting. They rely heavily on analyzing real-time data and using rules to adjust flight paths. They lack the ability to model and predict climate trends, making it difficult to accurately identify extreme weather phenomena such as strong winds, rainfall, or sudden changes in air pressure that may occur in the near future. This short-sighted response mechanism often renders the system passive in the face of sudden climate changes, making it unable to proactively avoid risky areas or dynamically adjust mission sequences. This reduces flight safety and mission success rates, especially in complex terrain, which can lead to increased risk of flight deviation or loss of control.

[0004] Therefore, in response to the above problems, there is an urgent need for an extreme climate adaptive flight control system for UAVs in complex terrain. Summary of the Invention

[0005] Technical problems solved In response to the shortcomings of the existing technology, the present invention provides an extreme climate adaptive flight control system for UAVs in complex terrain, which solves the problem in the existing technology that UAVs are difficult to adjust their flight in real time due to constantly changing climate conditions.

[0006] Technical Solution To achieve the above objectives, the present invention is implemented through the following technical solutions: an extreme climate adaptive flight control system for unmanned aerial vehicles (UAVs) operating in complex terrain, comprising a flight data acquisition and preprocessing module, a climate prediction and risk assessment module, an optimal path and task sequence adjustment module, a real-time flight control and path optimization module, and a safety emergency response and recording module. The flight data acquisition and preprocessing module acquires climate data, ground altitude, and positioning data through climate sensors, a Determined Elevation Model (DEM), and a GPS system, and preprocesses the climate data, ground altitude, and positioning data. The climate prediction and risk assessment module constructs a neural network model based on climate data to predict the climate, conducts a comprehensive risk assessment, and determines whether to change the path and adjust the task sequence. The optimal path and task sequence adjustment module performs path and task safety assessments based on the comprehensive risk assessment results, and changes the path and adjusts the task sequence based on the assessment results. The real-time flight control and path optimization module converts the output optimal path and task sequence into aircraft instructions, controls the UAV flight in real time, and dynamically updates and adjusts the instructions. The safety emergency response and recording module executes emergency strategies in emergencies and records the entire process to ensure flight safety and traceability.

[0007] Furthermore, the specific steps for obtaining climate data, ground height and positioning data through climate sensors, DEM and GPS systems are as follows: collect climate data in real time through climate sensors, where the climate data includes: temperature, humidity, air pressure, wind speed and wind direction, and obtain climate vectors in real time; obtain the ground height of the current flight position in real time by querying DEM; and obtain the positioning data of the UAV and the positioning data of the path target points in real time through the GPS system.

[0008] Furthermore, the specific steps for preprocessing climate data, ground height and positioning data are as follows: Remove noise from climate data: Use Kalman filtering technology to gradually correct the noise signal of climate data through weighted average of measurement errors to remove noise; Remove outliers from climate data: Use standard deviation method to detect and remove outliers in climate data. When climate data is more than 3 times the standard deviation away from its mean value, it is considered an outlier and removed; Normalize climate data, ground height and positioning data: Use minimum-maximum standardization to convert climate data, ground height and positioning data of different dimensions into a unified standard range and the data falls within the interval [0,1]; Calculate the default cruising speed: Use the GPS position coordinates of two adjacent time points to calculate the flight distance, and then divide it by the time interval between the two time points to obtain the default cruising speed of the drone.

[0009] Furthermore, the specific steps for predicting climate by building a neural network model are as follows: A neural network model is constructed using historical climate data and real-time temperature, humidity, air pressure, wind speed and wind direction data. The standardized climate vector is obtained from the historical climate data and input into the model training channel. The climate vector is labeled using a time-step sliding window to generate an input data vector as the training target. The model is trained using a neural network optimization algorithm and an accelerated computing platform, and cross-validation and hyperparameter tuning techniques are applied to improve the model's generalization ability and prevent overfitting. The climate vector received in real time is input into the trained neural network model, and a multi-dimensional hidden state vector is output as a climate dynamic perception vector. A single bias term is obtained by minimizing the loss function using a gradient descent optimization algorithm using historical climate data.

[0010] Furthermore, the specific steps for conducting a comprehensive risk assessment to determine whether to change the path and adjust the task sequence are as follows: obtain the climate dynamic perception vector, sum the squares of each hidden dimension of the climate dynamic perception vector and take the square root to obtain the Euclidean norm of the climate dynamic perception vector, sum the Euclidean norm and a single bias term, and then input the sum result into the activation function to obtain a comprehensive risk assessment value; compare the comprehensive risk assessment value with the danger threshold in real time. When the comprehensive risk assessment value is less than or equal to the danger threshold, the current path does not need to be adjusted; when the comprehensive risk assessment value is greater than the danger threshold, the current path needs to be adjusted, and the comprehensive risk assessment value is output to the optimal path and task sequence adjustment module.

[0011] Furthermore, the specific steps for conducting path safety assessment and task safety assessment based on the comprehensive risk assessment results are as follows: obtain the Euclidean norm of the climate dynamic perception vector and the comprehensive risk assessment value of the path segment, multiply the Euclidean norm of the climate dynamic perception vector by the comprehensive risk assessment value of the path segment, multiply it by the path adjustment factor, add 1 to the calculation result and take the inverse to obtain the path safety assessment value; multiply the Euclidean norm of the climate dynamic perception vector at the current moment by the comprehensive risk assessment value of the path segment, multiply it by the task adjustment factor, add 1 to the calculation result and take the inverse to obtain the task safety assessment value.

