Intelligent path planning and energy management system and method for long endurance fixed-wing unmanned aerial vehicles

Abstract

The present application relates to the technical field of unmanned aerial vehicle, in particular to an intelligent flight path planning and energy management system and method for long endurance fixed-wing unmanned aerial vehicle, comprising an environment perception unit; the flight path planning unit plans a flight trajectory meeting task requirements, safety requirements and energy constraints based on the unmanned aerial vehicle flight environment information obtained by the environment perception unit and the pre-stored unmanned aerial vehicle performance parameters; the energy management unit realizes dynamic energy management and optimization according to the real-time energy state of the unmanned aerial vehicle, the flight task and the planning result of the flight path planning unit. The flight path planning unit of the present application can call the energy consumption data of working conditions such as speed and height output by the energy management unit in real time, dynamically adjust the path weight to avoid high energy consumption sections; the energy management unit synchronously adjusts the main / backup battery charging and discharging strategy to optimize energy distribution according to the task time sequence of the flight path (such as waypoint priority, trajectory curvature).

Description

Intelligent trajectory planning and energy management system and method for long-endurance fixed-wing UAVs Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and more specifically, to an intelligent trajectory planning and energy management system and method for long-endurance fixed-wing UAVs. Background Technology

[0002] In the application of long-endurance fixed-wing UAVs in reconnaissance, surveying, and logistics, the coordinated optimization of trajectory planning and energy management directly determines their endurance and mission efficiency. Traditional technologies often prioritize path safety and mission completion in trajectory planning, while energy management independently focuses on battery consumption and recovery. This lack of dynamic coupling leads to problems such as energy imbalance and insufficient endurance for UAVs in complex weather conditions or long-endurance missions. Therefore, achieving integrated intelligent control of trajectory planning and energy management has become a key technical challenge in improving the environmental adaptability of UAVs.

[0003] For example, Chinese patent CN202110748173.1 discloses a UAV trajectory planning method based on an improved A* algorithm, including the following steps: Step S10, setting a starting point and a target point on a map, and randomly sampling points based on the starting point and target point; Step S20, creating a sampling tree, and expanding the sampling tree based on the sampling points; Step S30, calculating the node distances of the expanded sampling tree, and obtaining a node distance matrix based on the node distances; Step S40, using the A* algorithm and the node distance matrix to perform shortest path search, completing the UAV trajectory planning. The advantage of this invention is that it enables trajectory planning for UAVs with limited performance. Another example is Chinese patent CN202211029559.8, which discloses a UAV trajectory planning and obstacle avoidance method and system based on a spectrum map. This method first constructs a three-dimensional spectrum map on an edge server, the UAV pre-plans a path based on this three-dimensional spectrum map, and then the UAV dynamically avoids obstacles in real time based on the local spectrum map of its neighborhood. During the execution of a pre-planned path, the UAV receives a local spectrum map of its neighborhood in real time, distributed by the base station. When interference is detected, the UAV employs a perturbation fluid algorithm for dynamic obstacle avoidance. By adjusting the repulsion response coefficient, tangential response coefficient, and tangential direction coefficient in the perturbation fluid algorithm using a deep reinforcement learning algorithm based on a dual-delay deep deterministic policy gradient, the UAV can achieve real-time obstacle avoidance while minimizing the total flight path, all while satisfying motion constraints. The method provided in this invention offers UAVs global and local real-time spectrum map services, effectively improving the efficiency and robustness of UAV trajectory planning and dynamic obstacle avoidance.

[0004] Although the above technical solutions have their own design advantages, they also have the following technical defects: First, the track-energy unit coordination mechanism is missing: The improved A-satellite algorithm of Chinese patent CN202110748173.1 only realizes geometric path optimization through a track planning module, but does not establish a closed-loop interaction with the energy management module: it neither calls the energy unit's "energy consumption prediction model" (such as the energy consumption difference at different speeds / altitudes), nor dynamically adjusts the path weight according to the "remaining range" output by the energy unit (such as avoiding high-energy-consuming segments), resulting in a complete separation between track planning and energy supply, and long-endurance missions are prone to interruption due to energy consumption runaway; The spectrum obstacle avoidance strategy of Chinese patent CN202211029559.8, although involving dynamic track adjustment, only passively responds to obstacle avoidance energy consumption in energy management, does not integrate active strategies such as "gravitational potential energy recovery", and does not schedule waypoints according to the "real-time SOC" priority of the energy unit, and cannot achieve long-term balance between energy supply and demand. Secondly, the coupling depth of multi-source data across units is insufficient: The trajectory planning module of Chinese patent CN202110748173.1 only covers terrain geometry information in its sampled data, without integrating the equipment status data (such as battery aging coefficient) required by the energy unit and the meteorological data (the impact of wind speed on energy consumption) of the environment. This results in insufficient accuracy of the trajectory's "energy consumption prediction" and an inability to predict sudden changes in energy consumption during headwind / climbing sections. The energy management module of Chinese patent CN202211029559.8 only associates electromagnetic interference with location, without integrating the task timing constraints of the trajectory unit (such as waypoint priority) and the multi-dimensional status of the energy unit (such as charging and discharging characteristics). It cannot adjust the speed through wind speed prediction, nor can it allocate energy according to task priority, resulting in poor "track-energy" coordination. Third, the cross-unit adaptability in dynamic environments is lacking: The trajectory algorithm of Chinese patent CN202110748173.1 has a fixed process and lacks a cross-unit adaptive mechanism. When environmental complexity changes abruptly (e.g., dense obstacles, sudden weather changes), the trajectory unit does not coordinate with the energy unit to switch to an "energy consumption / efficiency priority" strategy (e.g., slowing down for energy saving in complex terrain), nor does it adjust the path length according to the energy unit's "endurance threshold." The energy strategy of Chinese patent CN202211029559.8 relies on offline training and is not integrated with online learning by the trajectory unit. Faced with unknown weather / equipment aging, the energy unit's "disturbance parameters" cannot infer the trajectory unit's path adjustment, and it does not link energy recovery and trajectory planning, making it difficult to cope with dynamic fluctuations during long-endurance operation. Therefore, we propose an intelligent trajectory planning and energy management system and method for long-endurance fixed-wing UAVs. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent trajectory planning and energy management system and method for long-endurance fixed-wing unmanned aerial vehicles (UAVs) to solve the problems mentioned in the background art.

[0006] To address the aforementioned technical problems, one objective of this invention is to provide an intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles, comprising:

[0007] The environmental perception unit is used to acquire real-time flight environment information of the UAV. Based on multi-sensor fusion technology, it integrates a data acquisition module to collect data on terrain, meteorological conditions, obstacle distribution and potential threat areas, and transmits the data to the flight path planning unit and the energy management unit.

[0008] The trajectory planning unit plans a flight trajectory that meets mission requirements, safety requirements and energy constraints based on the UAV flight environment information obtained by the environmental perception unit and the pre-stored UAV performance parameters. It uses a genetic algorithm combined with a simulated annealing algorithm for offline pre-planning and uses Dijkstra's algorithm for online dynamic adjustment. The planning results are then transmitted to the energy management unit and the decision control unit.