[0012] Furthermore, the specific steps of changing the path and adjusting the task order based on the evaluation results are as follows: calculate the path safety evaluation values ​​of all flight paths, select a benchmark value through the quick sort method, divide the path safety evaluation value sequence into two subsequences less than and greater than the benchmark value, recursively sort the two subsequences, and finally merge to obtain an ordered result. After the sorting is completed, select the flight path with the lowest path safety score; calculate the mission safety evaluation values ​​of all flight missions, select a benchmark value through the quick sort method, divide the mission safety evaluation value sequence into two subsequences less than and greater than the benchmark value, and recursively sort the two subsequences, and finally merge to obtain an ordered result. After the sorting is completed, give priority to executing the task with the highest mission safety evaluation value.

[0013] Furthermore, the specific steps for converting the output optimal path and task sequence into aircraft instructions are as follows: the heading setting value that needs to be adjusted is calculated using the heading angle formula according to the positioning data of the UAV and the positioning data of the path target point; the comprehensive risk assessment value is multiplied by the wind speed adjustment factor, and the difference between 1 and the product is multiplied by the default cruising speed of the UAV to calculate the rate detection value of the current path segment; the ground altitude of the path point is added to the basic safety height, and the height compensation is performed according to the product of the comprehensive risk assessment value of the path segment and the height adjustment factor to calculate the height detection value of the current path segment; the PID controller method is used to dynamically convert the heading setting value, the speed detection value and the height detection value into aircraft control instructions, and the control instructions are sent to the motor controller to dynamically adjust the aircraft attitude to achieve path adjustment; according to the task sequence, the next task target is automatically loaded after each task node is reached, and the next flight scheduling logic is entered.

[0014] Furthermore, the specific steps for real-time control of drone flight and dynamic update and adjustment are as follows: During the flight, the system continuously collects climate data and calculates the comprehensive risk assessment value in real time; sets a risk jump threshold, and when the difference between the comprehensive risk assessment value at the current moment and the comprehensive risk assessment value at the previous moment is greater than the risk jump threshold, the flight control system will transmit feedback information back to the optimal path and task sequence selection module, triggering path re-planning and task sequence adjustment to form a closed-loop control.

[0015] Furthermore, in case of emergencies, emergency strategies are implemented and the entire process is recorded to ensure flight safety and traceability. The specific steps are as follows: Match the corresponding emergency strategy according to the type of anomaly: Emergency return: Generate the fastest return path to a safe point; Temporary landing: Select an area with relatively stable climate for controlled landing; Altitude avoidance: Temporarily increase the flight altitude to avoid near-ground wind disturbance; Route deviation: Reselect a low-risk area for detour; All emergency paths can also ensure that the emergency plan has minimal risk by recalculating the path safety score; Operator notification and log recording: The system automatically records abnormal situations, and the recorded data includes: time, location, risk type, and emergency measures; Feedback alarm information to the ground station and operator in voice and text form; At the same time, it is written into the log system for subsequent analysis and model training.

[0016] Beneficial effects The present invention has the following beneficial effects: (1) The present invention integrates climate sensors, DEM and GPS data, and combines neural networks to dynamically predict climate conditions and conduct risk assessments, thereby achieving accurate prediction of flight risks in complex terrain and extreme climate environments, thereby effectively improving the flight safety and mission completion rate of UAVs in harsh environments.

[0017] (2) The present invention realizes intelligent optimization and adjustment of flight paths and task execution sequences by dynamically calculating and ranking path safety assessment values ​​and task safety assessment values, thereby improving the efficiency of multi-task execution and the rationality of path decision-making.

[0018] (3) The present invention constructs a closed-loop feedback mechanism during flight by setting a risk jump threshold, combines the real-time updated risk assessment value, and dynamically triggers path reconstruction and task sequence reordering, thereby realizing the adaptive control capability and dynamic response capability of the flight control system.

[0019] (4) The present invention, by presetting multiple emergency strategies and combining them with real-time risk scoring to select the optimal emergency path, automatically records the entire process of abnormal events and notifies the operator, thereby achieving full traceability of the flight mission and enhancing the emergency response capabilities, significantly improving the safety and robustness of the system.

[0020] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a structural diagram of the extreme climate adaptive flight control system for UAVs in complex terrain; Figure 2 It is a histogram of path safety assessment values; Figure 3 A bar chart showing task safety assessment values. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] See also Figure 1-Figure 3An embodiment of the present invention provides a technical solution: an extreme climate adaptive flight control system for unmanned aerial vehicles (UAVs) operating in complex terrain, comprising a flight data acquisition and preprocessing module, a climate prediction and risk assessment module, an optimal path and task sequence adjustment module, a real-time flight control and path optimization module, and a safety emergency response and recording module. The flight data acquisition and preprocessing module acquires climate data, ground altitude, and positioning data through climate sensors, DEM, and GPS systems, and preprocesses the climate data, ground altitude, and positioning data. The climate prediction and risk assessment module constructs a neural network model based on climate data to predict the climate, conducts a comprehensive risk assessment, and determines whether to change the path and adjust the task sequence. The optimal path and task sequence adjustment module conducts path safety assessment and task safety assessment based on the comprehensive risk assessment results, and changes the path and adjusts the task sequence based on the assessment results. The real-time flight control and path optimization module converts the output optimal path and task sequence into aircraft instructions, controls the UAV flight in real time, and dynamically updates and adjusts them. The safety emergency response and recording module executes emergency strategies in emergencies and records the entire process to ensure flight safety and traceability.