[0009] The energy management unit realizes dynamic energy management and optimization based on the real-time energy status of the UAV, flight mission, and the planning results of the trajectory planning unit. Based on energy monitoring, energy prediction, energy allocation and energy optimization, combined with energy recovery and reuse mechanisms, the management data is fed back to the decision control unit.

[0010] The decision control unit receives data from the environmental perception unit, the trajectory planning unit, and the energy management unit, and performs information fusion based on an adaptive weighted fusion algorithm. Based on the fused information, it generates flight status control commands and mission execution control commands using a hierarchical decision-making strategy. Through the flight status control submodule and the mission execution control submodule, it uses PID control technology to drive the UAV actuators and mission execution equipment.

[0011] The parameter storage unit is used to store UAV performance parameters (including but not limited to maximum flight speed, maximum ceiling, fuel consumption rate, power system power, etc.) for use by the trajectory planning unit.

[0012] As a further improvement to this technical solution, the data acquisition module includes a terrain and landform acquisition submodule, a meteorological parameter acquisition submodule, an obstacle detection submodule, and a threat perception submodule, wherein:

[0013] The terrain and landform acquisition submodule uses LiDAR and optical camera data fusion technology to generate a three-dimensional terrain model;

[0014] The meteorological parameter acquisition submodule includes a micro weather station and a turbulence sensor, which are used to collect wind speed, air pressure, temperature, humidity and turbulence intensity data in real time;

[0015] The obstacle detection submodule achieves dynamic obstacle detection within a preset range by fusing millimeter-wave radar with binocular vision and other multi-sensor technology.

[0016] The threat perception submodule uses electronic reconnaissance equipment and infrared thermal imagers to detect the location and intensity of electromagnetic radiation sources and heat source targets.

[0017] As a further improvement to this technical solution, the environmental perception unit further includes a data fusion module, which comprises a spatiotemporal registration submodule, a feature fusion submodule, and a decision fusion submodule, wherein:

[0018] The spatiotemporal registration submodule achieves precise alignment of multi-sensor data in time and space through hardware time synchronization and homogeneous coordinate transformation.

[0019] The feature fusion submodule uses deep learning algorithms to extract point cloud and image features and establish cross-modal data correlation.

[0020] The decision fusion submodule integrates multi-source decision information based on evidence theory to generate an environmental situation assessment result that includes threat level and response strategy.

[0021] As a further improvement to this technical solution, the trajectory planning unit includes a task parsing module, a path generation module, and a performance evaluation module, wherein:

[0022] The task parsing module is used to convert the task objective into an ordered sequence of waypoints and determine the priority of each waypoint, including the following steps:

[0023] S210.1. An improved target decomposition algorithm is used to generate the waypoint sequence. The calculation formula is as follows:

[0024] ;

[0025] in, Represents the three-dimensional coordinates and time constraints of the i-th waypoint;

[0026] S210.2. Establish the task constraint set C based on the improved genetic algorithm, and use the priority weight formula. Calculate the priority of each waypoint; among which, As a task importance factor, As a factor of time urgency, As a resource consumption factor, + =1 and [0,1]; This formula achieves dynamic allocation of waypoint execution priority by comprehensively considering multiple factors.

[0027] The path generation module is used to construct an environmental cost map and plan flight paths. It includes the following steps:

[0028] S220.1 Constructing a 3D Environmental Cost Map Its total cost function is:

[0029] ;

[0030] in, coordinates The environmental costs resulting from topographical factors; Representing coordinates The environmental costs arising from location and meteorological conditions; coordinates The environmental costs caused by obstacles; coordinates Environmental costs caused by regional threat factors; quantifying the environmental risks of drone flight paths by overlaying the costs of terrain, weather, obstacles, and threat areas;

[0031] S220.2. A modified genetic algorithm combined with simulated annealing is used for global path planning. The modified genetic algorithm has an adaptive crossover probability. and mutation probability The calculation formula is:

[0032] ;

[0033] ;

[0034] in, , These are the maximum and average fitness of the population, respectively. Let f be the fitness of the crossover individual, and let f be the fitness of the variant individual. [0,1]; By dynamically adjusting the crossover and mutation probabilities, the efficiency and accuracy of the algorithm in global path search are improved.

[0035] S220.3. Use an improved Dijkstra's algorithm for online dynamic adjustment, employing heuristic functions. To speed up the search; among them From the starting point to the node The actual cost, Let n be the estimated cost from node n to the target point. [0.5, 1.5] represents the weighting coefficients; this function introduces heuristic information to achieve a fast response when the path is dynamically adjusted.

[0036] The performance evaluation module is used to verify the feasibility of the flight path and predict energy consumption and risks, including the following steps:

[0037] S230.1, Based on the improved dynamic constraint verification algorithm, through formula Verify whether the turning radius of the flight path meets the minimum limit. ,in For flight speed, It is the acceleration due to gravity. The tilt angle (rad) is used to ensure the flight path conforms to the dynamic performance of the UAV.

[0038] S230.2. An improved genetic algorithm based on the energy gradient field is used to predict the trajectory energy consumption E. The calculation formula is as follows:

[0039] ;

[0040] in, Number of track segments; For segmented indexes; For the first The starting time of the flight path; For the first The end time of the flight path; It represents instantaneous power, which is related to parameters such as speed and altitude. The total energy consumption of the flight path is calculated by integration.

[0041] S230.3, Risk index update formula based on Dijkstra's algorithm Quantify flight path risks; among them, Total risk of the flight path; For a moment The location of the drone; For position The risk density function at the location enables dynamic assessment of risk throughout the entire flight path.

[0042] As a further improvement to this technical solution, the path generation module includes an environmental cost map. A hierarchical modeling strategy is adopted, specifically including:

[0043] Static cost layer It is constructed using the octree algorithm, integrating terrain elevation cost and obstacle cost, and the calculation formula is as follows:

[0044] ;

[0045] in, Terrain elevation cost (dimensionless, used to quantify the impact of terrain undulation on flight), Terrain elevation cost Using meters as the unit of measurement, terrain point cloud data is collected by lidar on a drone, and a digital elevation model (DEM) of the mission area is constructed using the Kriging interpolation method. The value is the difference between the elevation of the coordinate point in the DEM and the reference elevation (such as the lowest elevation of the mission area), used to quantify the impact of terrain undulation on UAV flight. Obstacle cost (dimensionless, used to mark the passage cost of an obstacle area); weighting coefficient , satisfy + =1 and [0,1] (By adjusting the weights, the influence of terrain and obstacles on path planning is balanced);

[0046] Dynamic cost layer Based on real-time updates using a meteorological field interpolation algorithm, and integrating the dynamic effects of headwind and turbulence, the calculation formula is as follows:

[0047] ;

[0048] in, The cost of headwind drag (dimensionless, reflecting the impact of headwind on flight energy consumption and speed). Cost of turbulence impact (dimensionless, quantifying the interference of turbulence on flight stability); weighting coefficient , satisfy and [0,1] (Adapting to different meteorological conditions, prioritizing headwinds and turbulence for path planning);

[0049] Threat Cost Layer The threat region is described based on a Gaussian mixture model, and the calculation formula is as follows:

[0050] ;

[0051] in, For the first Threat source intensity (dimensionless, reflecting the degree of danger of the threat source). For position To the Euclidean distances between threat sources (unit: meters, used to calculate threat attenuation). For the first The impact range parameter of a threat source (unit: meters, which determines the spatial range of the threat's effect). The total number of threat sources (a positive integer, representing the number of threat sources to be avoided); Gaussian mixture model parameters. , The calibration method involves selecting n typical electromagnetic threat source test points in the mission area and using an electromagnetic radiation detector to measure the electromagnetic radiation intensity values ​​at different distances and angles. ( For test point number, (as radiation direction number); or as radiation intensity value Using samples, the maximum likelihood estimation method is used to fit a Gaussian mixture model to establish the relationship between measured electromagnetic radiation intensity and... (Standard deviation of distribution, reflecting the degree of dispersion of radiation intensity) The mapping relationship between (distribution mean and the direction of the threat source center) is used to calibrate parameters for subsequent probability distribution modeling and cost calculation of threat areas.