[0024] Specifically, the steps for acquiring climate data, ground elevation, and positioning data using climate sensors, DEMs, and GPS are as follows: First, climate sensors collect real-time climate data. This data includes temperature, humidity, air pressure, wind speed, and direction, encompassing key meteorological parameters that influence the drone's flight state and forming a comprehensive environmental perception dimension. The system integrates and synchronously processes these climate elements through multiple channels, generating high-dimensional, structured climate vectors in real time. This provides highly accurate and timely input feature data for the subsequent neural network prediction model, ensuring the accuracy and dynamic nature of climate modeling. Second, by querying the DEM (Digital Elevation Model), ground elevation information at the current flight location is obtained in real time, enabling precise perception of complex terrain fluctuations. As a static geospatial reference, DEM data provides basic terrain constraints for flight path planning and altitude safety assessment, helping to avoid areas of sudden terrain changes and improving the terrain adaptability and overall safety of the flight trajectory. Finally, GPS is used to obtain real-time positioning data for the drone and path target points, ensuring high-precision spatial tracking and dynamic updates of the aircraft's position during mission execution. This positioning information is not only used to assist in trajectory planning, path correction and attitude adjustment, but also provides key position information support for the calculation and execution of flight control instructions, ensuring the continuity of task scheduling and the accuracy of the execution process.

[0025] This implementation integrates climate sensors, DEM, and GPS systems to collect real-time climate data, ground altitude, and positioning data. This creates a multi-dimensional, highly continuous environmental perception data system, enhancing the flight control system's ability to accurately model the flight environment. The acquired climate vectors, terrain altitude, and spatial positioning information provide highly accurate and timely data support for subsequent climate forecasting, risk assessment, and route optimization, thereby enhancing the system's adaptability and safety in complex terrain and extreme climate conditions.

[0026] Specifically, the steps for preprocessing climate data, ground height and positioning data are as follows: Denoising climate data: Using Kalman filtering technology, the algorithm is based on the optimal estimation theory of linear systems. By performing weighted averaging on the measurement errors of meteorological sensor input data, the random noise components in the climate data are gradually corrected, thereby obtaining smoother and more reliable meteorological data, providing a stable data basis for subsequent model training and control command generation; Denoising climate data: Using the standard deviation method to detect and remove outliers in climate data, that is, statistically calculating the mean and standard deviation of climate data. When the climate data at a certain moment deviates from its historical average by more than 3 times the standard deviation, it is identified as an outlier and removed to eliminate the interference of extreme weather fluctuations on system judgment and prediction accuracy; Climate data, ground height and positioning Data normalization: Using the maximum and minimum normalization methods, multi-source data with different physical dimensions and numerical ranges are uniformly processed and compressed to the [0, 1] interval. This normalization process can effectively improve the training efficiency and convergence speed of neural network models when processing heterogeneous data, while also avoiding bias issues caused by data scale differences. Calculating the default cruise speed: Based on the position coordinate data provided by the GPS system at two consecutive time points, the spherical distance formula or the plane Euclidean distance is used to calculate the displacement distance of the current flight segment. This distance is then divided by the corresponding time interval to obtain the actual average flight speed of the drone during that time period. This is used as the default cruise speed parameter, providing the basic input value for path planning and rate control.

[0027] In this implementation plan, the collected climate data, ground altitude and positioning data are systematically preprocessed, including using Kalman filtering to remove noise, standard deviation method to eliminate outliers, minimum-maximum normalization to achieve dimensional unification, and calculation of the default cruising speed based on GPS position information. This effectively improves the quality and consistency of multi-source data, enhances the flight control system's adaptability to environmental changes and the rationality of flight control parameter settings, thereby providing stable and reliable basic data support for subsequent climate forecasting, path planning and flight control.

[0028] Specifically, the specific steps for predicting climate by building a neural network model are as follows: First, a neural network model is built through historical climate data and real-time collected temperature, humidity, air pressure, wind speed and wind direction data, and a data-driven prediction framework is established based on multi-dimensional time series characteristics; standardized climate vectors are obtained from historical climate data and input into the model training channel, and a sliding window mechanism with a certain time span is combined with a time step sliding window to annotate the climate vectors, and an input data vector with time series characteristics is generated as a training target to improve the model's sensitivity to dynamic changes in time; secondly, a neural network optimization algorithm and an accelerated computing platform are used to The model is efficiently trained, and cross-validation and hyperparameter tuning techniques are applied to improve the generalization ability of the model, enhance the stability of the model under different environmental conditions, and effectively prevent the occurrence of overfitting; then, the climate vector received in real time is input into the trained neural network model, and a multi-dimensional hidden state vector is output through the deep network structure as a high-dimensional, nonlinear expression of the current environmental state, namely the climate dynamic perception vector; finally, a gradient descent optimization algorithm is used to minimize the loss function through historical climate data, so as to obtain a single bias term that reflects the overall climate risk trend, further enhancing the accuracy of risk assessment and model interpretability.

[0029] In this implementation plan, a neural network prediction model that integrates historical and real-time climate data is constructed, the input data is labeled using a time-step sliding window mechanism, and optimization algorithms and accelerated computing platforms are used for efficient training. At the same time, cross-validation and hyperparameter tuning strategies are introduced to enhance the generalization capability of the model. This model can output climate perception vectors and a single bias term with dynamic environmental characteristics in real time, enabling accurate modeling and efficient prediction of climate change trends. This significantly improves the flight control system's sensitivity to extreme climate and its forward-looking response, and enhances the overall intelligence level of flight path planning and risk control.

[0030] Specifically, the steps for conducting a comprehensive risk assessment to determine whether to change the path and adjust the task sequence are as follows: First, obtain the climate dynamic perception vector as a high-dimensional abstract expression of the current environmental state, where the vector is composed of multi-dimensional hidden states output by the neural network model; sum the squares of the hidden dimensions of the climate dynamic perception vector and take the square root to calculate its Euclidean norm, which is used to measure the overall climate change intensity and the degree of environmental disturbance; then, sum the calculated Euclidean norm with the single bias term obtained through historical data optimization to form a risk-based indicator that integrates the current state and long-term trends; and then input the summation result into the activation function. Complete nonlinear mapping, output comprehensive risk assessment value, and enhance the model's ability to express complex climatic conditions; then, use the Monte Carlo simulation method to comprehensively analyze historical flight data to set a danger threshold, and the system compares the comprehensive risk assessment value with the set danger threshold in real time. When the comprehensive risk assessment value is less than or equal to the danger threshold, it is judged that the current path is within the safe range, and the current path does not need to be adjusted; when the comprehensive risk assessment value is greater than the danger threshold, the system determines that there are risk hazards on the current path, and it is necessary to trigger the path reconstruction mechanism, and output the comprehensive risk assessment value as the decision basis to the optimal path and task sequence adjustment module, and start the path and task re-arrangement process.

[0031] The specific calculation formula for the comprehensive risk assessment value is:

[0032] Where, represents the comprehensive risk assessment value, represents the climate dynamic perception vector, Represents the climate dynamic perception vector The Euclidean norm of represents a single bias term.

[0033] This implementation calculates the Euclidean norm of the climate dynamic perception vector output by the neural network and constructs a risk assessment index system based on a single bias term derived from historical training. A nonlinear activation function is further introduced to enhance the risk assessment model's ability to represent complex climate conditions, enabling real-time calculation and determination of comprehensive risk assessment values. By comparing this assessment value with a preset risk threshold, it dynamically determines whether the current path presents climate risk, triggering path adjustments and task rescheduling mechanisms. This effectively enhances the flight control system's intelligent response capabilities and flight decision-making flexibility in complex terrain and extreme climate conditions.

[0034] Specifically, the specific steps for conducting path safety assessment and mission safety assessment based on the comprehensive risk assessment results are as follows: First, obtain the Euclidean norm of the climate dynamic perception vector and the comprehensive risk assessment value corresponding to the path segment. The former reflects the overall intensity of the current environmental disturbance, and the latter reflects the potential risk level of the path area; then, multiply the Euclidean norm of the climate dynamic perception vector by the comprehensive risk assessment value of the path segment to form a risk intensity product index, and further multiply it by the path adjustment factor set by the system. The path adjustment factor is obtained through offline evaluation of historical flight data and automatic parameter adjustment of the Bayesian optimization algorithm. The value range of the path adjustment factor is 0-1. This factor is used to perform weight correction on the terrain complexity, traffic importance, etc. of different paths; the above multiplication is carried out. The product result is added by 1 and the inverse is taken to eliminate the influence of extreme values, thereby obtaining the path safety assessment value. The higher the value, the safer the path, which facilitates subsequent path sorting and optimization. Next, the Euclidean norm of the climate dynamic perception vector at the current moment is multiplied again by the comprehensive risk assessment value of the path segment to reflect the mission execution risk level in the current time window, and multiplied by the mission adjustment factor. The mission adjustment factor is obtained through offline evaluation of historical flight data and automatic parameter adjustment of the Bayesian optimization algorithm. The value range of the path adjustment factor is 0-1, and the factor is dynamically assigned based on the mission type, priority and completion conditions. The calculation result is also added by 1 and the inverse is taken to obtain the mission safety assessment value, which provides a reliable risk basis and quantitative support for the subsequent optimization of the mission execution sequence.

[0035] The specific calculation formula of the path safety assessment value is:

[0036] Where, Indicates the path The path safety assessment value, represents the climate dynamic perception vector, Represents the climate dynamic perception vector The Euclidean norm of represents the path adjustment factor, Indicates the path Comprehensive risk assessment value; Among them, the Euclidean norm of the climate dynamic perception vector, the comprehensive risk assessment value of the path, and the path adjustment factor respectively represent the numerical values ​​of the impact of climate disturbances, geographical environmental risks, and path weights on the path safety assessment value, and can be directly called from the above modules when used.