[0052] The total cost of the environmental cost map is the sum of the costs from each layer, that is:

[0053] = + + This is used to provide a quantitative basis for environmental risk in flight path planning; among which, obstacle costs have been approved. Integrate into the static cost layer middle.

[0054] As a further improvement to this technical solution, the performance evaluation module includes a feasibility verification submodule, an energy consumption prediction submodule, and a risk assessment submodule, wherein:

[0055] The feasibility verification submodule uses the geometric constraint method to verify whether the turning radius of the UAV's trajectory meets the minimum turning radius requirement of the UAV, and to verify whether the climb rate and descent rate are within the allowable range.

[0056] The energy consumption prediction submodule estimates the total energy consumption of the flight path based on the energy consumption model of the segmented flight path, combined with flight speed, altitude and weather conditions.

[0057] The risk assessment submodule uses a risk weight field algorithm to transform risk factors (such as terrain obstacles, weather threats, and no-fly zones) into path costs and calculate the comprehensive risk index of the flight path.

[0058] As a further improvement to this technical solution, the energy management unit includes an energy monitoring module, an energy prediction module, an energy allocation module, and an energy optimization module, wherein:

[0059] The energy monitoring module collects the state of charge, charging and discharging power and temperature parameters of the battery pack in real time through voltage sensors, current sensors and temperature sensors, and estimates the remaining battery capacity by combining the Kalman filter algorithm.

[0060] The energy prediction module is based on a long short-term memory network model trained on historical flight data. Combined with the current flight path planning results, it predicts the energy consumption distribution of each flight segment and generates a prediction curve for the remaining flight range.

[0061] The energy allocation module optimizes flight parameters (such as airspeed and altitude) using a dynamic programming algorithm based on task priority and real-time energy status, balancing cruise energy consumption and mission completion efficiency.

[0062] The energy optimization module achieves efficient charging of solar panels through maximum power point tracking technology and maximizes solar energy collection efficiency by combining flight attitude adjustment algorithms.

[0063] As a further improvement to this technical solution, the energy recovery and reuse mechanism of the energy management unit specifically includes:

[0064] When the trajectory planning unit outputs trajectory adjustment commands, the energy management unit monitors the charge status of the main and backup batteries in real time. During the cruise phase, the main battery prioritizes powering the drone. When the trajectory enters a recyclable energy scenario (such as the gliding phase), it automatically switches to the backup battery for charging, storing redundant energy during flight. The charging and discharging switching is based on a simple power threshold preset by the flight controller (the backup battery is activated to replenish power when the main battery SOC is low to about 10%, and the backup battery is paused when the backup battery SOC is high to about 90%), balancing battery usage and energy retention.

[0065] As a further improvement to this technical solution, the decision control unit includes an information fusion module, a decision generation module, and an instruction execution module, wherein:

[0066] The information fusion module is used to purify multi-source data and dynamically allocate fusion weights to ensure the accuracy and real-time nature of decision input. The information fusion module includes a data preprocessing submodule and a dynamic weight allocation submodule, wherein:

[0067] The data preprocessing submodule uses the Kalman filter algorithm to reduce noise in the environmental perception data, aligns the trajectory planning data through timestamp synchronization technology, and uses the threshold detection method to filter out abnormal values ​​in the energy management data.

[0068] The dynamic weight allocation submodule constructs a confidence model based on the sensor's historical error rate and current working status, and dynamically adjusts the fusion ratio in conjunction with the task stage weight matrix.

[0069] The decision generation module generates global strategies and control commands based on fused data, constructs a decision model using Markov decision process, and uses UAV position, remaining energy, and mission progress as state variables to generate global strategies; and uses model predictive control algorithm to continuously optimize flight parameters within a preset prediction time domain, and outputs trajectory tracking commands in combination with UAV dynamic constraints.

[0070] The command execution module is used to convert decision commands into controllable signals for the actuators, ensuring safe and accurate execution. The command execution module includes a flight status control submodule and a mission execution control submodule, wherein:

[0071] The flight status control submodule adopts a PID+feedforward composite control structure. The feedforward part pre-calculates the ideal control quantity based on the UAV dynamics model, and the feedback part performs PID adjustment through real-time error, outputting a flight control signal including attitude angle and throttle value.

[0072] The task execution control submodule presets a device load safety threshold. When the instruction exceeds the threshold, a proportional attenuation strategy is triggered and fed back to the decision generation module to adjust the strategy.

[0073] The second objective of this invention is to provide an intelligent trajectory planning and energy management method for long-endurance fixed-wing unmanned aerial vehicles (UAVs). Based on the aforementioned intelligent trajectory planning and energy management system for long-endurance fixed-wing UAVs, the method includes the following steps:

[0074] S100: Through the multi-sensor fusion technology of the environmental perception unit, real-time data on terrain, weather conditions, obstacle distribution and threat areas are collected. After spatiotemporal registration, feature fusion and decision fusion processing, environmental situation assessment results are generated.

[0075] S200: Based on environmental situation assessment results and UAV performance parameters, offline trajectory pre-planning is carried out using a combination of genetic algorithm and simulated annealing algorithm. Online dynamic adjustment is achieved through Dijkstra algorithm. The feasibility of the trajectory is verified by turning radius constraint, energy consumption prediction and risk assessment.

[0076] S300 collects battery parameters and estimates remaining capacity through the energy monitoring module based on real-time flight track and energy status, uses a long short-term memory network to predict energy consumption for each flight segment, optimizes energy allocation using a dynamic programming algorithm, dynamically switches between primary and backup battery charging and discharging based on flight track scenarios, and achieves energy reuse by relying on flight redundancy energy storage and power threshold control.

[0077] The S400 inputs environmental perception, trajectory planning, and energy management data into the decision control unit. After processing by an adaptive weighted fusion algorithm, it generates a global mission strategy using a Markov decision process. The system then outputs flight control commands through model predictive control and PID regulation to drive the actuators to complete the mission.

[0078] The S500 stores and updates UAV performance parameters (such as maximum flight speed and fuel consumption rate) in real time through a parameter storage unit, which can be dynamically called by the trajectory planning unit during the trajectory planning process of the S200.