[0037] In this embodiment, the disturbance intensity of path segment A is set to 2.0, the comprehensive risk value is set to 0.3, and the path adjustment factor is set to 0.8; the disturbance intensity of path segment B is set to 1.5, the risk value is set to 0.5, and the adjustment factor is set to 0.6; the disturbance intensity of path segment C is set to 3.0, the comprehensive risk value is set to 0.4, and the path adjustment factor is set to 0.9; the disturbance intensity of path segment D is set to 2.5, the comprehensive risk value is set to 0.6, and the path adjustment factor is set to 0.7; the disturbance intensity of path segment E is set to 1.0, the comprehensive risk value is set to 0.2, and the path adjustment factor is set to 0.5, as shown in Table 1. Table 1 Path safety assessment value data table

[0038] like Figure 2 As shown in the figure, the path safety evaluation value bar chart provided in the embodiment of the present application is shown in Table 1 and Figure 2 It can be seen that path segment E has the highest safety assessment value, indicating that its climate disturbance is the weakest, the risk is the lowest, and the overall flight safety is the best; path segment C has the lowest safety assessment value, mainly because its corresponding disturbance intensity is large, the risk value is relatively high, and the adjustment factor is large, which comprehensively manifests itself as a high-risk path at the current moment; the system will dynamically sort the flight path priorities according to the scoring graph, and give priority to path segments with high scores and low flight risks for task scheduling and flight control instruction generation.

[0039] The specific calculation formula for the task safety assessment value is:

[0040] Where, Indicates a task The mission safety assessment value, represents the climate dynamic perception vector, Represents the climate dynamic perception vector The Euclidean norm of represents the task adjustment factor, Indicates a task comprehensive risk assessment value.

[0041] Among them, the Euclidean norm of the climate dynamic perception vector, the comprehensive risk assessment value of the task and the task adjustment factor respectively represent the numerical values ​​of the impact of climate disturbance, geographical environment risk and task weight on the task safety assessment value, and can be called directly from the above modules when used.

[0042] In this embodiment, the disturbance intensity of task number T6 is set to 2.8, the comprehensive risk value of the task is set to 0.35, and the task adjustment factor is set to 0.85; the disturbance intensity of task number T7 is set to 1.1, the risk value is set to 0.53, and the adjustment factor is set to 0.75; the disturbance intensity of task number T8 is set to 2.2, the comprehensive risk value is set to 0.45, and the task adjustment factor is set to 0.65; the disturbance intensity of task number T9 is set to 1.6, the comprehensive risk value is set to 0.25, and the task adjustment factor is set to 0.90; the disturbance intensity of task number T10 is set to 3.1, the comprehensive risk value is set to 0.50, and the task adjustment factor is set to 0.60, as shown in Table 2: Table 2 Task safety assessment value data table

[0043] like Figure 3 As shown in the figure, it is a histogram of task safety evaluation values ​​provided in the embodiment of the present application. According to Table 2 and Figure 3 It can be seen that mission T9 has the highest safety score, mainly because its disturbance intensity and risk value are relatively low. Although the adjustment factor is large, it still obtains the best score; mission T10 has the lowest score, because its climate disturbance intensity is the largest and the risk value is high, which leads to a low safety assessment value; the system will dynamically sort the flight mission priorities according to the score map, and give priority to path segments with high scores and low flight risks for task scheduling and flight control instruction generation.

[0044] This implementation combines the Euclidean norm of the climate dynamic perception vector with the comprehensive risk assessment value of each path segment to construct a safety assessment model for both path and task dimensions. Path and task adjustment factors are then introduced to weight the assessment results. By adding 1 and taking the inverse, the impact of extreme values ​​on the assessment results is mitigated, achieving a quantitative safety assessment of each flight path and task execution priority. This method accurately determines the path risk level and task feasibility based on real-time risk status, providing a quantitative basis for path reconstruction and task rescheduling, significantly improving the system's flight strategy optimization capabilities and mission execution safety in highly dynamic climate environments.

[0045] Specifically, the specific steps of changing the path and adjusting the task order based on the evaluation results are as follows: First, calculate the path safety evaluation values ​​of all flight paths to form a path safety value sequence, which is used to measure the relative safety of each path under the current environmental conditions; then, select a benchmark value through the quick sorting method, divide the path safety evaluation value sequence into two subsequences that are less than and greater than the benchmark value, and continue to perform the quick sorting process on the two subsequences in a recursive manner to ensure sorting efficiency and stability; finally, merge the sorted subsequences to obtain a complete ordered path safety sequence, and select the flight path with the lowest path safety score value according to the sorting result, which means the flight path with the highest risk level. The path plan with the lowest risk and the highest airworthiness is selected as the preferred target for the current flight path; then, the mission safety assessment values ​​of all flight missions are calculated to judge the feasibility and risk level of each mission in the current environment; similarly, a benchmark value is selected through the quick sorting method, and the mission safety assessment value sequence is divided into two subsequences smaller than and larger than the benchmark value, and the two subsequences are sorted using a recursive sorting algorithm; after the sorting is completed, all results are merged to generate an ordered list of mission safety assessment values. According to the sorting results, the system gives priority to executing the tasks with the highest mission safety assessment values, thereby ensuring the safety priority principle in the task scheduling process and the stability of task completion.