[0079] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0080] 1. The trajectory planning unit of this invention can call upon the operating condition energy consumption data such as airspeed and altitude output by the energy management unit in real time, and dynamically adjust the path weight to avoid high-energy-consuming segments; the energy management unit simultaneously optimizes energy allocation by intelligently adjusting the charging and discharging strategies of the main / backup batteries according to the task sequence of the trajectory (such as waypoint priority and trajectory curvature). Through the two-way feedback design of "dynamic energy consumption control in trajectory planning + energy management adapting to trajectory requirements", the limitations of the separation between trajectory and energy management in traditional solutions are broken, upgrading trajectory planning from a single geometric optimization to a dual-dimensional optimization of "path + energy consumption"; energy management evolves from passive energy replenishment to actively matching trajectory task requirements, constructing a dynamic collaborative mechanism between trajectory and energy, effectively supporting the energy consumption balance of long-endurance missions, and improving the reliability of UAV endurance and mission execution;

[0081] 2. This invention integrates meteorological (wind speed, air pressure) and terrain data from environmental perception, the battery's full life-cycle status (aging coefficient, charge / discharge characteristics) from the energy management unit, and the task constraints (timing, priority) from the trajectory planning unit to construct a "meteorology-energy consumption-task" correlation model. This model can accurately predict sudden changes in energy consumption in scenarios such as headwinds and uphill climbs, and also supports the energy management unit in allocating resources according to task priority. It overcomes the shortcomings of traditional solutions, such as single data dimensions and poor coordination, and improves the accuracy of cooperation between the trajectory planning unit and the energy management unit.

[0082] 3. Based on environmental complexity (dynamic obstacle density, intensity of meteorological changes), the trajectory planning unit and energy management unit can collaboratively switch working modes (such as simultaneously triggering "energy consumption priority" trajectory optimization and energy recovery enhancement strategies under complex terrain); through an online learning mechanism, the energy consumption prediction parameters of the energy management unit can feed back into the trajectory planning algorithm, dynamically adjust the path length and attitude, and simultaneously link the energy recovery strategy to match trajectory changes, effectively cope with energy fluctuations and trajectory adjustment needs under unknown operating conditions, and ensure the continuous execution reliability of long-endurance missions. Attached Figure Description

[0083] Figure 1 is a schematic diagram of the system framework of the present invention;

[0084] Figure 2 is a schematic diagram of the method steps of the present invention;

[0085] Figure 3 is a schematic diagram of the system framework of the environmental sensing unit in this invention;

[0086] Figure 4 is a schematic diagram of the system framework of the trajectory planning unit in this invention;

[0087] Figure 5 is a schematic diagram of the system framework of the decision control unit in this invention;

[0088] The meanings of the labels in the diagram are as follows:

[0089] 100. Environmental Perception Unit; 110. Data Acquisition Module; 111. Topography and Geomorphology Acquisition Submodule; 112. Meteorological Parameter Acquisition Submodule; 113. Obstacle Detection Submodule; 114. Threat Perception Submodule; 120. Data Fusion Module; 121. Spatiotemporal Registration Submodule; 122. Feature Fusion Submodule; 123. Decision Fusion Submodule;

[0090] 200. Flight path planning unit; 210. Mission parsing module; 220. Path generation module; 230. Performance evaluation module; 231. Feasibility verification submodule; 232. Energy consumption prediction submodule; 233. Risk assessment submodule;

[0091] 300. Energy Management Unit; 310. Energy Monitoring Module; 320. Energy Prediction Module; 330. Energy Allocation Module; 340. Energy Optimization Module;

[0092] 400. Decision control unit; 410. Information fusion module; 411. Data preprocessing submodule; 412. Dynamic weight allocation submodule; 420. Decision generation module; 430. Command execution module; 431. Flight status control submodule; 432. Mission execution control submodule;

[0093] 500. Parameter storage unit. Detailed Implementation

[0094] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all 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.

[0095] Example 1

[0096] As shown in Figure 1, this embodiment provides an intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles, including:

[0097] As shown in Figure 3, the environmental perception unit 100 is used to acquire UAV flight environment information in real time. Based on multi-sensor fusion technology, it integrates a data acquisition module 110 to collect data on terrain, meteorological conditions, obstacle distribution and potential threat areas, and transmits them to the flight path planning unit 200 and the energy management unit 300.

[0098] In this embodiment, the data acquisition module 110 includes a terrain and landform acquisition submodule 111, a meteorological parameter acquisition submodule 112, an obstacle detection submodule 113, and a threat perception submodule 114, wherein:

[0099] The terrain and landform acquisition submodule 111 uses LiDAR and optical camera data fusion technology to generate a three-dimensional terrain model;

[0100] As a further explanation of this embodiment, the lidar in this embodiment can be a 16-line mechanical lidar, generating point cloud data at a scanning frequency of 10-20Hz; the optical camera can be an industrial-grade area array camera, coupled with a 1280×960 resolution image sensor, to collect surface texture information in real time. The two devices are synchronized through hardware timestamps, and after spatiotemporal registration of point cloud and image features, a three-dimensional terrain mesh model is constructed based on the octree algorithm. The octree algorithm is used for terrain mesh construction, and its hierarchical storage structure can dynamically adjust the mesh resolution according to the flight altitude, reducing data storage while ensuring modeling accuracy.

[0101] The meteorological parameter acquisition submodule 112 includes a miniature weather station and a turbulence sensor, which are used to collect wind speed, air pressure, temperature, humidity and turbulence intensity data in real time;

[0102] As a further explanation of this embodiment, the meteorological parameter acquisition submodule 112 integrates a miniature weather station and a turbulence sensor. The weather station is equipped with a composite sensor for wind speed, air pressure, temperature, and humidity to collect environmental parameters at an appropriate frequency. The turbulence sensor adopts a high-frequency sampling mechanism to monitor the intensity of atmospheric turbulence in real time. When the wind speed exceeds a preset threshold, the system automatically triggers the high-frequency sampling mode to ensure the real-time performance of data under strong wind conditions.

[0103] The obstacle detection submodule 113 achieves dynamic obstacle detection within a preset range through multi-sensor fusion of millimeter-wave radar and binocular vision;

[0104] As a further explanation of this embodiment, the millimeter-wave radar in this embodiment achieves long-distance obstacle ranging through FMCW modulation technology. The binocular camera constructs a stereo vision system with an appropriate baseline distance, and works with a lens of corresponding focal length to achieve field of view coverage. Sensor data can be updated through Kalman filtering, and joint detection of dynamic obstacles is achieved based on the Mahalanobis distance correlation threshold. The detection results are transmitted to the data fusion module 120 in real time.

[0105] The threat perception submodule 114 uses electronic reconnaissance equipment and an infrared thermal imager to detect the location and intensity of electromagnetic radiation sources and heat source targets. The threat perception range of the threat perception submodule 114 is limited to the civilian electromagnetic environment.

[0106] As a further explanation of this embodiment, the electronic reconnaissance equipment and infrared thermal imager can be composed of an ultra-wideband spectrum analyzer and an infrared thermal imager. The spectrum analyzer covers a wide frequency range and uses the Time Difference of Arrival (TDoA) algorithm to locate electromagnetic radiation sources; the infrared thermal imager uses a high-resolution detector to perform real-time imaging of heat source targets above a specific temperature. When the detected electromagnetic radiation intensity exceeds a preset threshold, the system automatically expands the spectrum analysis bandwidth to improve the accuracy of threat source identification.

[0107] In this embodiment, the environmental perception unit 100 further includes a data fusion module 120, which includes a spatiotemporal registration submodule 121, a feature fusion submodule 122, and a decision fusion submodule 123, wherein:

[0108] The spatiotemporal registration submodule 121 achieves precise alignment of multi-sensor data in time and space through hardware time synchronization and homogeneous coordinate transformation;

[0109] As a further explanation of this embodiment, a precise clock synchronization protocol is used to achieve system-wide clock synchronization, with nanosecond-level timestamps generated by a dedicated chip. Spatial registration can be accomplished through a homogeneous coordinate transformation matrix, where the rotation matrix is ​​calculated based on quaternion interpolation, and the translation vector is calculated by the fusion of GPS and IMU, ensuring that data from various sensors are aligned in a unified spatiotemporal coordinate system.

[0110] The feature fusion submodule 122 uses deep learning algorithms to extract point cloud and image features and establish cross-modal data correlation.

[0111] As a further explanation of this embodiment, a cross-modal deep learning network is constructed. Point cloud data is processed by extracting geometric features using a specialized network model, while image data is processed using a semantic segmentation network. Cross-modal matching of point clouds and images is achieved through feature alignment and association algorithms. The network model is optimized based on a large amount of training data containing different scenes, achieving high-precision feature point matching and environmental semantic understanding.

[0112] The decision fusion submodule 123 integrates multi-source decision information based on evidence theory to generate an environmental situation assessment result that includes threat level and response strategy.

[0113] As a further explanation of this embodiment, this embodiment constructs a threat assessment model based on evidence theory and defines an identification framework. ={Safe, Low Threat, Medium Threat, High Threat}, and integrates multi-source decision-making information through basic trust allocation. When the threat level is determined to be medium or high threat, an environmental situation assessment result including obstacle avoidance path suggestions is automatically generated and simultaneously output to the trajectory planning unit 200 and the decision control unit 400.

[0114] As shown in Figure 4, the trajectory planning unit 200 plans a flight trajectory that meets the mission requirements, safety requirements and energy constraints based on the UAV flight environment information obtained by the environmental perception unit 100 and the pre-stored UAV performance parameters. It uses a genetic algorithm combined with a simulated annealing algorithm for offline pre-planning and uses the Dijkstra algorithm for online dynamic adjustment. The planning results are transmitted to the energy management unit 300 and the decision control unit 400.

[0115] In this step, the trajectory planning unit 200 includes a task parsing module 210, a path generation module 220, and a performance evaluation module 230, wherein:

[0116] The task parsing module 210 is used to convert the task objective into an ordered sequence of waypoints and determine the priority of each waypoint, including the following steps:

[0117] S210.1. An improved target decomposition algorithm is used to generate the waypoint sequence. The calculation formula is as follows:

[0118] ;

[0119] in, Represents the three-dimensional coordinates and time constraints of the i-th waypoint;

[0120] As a further explanation of this step, the time constraint is dynamically set according to the task type:

[0121] Emergency reconnaissance mission: The time window tolerance is set to ±5% in the example, and the latest arrival time of each waypoint is calculated by back-calculating the mission deadline.

[0122] Routine inspection tasks: The time window tolerance is widened to ±15% in the previous example, and an equal-interval time allocation strategy is adopted.

[0123] For rectangular reconnaissance areas, a checkerboard sampling method is used to generate waypoints, and the sampling interval is calculated based on camera parameters.

[0124] S210.2. Establish the task constraint set C based on the improved genetic algorithm, and use the priority weight formula. Calculate the priority of each waypoint; among which, As a task importance factor, As a factor of time urgency, As a resource consumption factor, + =1 and [0,1]; This formula achieves dynamic allocation of waypoint execution priority by comprehensively considering multiple factors.

[0125] As a further explanation of this step, the weighting coefficients are adjusted as follows in the calculation of the priority of each waypoint: In the disaster relief scenario, let... =0.5 (task importance) =0.3 (Time urgency) =0.2 (resource consumption); Daily inspection scenario adjusted to =0.3、 =0.2、 =0.5;

[0126] Furthermore, the task importance factor Mapped by task type (human rescue is 1.0, material transport is 0.7); time urgency factor. (Normalized to [0,1]); Resource consumption factor Calculations based on the mapping relationship between waypoint altitude and energy consumption (e.g., when altitude > 3000m). =0.8).

[0127] The path generation module 220 is used to construct an environmental cost map and plan flight paths. It includes the following steps:

[0128] S220.1 Constructing a 3D Environmental Cost Map Its total cost function is:

[0129] ;

[0130] in, coordinates The environmental costs resulting from topographical factors; Representing coordinates The environmental costs arising from location and meteorological conditions; coordinates The environmental costs caused by obstacles; coordinates Environmental costs caused by regional threat factors; quantifying the environmental risks of drone flight paths by overlaying the costs of terrain, weather, obstacles, and threat areas;

[0131] S220.2. A modified genetic algorithm combined with simulated annealing is used for global path planning. The modified genetic algorithm has an adaptive crossover probability. and mutation probability The calculation formula is:

[0132] ;

[0133] ;

[0134] in, , These are the maximum and average fitness of the population, respectively. Let f be the fitness of the crossover individual, and let f be the fitness of the variant individual. [0,1];

[0135] S220.3. Use an improved Dijkstra's algorithm for online dynamic adjustment, employing heuristic functions. To speed up the search; among them From the starting point to the node The actual cost, Let n be the estimated cost from node n to the target point. [0.5, 1.5] represents the weighting coefficients. The algorithm dynamically adjusts to environmental complexity; by introducing heuristic information, it achieves rapid response during dynamic path adjustments. By dynamically adjusting crossover and mutation probabilities, the efficiency and accuracy of the algorithm in global path search are improved. Offline pre-planning is performed using a combination of genetic and simulated annealing algorithms. The genetic algorithm searches for multiple potential trajectory solutions based on prior data such as terrain and weather conditions in the UAV mission area. The simulated annealing algorithm iteratively optimizes these trajectory solutions, outputting an initial trajectory set that satisfies basic constraints (such as range and no-fly zone avoidance). In the online dynamic adjustment phase, Dijkstra's algorithm uses key nodes (such as segment turning point coordinates) of each trajectory in the initial trajectory set as input, combined with real-time perceived environmental dynamic information (such as sudden obstacles and temporary no-fly zones), to quickly search for locally optimal trajectory adjustment paths, achieving online trajectory updates.

[0136] The performance evaluation module 230 is used to verify the feasibility of the flight path and predict energy consumption and risks, including the following steps:

[0137] S230.1, Based on the improved dynamic constraint verification algorithm, through formula Verify whether the turning radius of the flight path meets the minimum limit. ,in For flight speed, It is the acceleration due to gravity. The tilt angle (rad) is used to ensure the flight path conforms to the dynamic performance of the UAV.

[0138] S230.2. An improved genetic algorithm based on the energy gradient field is used to predict the trajectory energy consumption E. The calculation formula is as follows:

[0139] ;

[0140] in, Number of track segments; For segmented indexes; For the first The starting time of the flight path; For the first The end time of the flight path; It represents instantaneous power, which is related to parameters such as speed and altitude. The total energy consumption of the flight path is calculated by integration.