[0046] This implementation rapidly sorts the safety assessment values ​​of all flight paths and tasks, employing an efficient recursive partitioning and merging strategy to quickly and orderly arrange the assessment sequences. This method then selects the optimal flight path based on the lowest path safety score, and prioritizes tasks based on the highest task safety score. This ensures that path and task scheduling decisions are dynamically optimized based on risk minimization. This approach not only improves the responsiveness of path and task switching but also enhances the system's global scheduling capabilities and the safety and reliability of flight decisions in a multi-path, multi-task concurrent environment.

[0047] Specifically, the specific steps for converting the output optimal path and task sequence into aircraft instructions are as follows: the heading setting value that needs to be adjusted is calculated using the heading angle formula according to the positioning data of the UAV and the positioning data of the path target point; the comprehensive risk assessment value is multiplied by the wind speed adjustment factor. The speed adjustment factor is obtained by fitting the relationship between the task effect and the risk level through regression analysis in multiple rounds of flight simulation using the minimum error optimization algorithm. The speed adjustment factor has a value range of 0-1. The difference between 1 and the product is multiplied by the default cruising speed of the UAV to calculate the rate detection value of the current path segment; the ground altitude of the path point is added to the basic safety altitude, and the speed adjustment factor is calculated according to the path. The height compensation is performed by multiplying the comprehensive risk assessment value of the path segment by the height adjustment factor. The height adjustment factor is obtained by simulating the minimum safe height increase required for flight under different risk levels, using regression fitting and fault-tolerant analysis algorithms. The value range of the height adjustment factor is 0-1, so as to calculate the height detection value of the current path segment; the PID controller method is used to dynamically convert the heading setting value, speed detection value and height detection value into aircraft control instructions, and the control instructions are sent to the motor controller to dynamically adjust the aircraft attitude to achieve path adjustment; according to the task sequence, the next task target is automatically loaded after each task node is reached, and the next flight scheduling logic is entered.

[0048] The specific calculation formula for the rate detection value is:

[0049] Where, Indicates the path The rate detection value, Indicates the default cruising speed value of the drone. represents the speed adjustment factor, Indicates the path Comprehensive risk assessment value; The specific calculation formula for the height detection value is:

[0050] Where, Indicates the path The height detection value of Indicates the ground altitude of the current flight position. Indicates the basic safety height, represents the height adjustment factor, Indicates the path comprehensive risk assessment value.

[0051] In this implementation, the optimal path and task sequence are converted into control instructions that can be executed by the aircraft. The heading angle formula is used to calculate the heading setting value. Based on the comprehensive risk assessment value combined with the wind speed adjustment factor and the altitude adjustment factor, the rate detection value and altitude detection value of the path segment are calculated respectively. Then, the PID controller is used to dynamically convert the above parameters into control instructions to achieve real-time adjustment of the aircraft's attitude, speed, and altitude. At the same time, the system automatically loads subsequent task objectives according to the task execution order to ensure the continuity and autonomy of the flight process. This method effectively enhances the flight control accuracy and task scheduling efficiency of UAVs under complex climate and terrain conditions, and improves the overall intelligence level and operational stability of the flight system.

[0052] Specifically, the specific steps for real-time control of drone flight and dynamic update and adjustment are as follows: During the flight, the system continuously collects climate data, covering multi-dimensional environmental parameters such as temperature, humidity, air pressure, wind speed and wind direction, and combines positioning and altitude information to build a flight environment perception model in real time. Based on this data, a comprehensive risk assessment value is calculated in real time to dynamically reflect the safety status of the current flight environment; the system sets a risk jump threshold as a risk-sensitive criterion. When the difference between the comprehensive risk assessment value at the current moment and the comprehensive risk assessment value at the previous moment exceeds the risk jump threshold, it is judged that the environmental state has changed significantly; at this time, the flight control system immediately transmits the change information as feedback to the optimal path and task sequence selection module, initiates the path re-planning and task sequence adjustment mechanism, and realizes information linkage between the flight control layer and the decision-making scheduling layer; through the above mechanism, the system forms a closed-loop control logic based on dynamic changes in risks, ensuring that the continuity and safety of flight missions can be maintained under complex or sudden climate conditions, and realizing intelligent and autonomous flight strategy adjustment capabilities.

[0053] This implementation continuously collects multidimensional climate data during flight and calculates a comprehensive risk assessment in real time. This, combined with a pre-defined risk transition threshold, enables a sensitive response to sudden environmental changes. When a dramatic risk shift is detected, the system immediately transmits feedback to the routing and task scheduling module, triggering path replanning and task sequence adjustments, thereby establishing a risk-driven closed-loop flight control system. This mechanism effectively enhances the flight system's adaptability to dynamic climate changes and its risk response speed, ensuring the safety and stability of flight missions in complex environments.

[0054] Specifically, the specific steps for executing emergency strategies and recording the entire process in emergency situations to ensure flight safety and traceability are as follows: First, the corresponding emergency strategies are matched according to the type of anomaly. The system presets a variety of emergency response scenarios, automatically identifies the type of fault or high-risk event through the anomaly recognition module, and calls the corresponding measures in the emergency strategy library, including: Emergency return: The system generates the fastest return path in real time based on the spatial topological relationship between the current flight position and the safety point, ensuring that the drone leaves the dangerous area as soon as possible and returns safely; Temporary landing: When the flight risk continues to rise and returning is not feasible, the system will choose an area with relatively stable climate conditions and small terrain to implement a controlled landing to reduce the safety hazards brought about by continued flight. ; Altitude avoidance: When low-altitude wind disturbance or obstacle threats appear, the system instructs the drone to temporarily increase its flight altitude and avoid the dangerous area to achieve rapid avoidance of near-ground risks; Route deviation: If the current path segment is in a high-risk state as a whole, the system will reselect the path segment in the area with a lower risk value to detour and achieve dynamic route adjustment; All emergency paths can be evaluated by recalculating the path safety score value to ensure that the selected emergency plan still has the minimum risk path characteristics under emergency response, thereby ensuring the effectiveness and safety of emergency measures; Secondly, in terms of operator notification and log recording: the system will automatically record abnormal situations, including the time of the event, the location of the drone, the type of risk detected, and the corresponding emergency treatment measures; Warning information is promptly fed back to the ground station and remote operators in the form of voice broadcasts and text messages, realizing a collaborative human-machine decision-making mechanism; at the same time, relevant data is fully written into the log recording system to provide high-quality data support for subsequent flight data analysis, system optimization and model retraining, achieving traceability and closed-loop optimization of the entire flight process.

[0055] In this implementation plan, by establishing a matching mechanism between abnormality types and various emergency strategies, combined with real-time risk assessment and path reconstruction capabilities, a variety of emergency response plans such as emergency return, temporary landing, altitude avoidance and route deviation are implemented, and the minimum risk characteristics of the emergency path are guaranteed by recalculating the path safety score value. At the same time, the system has the function of automatic recording and feedback of abnormal events, and can simultaneously convey the time, location, type and response measures of risk events to the operator in voice and text form, and write them completely into the log system for subsequent model training and flight behavior backtracking. The above mechanism significantly improves the safety assurance capability of flight missions under sudden risks and the traceability of the system, and enhances the intelligence and reliability of the entire flight control system.

[0056] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0057] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An extreme climate adaptive flight control system for UAVs operating in complex terrain, characterized by: It includes flight data acquisition and preprocessing module, climate prediction and risk assessment module, optimal path and task sequence adjustment module, real-time flight control and path optimization module and safety emergency response and recording module, among which: The flight data acquisition and preprocessing module acquires climate data, ground height and positioning data through climate sensors, DEM and GPS system, and preprocesses the climate data, ground height and positioning data; The climate prediction and risk assessment module uses climate data to build a neural network model to predict the climate and conduct a comprehensive risk assessment to determine whether to change the path and adjust the task sequence; The optimal path and task sequence adjustment module performs path safety assessment and task safety assessment according to the comprehensive risk assessment results, and changes the path and adjusts the task sequence based on the assessment results; The real-time flight control and path optimization module converts the output optimal path and task sequence into aircraft instructions, controls the UAV flight in real time, and dynamically updates and adjusts it; The safety emergency response and recording module executes emergency strategies in emergency situations and records the entire process to ensure flight safety and traceability.

2. The extreme climate adaptive flight control system for complex terrain of a UAV according to claim 1, characterized in that: The specific steps of obtaining climate data, ground height and positioning data through climate sensors, DEM and GPS system are as follows: Real-time climate data collection through climate sensors, including temperature, humidity, air pressure, wind speed and direction, and real-time climate vector acquisition; By querying DEM, the ground altitude of the current flight position can be obtained in real time; Through the GPS system, the positioning data of the drone and the positioning data of the path target points are obtained in real time.

3. The UAV complex terrain extreme climate adaptive flight control system according to claim 1, characterized in that: The specific steps of preprocessing the climate data, ground height and positioning data are as follows: Denoising climate data: Using Kalman filtering technology, the noise signal of climate data is gradually corrected and removed through the weighted average of measurement errors; Remove outliers from climate data: Use the standard deviation method to detect and remove outliers in climate data. When climate data deviates from its mean by more than 3 standard deviations, it is considered an outlier and removed. Normalization of climate data, ground height, and positioning data: The climate data, ground height, and positioning data of different dimensions are converted to a unified standard range through minimum-maximum normalization, and the data fall within the interval [0,1]; Calculate the default cruising speed: Use the GPS position coordinates of two adjacent time points to calculate the flight distance, and then divide it by the time interval between the two time points to get the default cruising speed of the drone.

4. The extreme climate adaptive flight control system for complex terrain of a UAV according to claim 1, characterized in that: The specific steps of predicting climate by building a neural network model are as follows: A neural network model is constructed using historical climate data and real-time temperature, humidity, air pressure, wind speed, and wind direction data. Standardized climate vectors are obtained from the historical climate data and input into the model training pipeline. A time-step sliding window is used to annotate the climate vectors and generate input data vectors as training targets. Use neural network optimization algorithms and accelerated computing platforms to train models, and apply cross-validation and hyperparameter tuning techniques to improve model generalization capabilities and prevent overfitting; The real-time climate vector is input into the trained neural network model, and a multi-dimensional hidden state vector is output as the climate dynamic perception vector. A single bias term is obtained by minimizing the loss function using a gradient descent optimization algorithm based on historical climate data.