[0141] S230.3, Risk index update formula based on Dijkstra's algorithm Quantify flight path risks; among them, Total risk of the flight path; For a moment The location of the drone; For position The risk density function at the location enables dynamic assessment of risk throughout the entire flight path.

[0142] In this step, the path generation module 220 generates an environmental cost map. A hierarchical modeling strategy is adopted, specifically including:

[0143] Static cost layer It is constructed using the octree algorithm, integrating terrain elevation cost and obstacle cost, and the calculation formula is as follows:

[0144] ;

[0145] in, The terrain elevation cost (dimensionless, used to quantify the impact of terrain undulation on flight). Obstacle cost (dimensionless, used to mark the passage cost of an obstacle area); weighting coefficient , satisfy + =1 and [0,1] (By adjusting the weights, the influence of terrain and obstacles on path planning is balanced);

[0146] Dynamic cost layer Based on real-time updates using a meteorological field interpolation algorithm, and integrating the dynamic effects of headwind and turbulence, the calculation formula is as follows:

[0147] ;

[0148] in, The cost of headwind drag (dimensionless, reflecting the impact of headwind on flight energy consumption and speed). Cost of turbulence impact (dimensionless, quantifying the interference of turbulence on flight stability); weighting coefficient , satisfy and [0,1] (Adapting to different meteorological conditions, prioritizing headwinds and turbulence for path planning);

[0149] Threat Cost Layer The threat region is described based on a Gaussian mixture model, and the calculation formula is as follows:

[0150] ;

[0151] in, For the first Threat source intensity (dimensionless, reflecting the degree of danger of the threat source). For position To the Euclidean distances between threat sources (unit: meters, used to calculate threat attenuation). For the first The impact range parameter of a threat source (unit: meters, which determines the spatial range of the threat's effect). The total number of threat sources (a positive integer, representing the number of threat sources that need to be avoided).

[0152] The total cost of the environmental cost map is the sum of the costs from each layer, that is:

[0153] = + + This is used to provide a quantitative basis for environmental risk in flight path planning; among which, obstacle costs have been approved. Integrate into the static cost layer middle.

[0154] In this step, the performance evaluation module 230 includes a feasibility verification submodule 231, an energy consumption prediction submodule 232, and a risk assessment submodule 233, wherein:

[0155] The feasibility verification submodule 231 uses the geometric constraint method to verify whether the UAV's turning radius meets the minimum turning radius requirement for UAVs, and to verify whether the climb rate and descent rate are within the allowable range; for example, the minimum turning radius of a fixed-wing UAV. Set to 500m, when < The system will trigger a trajectory correction. At the same time, the vertical speed of the planned trajectory will be compared with the UAV's performance parameters (such as a maximum climb rate of 15 m / s and a maximum descent rate of 10 m / s). If the speed exceeds the range, the trajectory slope will be adjusted by 10%.

[0156] The energy consumption prediction submodule 232 estimates the total energy consumption of the flight path based on the energy consumption model of the segmented flight path, combined with flight speed, altitude and weather conditions.

[0157] As a further explanation of this step, the energy consumption prediction submodule 232 in this embodiment is based on a segmented flight path energy model. By dividing the complete flight path into several independent segments according to terrain and meteorological features, it achieves a refined estimation of energy consumption throughout the entire journey. Its core principle is to use a UAV dynamics model to establish a mapping relationship between flight speed, altitude, meteorological conditions (such as wind speed and air pressure), and energy consumption.

[0158] Segmented modeling strategy: Based on the changing characteristics of the environmental cost map (such as terrain undulation and sudden changes in wind speed), the flight path is divided into segments of 10-20km (which can be shortened to 5km for complex terrain), and the energy consumption of each segment is calculated independently and then accumulated.

[0159] Application of physical models: Based on the aerodynamic characteristics of fixed-wing UAVs, a three-dimensional mapping table of "velocity-altitude-power" is constructed. For example, at a cruising altitude of 3000m and a speed of 100m / s, the propulsion power is estimated by referring to empirical formulas (these formulas are mature dynamic equations in the field, derived based on lift and drag coefficients).

[0160] Meteorological impact correction: Introducing the headwind drag coefficient (range 0.1-1.0) and turbulence energy consumption compensation factor (0-0.3), when the environmental sensing unit 100 detects a wind speed ≥15m / s, it automatically calls the high wind speed energy consumption correction model (this model is based on wind tunnel test data fitting).

[0161] The function of the energy consumption prediction submodule 232 is to provide energy constraints for trajectory planning: when the predicted energy consumption exceeds 80% of the UAV's remaining power, trajectory replanning is triggered to ensure the dynamic balance of energy supply and demand during long-endurance missions.

[0162] The risk assessment submodule 233 uses a risk weight field algorithm to transform risk factors (such as terrain obstacles, weather threats, and no-fly zones) into path costs and calculate the comprehensive risk index of the flight path.

[0163] As a further explanation of this step, the risk assessment submodule 233 in this embodiment employs a risk weight field algorithm to transform risk factors such as terrain obstacles, weather threats, and no-fly zones into a cost field in three-dimensional space, and calculates the comprehensive risk index of the flight path through integration. Its core principle is as follows:

[0164] Risk factor quantification: Obstacles, strong wind areas, electromagnetic threat sources, etc. are defined as "risk sources". A decay model similar to physical fields (such as Gaussian distribution) is used to describe the decay law of risk with distance (a mature spatial risk modeling method in the field).

[0165] Multi-source risk fusion: By superimposing different types of risk fields with weighting coefficients (such as terrain obstacles 0.4, weather threats 0.3, and no-fly zones 0.3), a unified risk cost map is formed;

[0166] Dynamic update mechanism: When the environmental perception unit 100 detects a new threat source (such as a sudden obstacle), it generates a risk impact area with a radius of 5km centered on the threat source, and the risk value decays exponentially (the decay coefficient is determined by the threat type).

[0167] The output of the risk assessment submodule 233 is used to determine the priority of trajectory optimization: when the comprehensive risk index exceeds a preset threshold (such as 0.7), the weight of "risk avoidance" in path planning is automatically increased to ensure that the UAV avoids high-risk areas and improves the safety of mission execution.

[0168] Furthermore, when the environmental perception unit 100 detects a sudden threat, the trajectory planning unit 200 executes the following procedures:

[0169] First, environmental update data is received via gigabit network, and the cost map is recalculated.

[0170] Subsequently, the improved Dijkstra algorithm initiates a heuristic search. Temporarily upgraded to 1.5, prioritizing the search of paths with low threat cost;

[0171] Then, the performance evaluation module 230 completes the dynamic constraint verification of the new trajectory within 1 second. If the turning radius does not meet the requirements, the search range is expanded by 10% and replanned.

[0172] Finally, the replanned flight path is transmitted to the Energy Management Unit 300 (to update the energy consumption prediction model) and the Decision Control Unit (to adjust flight parameters) via the SPI bus, with the entire process taking ≤2 seconds.

[0173] The energy management unit 300 realizes dynamic energy management and optimization based on the real-time energy status of the UAV, the flight mission and the planning results of the trajectory planning unit 200. Based on energy monitoring, energy prediction, energy allocation and energy optimization, combined with the energy recovery and reuse mechanism, the energy management unit 300 feeds the management data back to the decision control unit 400.

[0174] In this step, the energy management unit 300 includes an energy monitoring module 310, an energy prediction module 320, an energy allocation module 330, and an energy optimization module 340, wherein:

[0175] The energy monitoring module 310 collects the state of charge, charging and discharging power and temperature parameters of the battery pack in real time through voltage sensors, current sensors and temperature sensors, and estimates the remaining battery capacity by combining the Kalman filter algorithm.

[0176] As a further explanation of this embodiment, the voltage sensor and current sensor in this embodiment acquire the battery pack terminal voltage and charging / discharging current at a frequency of kHz, and the high-frequency noise is eliminated by hardware filtering; the temperature sensor in this embodiment has multiple temperature measuring points distributed inside the battery pack, and acquires temperature data in real time using a high-frequency sampling mechanism. The signal is processed by an amplification circuit and then input to an analog-to-digital converter.

[0177] Furthermore, this embodiment employs the Extended Kalman Filter (EKF) algorithm, which estimates the State of Charge (SOC) in real time based on the battery equivalent circuit model and combined with voltage, current, and temperature data. The EKF algorithm improves the estimation accuracy through state update and observation correction mechanisms, and the estimation results are synchronously transmitted to the energy prediction module 320 and the decision control unit 400.

[0178] The energy prediction module 320 uses a long short-term memory network model trained on historical flight data and combines it with the current flight path planning results to predict the energy consumption distribution of each flight segment and generate a prediction curve for the remaining flight range.

[0179] As a further explanation of this embodiment, the energy prediction module 320 receives segmented trajectory data from the trajectory planning unit 200 and meteorological data from the environmental perception unit, extracts characteristic parameters such as airspeed, altitude change rate, and wind direction angle, and normalizes them as inputs to a two-layer long short-term memory network (LSTM). The energy prediction module 320 trains an LSTM model based on historical flight data to predict the energy consumption density of each segment, and then combines the current trajectory planning results to generate a remaining range curve, providing data support for the optimization decision of the energy allocation module 330.

[0180] Furthermore, in the energy management unit 300, an LSTM model is used to predict and regulate the UAV's energy state. The LSTM model is embedded in the FPGA of the flight control module. Leveraging the parallel computing advantages of the FPGA, it quickly processes multi-dimensional inputs such as historical flight data and energy consumption in different flight segments. Through the parameters obtained from model training, it accurately predicts the energy consumption and remaining power changes in different flight segments. In conjunction with the main and backup battery energy management strategies, when the main battery's SOC is below 10%, the backup battery replenishment logic is triggered in a timely manner to ensure a stable energy supply for the UAV's long-endurance flight.

[0181] The energy distribution module 330 optimizes flight parameters (such as airspeed and altitude) based on mission priority and real-time energy status using a dynamic programming algorithm to balance cruise energy consumption and mission completion efficiency.

[0182] As a further explanation of this embodiment, the energy allocation module 330 in this embodiment uses the current state of charge (SOC), remaining mission time, and segment index output by the energy monitoring module 310 as state variables, and flight speed and flight altitude as decision variables to construct a multi-objective optimization model. Its cost function comprehensively considers energy consumption, mission completion time, and risk index, adjusts the priority of objectives through weight coefficients, and adopts a rolling optimization strategy to solve for the optimal combination of flight parameters for the current mission stage. The optimization process satisfies constraints such as battery capacity and flight performance, and finally transmits the results to the decision control unit 400 and the energy optimization module 340 to support the dynamic adjustment of energy allocation and recovery strategies.

[0183] The energy optimization module 340 achieves efficient charging of solar panels through maximum power point tracking technology and maximizes solar energy collection efficiency by combining flight attitude adjustment algorithms.

[0184] As a further explanation of this embodiment, the energy optimization module 340 in this embodiment receives the flight parameters from the energy distribution module 330 and the battery status from the energy monitoring module 310. It tracks the peak power generation of the solar panel through maximum power point tracking (MPPT) technology and dynamically optimizes the wing's light-receiving angle by combining the flight attitude adjustment algorithm to improve the light energy collection efficiency. It uses the waste heat of the power system to drive the thermoelectric generator to recover energy and store it in the backup battery. At the same time, it dynamically adjusts the charging and discharging strategy based on the status of the main and backup batteries: that is, during the cruise phase, the energy of the main battery is released first, and during the recovery phase, the backup battery is replenished first. The adjustment results are fed back to the decision control unit 400 to support the dynamic optimization of energy flow.

[0185] In this embodiment, the energy recovery and reuse mechanism of the energy management unit 300 specifically includes:

[0186] When the trajectory planning unit 200 outputs trajectory adjustment commands, the energy management unit 300 monitors the charge status of the main and backup batteries in real time. During the cruise phase, the main battery prioritizes powering the UAV. When the trajectory enters a recyclable energy scenario (such as the gliding phase), it automatically switches to charging the backup battery to store redundant energy during flight. The charging and discharging switching is based on a simple power threshold preset by the flight controller (the backup battery is activated to replenish power when the main battery SOC is low to about 10%, and the backup battery is paused when the backup battery SOC is high to about 90%), balancing battery usage and energy retention.

[0187] As a further explanation of this embodiment, the energy management unit 300 in this embodiment ensures battery safety and energy utilization by switching charge and discharge through a simple power threshold:

[0188] Main battery recharge trigger: When the main battery's state of charge (SOC) drops to around 10%, the backup battery is activated for recharge. This threshold reserves emergency power (the main battery needs to maintain basic power supply for critical systems such as flight control and navigation) to prevent damage to the battery from deep discharge.

[0189] Backup battery charging pause: Charging is paused when the backup battery's SOC reaches approximately 90%. This threshold prevents the backup battery (such as a lithium battery) from overheating due to full charging and covers the storage needs of most energy recovery scenarios.

[0190] The threshold can be configured directly through the flight control ground station software (e.g., input "main battery charging threshold: 10" or "backup battery pause threshold: 90") to adapt to different battery types.

[0191] Furthermore, after the trajectory planning unit 200 outputs the trajectory adjustment command in this embodiment, the energy management unit 300 works in conjunction with the flight control subsystem to identify energy recovery scenarios such as the "gliding phase" in the following ways:

[0192] Real-time parameter acquisition: The flight control subsystem acquires 2 sets of core parameters per second.

[0193] Altitude change rate: Calculated by GPS or barometer, reflecting the drone's ascent and descent trend;

[0194] Engine thrust command: Output by the powertrain ECU, reflecting the active power load.

[0195] Scene triggering conditions: When the following conditions are met for 3 consecutive seconds, it is determined to be an energy recovery scene:

[0196] The altitude change rate remains <0 (the drone is descending).

[0197] Engine thrust command remains below 10% of full thrust (low load on the power system).

[0198] This logic is designed based on the gliding characteristics of fixed-wing UAVs, and the flight control parameter interface allows for customizable thresholds (e.g., for heavy-load UAVs, the thrust command threshold can be adjusted to 15%).