5. The UAV complex terrain extreme climate adaptive flight control system according to claim 1, characterized in that: The specific steps for conducting a comprehensive risk assessment to determine whether to change the path and adjust the task sequence are as follows: Obtain a climate dynamic perception vector, sum the squares of each hidden dimension of the climate dynamic perception vector and take the square root to obtain the Euclidean norm of the climate dynamic perception vector, sum the Euclidean norm and a single bias term, and then input the summation result into an activation function to obtain a comprehensive risk assessment value; Compare the comprehensive risk assessment value with the danger threshold in real time. If the comprehensive risk assessment value is less than or equal to the danger threshold, the current path does not need to be adjusted. When the comprehensive risk assessment value is greater than the danger threshold, the current path needs to be adjusted, and the comprehensive risk assessment value is output to the optimal path and task sequence adjustment module.

6. The extreme climate adaptive flight control system for complex terrain of a UAV according to claim 1, characterized in that: The specific steps for performing path safety assessment and task safety assessment based on the comprehensive risk assessment results are as follows: Obtain the Euclidean norm of the climate dynamic perception vector and the comprehensive risk assessment value of the path segment, multiply the Euclidean norm of the climate dynamic perception vector and the comprehensive risk assessment value of the path segment, multiply the result by the path adjustment factor, add 1 to the result, and take the inverse to obtain the path safety assessment value; Multiply the Euclidean norm of the climate dynamic perception vector at the current moment by the comprehensive risk assessment value of the path segment, multiply it by the task adjustment factor, add 1 to the calculated result and take the inverse to obtain the task safety assessment value.

7. The extreme climate adaptive flight control system for complex terrain of a UAV according to claim 1, characterized in that: The specific steps of changing the path and adjusting the task order based on the evaluation results are as follows: Calculate the path safety assessment value of all flight paths, select a benchmark value through quick sorting, divide the path safety assessment value sequence into two subsequences, one smaller than and one larger than the benchmark value, recursively sort these two subsequences, and finally merge them to obtain an ordered result. After the sorting is completed, select the flight path with the lowest path safety score; Calculate the mission safety assessment values ​​of all flight missions, select a benchmark value through the quick sort method, divide the mission safety assessment value sequence into two subsequences that are less than and greater than the benchmark value, and recursively sort these two subsequences. Finally, merge them to obtain an ordered result. After the sorting is completed, the task with the highest mission safety assessment value is executed first.

8. The extreme climate adaptive flight control system for complex terrain of a UAV according to claim 1, characterized in that: The specific steps of converting the output optimal path and task sequence into aircraft instructions are as follows: The heading angle formula is used to calculate the heading setting value that needs to be adjusted based on the positioning data of the UAV and the positioning data of the path target point; Multiply the comprehensive risk assessment value by the wind speed adjustment factor, and multiply the difference between 1 and the product by the drone's default cruising speed to calculate the rate detection value for the current path segment. The ground elevation of the path point is added to the basic safety height, and the height compensation is performed based on the product of the comprehensive risk assessment value of the path segment and the height adjustment factor to calculate the height detection value of the current path segment; The PID controller method is used to dynamically convert the heading setting value, speed detection value and altitude detection value into aircraft control instructions. The control instructions are sent to the motor controller to dynamically adjust the aircraft attitude and realize path adjustment. According to the task sequence, the next task target is automatically loaded after each task node is reached, and the next flight scheduling logic is entered.

9. The UAV complex terrain extreme climate adaptive flight control system according to claim 1, characterized in that: The specific steps for controlling the drone flight in real time and dynamically updating and adjusting are as follows: During the flight, the system continuously collects climate data and calculates comprehensive risk assessment values ​​in real time; A risk jump threshold is set. When the difference between the comprehensive risk assessment value at the current moment and the comprehensive risk assessment value at the previous moment is greater than the risk jump threshold, the flight control system will transmit feedback information back to the optimal path and task sequence selection module, triggering path re-planning and task sequence adjustment to form a closed-loop control.

10. The extreme climate adaptive flight control system for complex terrain of a UAV according to claim 1, characterized in that: The specific steps for implementing emergency strategies and recording the entire process to ensure flight safety and traceability are as follows: Match the corresponding emergency strategy according to the exception type: Emergency return: Generate the fastest return route to a safe point; Temporary landing: select an area with relatively stable climate for controlled landing; Altitude avoidance: Temporarily increase the flight altitude to avoid wind disturbance near the ground; Route deviation: reselect the route segment in the low-risk area to detour; All emergency paths can also be recalculated to ensure that the emergency plan has the minimum risk; Operator Notification and Logging: The system automatically records abnormal situations, and the recorded data includes: time, location, risk type, and emergency measures; Feedback warning information to ground stations and operators via voice and text; At the same time, it is written into the log system for subsequent analysis and model training.

Images