[0199] Furthermore, in energy recovery scenarios (such as the gliding phase), energy conversion can be achieved through the following existing hardware collaborations:

[0200] Mechanical energy harvesting: When the drone glides, gravitational potential energy is converted into flight kinetic energy, which drives the propeller to rotate passively (the engine is idling, and the propeller is coaxially connected with the generator).

[0201] Energy conversion and storage: The propeller rotation drives the generator to generate AC power, which is rectified and regulated by the flight control power management module and stored in the backup battery in the form of DC power.

[0202] This process relies on the inherent power structure of fixed-wing UAVs (propellers, generators, batteries) and requires no additional complex modifications. Those skilled in the art can reproduce it through flight control code (such as Ardupilot battery management logic).

[0203] As shown in Figure 5, the decision control unit 400 receives data from the environmental perception unit 100, the trajectory planning unit 200, and the energy management unit 300, and performs information fusion based on an adaptive weighted fusion algorithm. Based on the fused information, it generates flight status control commands and mission execution control commands using a hierarchical decision-making strategy. Through the flight status control submodule 431 and the mission execution control submodule 432, it uses PID control technology to drive the UAV's actuators and mission execution equipment.

[0204] In this embodiment, the decision control unit 400 includes an information fusion module 410, a decision generation module 420, and an instruction execution module 430, wherein:

[0205] The information fusion module 410 is used to purify multi-source data and dynamically allocate fusion weights to ensure the accuracy and real-time nature of decision input. The information fusion module 410 includes a data preprocessing submodule 411 and a dynamic weight allocation submodule 412, wherein:

[0206] Data preprocessing submodule 411 uses the Kalman filter algorithm to reduce noise in environmental perception data, aligns trajectory planning data through timestamp synchronization technology, and uses threshold detection method to filter out outliers in energy management data;

[0207] As a further explanation of this embodiment, in this embodiment, data such as lidar point clouds and visual images first enter the data preprocessing submodule 411, where a Kalman filter algorithm (a recursive estimation algorithm based on state prediction and observation updates) is used to eliminate noise. For example, for the lidar's three-dimensional coordinate data, the filtering intensity is dynamically adjusted through the process noise covariance matrix to ensure smooth point cloud coordinates. At the same time, the path point data of the trajectory planning unit 200 and the SOC data of the energy management unit 300 are synchronized with the clock through the NTP protocol. For asynchronous data, a linear interpolation method is used to generate synchronized data points within the time window. In addition, energy data is filtered for fluctuation anomalies (such as when SOC suddenly exceeds a preset range) through a threshold detection method, and a weighted average of historical effective values ​​is used to replace them to ensure the reliability of the data input for decision-making.

[0208] The dynamic weight allocation submodule 412 constructs a confidence model based on the sensor's historical error rate and current working status, and dynamically adjusts the fusion ratio in conjunction with the task stage weight matrix.

[0209] The decision generation module 420 generates global strategies and control commands based on fused data. It uses Markov decision process to build a decision model and uses UAV position, remaining energy, and mission progress as state variables to generate global strategies. It also uses model predictive control algorithm to continuously optimize flight parameters in a preset prediction time domain and outputs trajectory tracking commands in combination with UAV dynamic constraints.

[0210] As a further explanation of this embodiment, the state space is constructed using the UAV's three-dimensional position, remaining energy (SOC), and mission progress (waypoint completion percentage). Speed ​​and heading adjustments are defined as the action space. Objectives such as maintaining safe distance, improving mission efficiency, and optimizing energy consumption are quantified through a reward function. An iterative value algorithm is used to solve for the optimal strategy, balancing current decisions with future gains (typically 0.95) during the iteration process. For example, when the energy management unit 300 issues a low battery warning, the MDP model prioritizes outputting the global strategy of "returning to the nearest landing point," clarifying the long-term decision direction and providing guidance for subsequent control optimization.

[0211] Command execution module 430 is used to convert decision commands into controllable signals for the actuators, ensuring safe and accurate execution. Command execution module 430 includes flight status control submodule 431 and mission execution control submodule 432, wherein:

[0212] The flight status control submodule 431 adopts a PID+feedforward composite control structure. The feedforward part pre-calculates the ideal control quantity based on the UAV dynamics model, and the feedback part performs PID adjustment through real-time error, outputting flight control signals including attitude angle and throttle value.

[0213] As a further explanation of this embodiment, the PID + feedforward composite control strategy adopted in this embodiment specifically includes:

[0214] Feedforward mechanism: Based on the UAV dynamic model (such as the six-degree-of-freedom motion equation), ideal control quantities such as pitch angle and throttle opening are pre-calculated to compensate for known disturbances (such as airflow and load changes) in advance.

[0215] Feedback mechanism: Real-time correction of state errors (such as altitude and attitude deviations) using a PID algorithm, with typical parameters (such as altitude channel) being adjusted accordingly. , , Based on the device model and task scenario, the integral limit avoids the accumulation of long-term errors.

[0216] Furthermore, the control signal is converted from digital to analog and outputs a 1000–2000μs PWM signal to drive the ESC; at the same time, the flight control cycle and the MPC optimization process of the decision generation module 420 are kept synchronized at a frequency of 10Hz to ensure real-time connection between command updates and sensor status feedback, supporting dynamic trajectory tracking.

[0217] The task execution control submodule 432 presets a device load safety threshold. When the instruction exceeds the threshold, a proportional attenuation strategy is triggered and fed back to the decision generation module 420 to adjust the strategy.

[0218] As a further explanation of this embodiment, the preset device load safety threshold in this embodiment includes: preset safety thresholds such as robotic arm torque and camera gimbal angular velocity (e.g., maximum robotic arm load and maximum gimbal rotation rate), which serve as the execution boundary of the task instruction;

[0219] When the decision instruction exceeds the threshold:

[0220] First, adjust the execution amount using a linear decay function (e.g., if the command torque exceeds the limit, decay it proportionally to a safe range).

[0221] Synchronously, abnormal information is fed back to the decision generation module 420 via the bus, triggering the strategy replanning mechanism (such as reducing task load priority and adjusting execution timing) to achieve closed-loop control of execution-feedback-decision correction.

[0222] Understandably, during the entire command execution process, real-time data on flight status and mission execution are continuously transmitted back to the decision generation module 420, providing a basis for the dynamic optimization of control strategies and ensuring the safety and accuracy of UAV flight and mission execution.

[0223] The parameter storage unit 500 is used to store UAV performance parameters (including but not limited to maximum flight speed, maximum ceiling, fuel consumption rate, power system power, etc.) for the trajectory planning unit 200 to call.

[0224] As a further explanation of this embodiment, the parameter storage unit 500 in this embodiment uses industrial-grade SPI flash memory (such as W25Q128, with a capacity of 128Mbit); and the parameter storage unit 500 can store the core parameters of the UAV according to functional modules, including:

[0225] Flight performance parameters: maximum flight speed, maximum service ceiling, minimum turning radius, etc.;

[0226] Powertrain parameters: rated power, fuel consumption rate (internal combustion engine), battery capacity (electric motor type), etc.

[0227] Structural constraint parameters: unloaded mass, moment of inertia, upper limit of task load, etc.

[0228] As a further explanation of this step,

[0229] Furthermore, the parameter storage unit 500 interacts with the trajectory planning unit 200 and the energy management unit 300 via the CAN bus:

[0230] First, before path planning, the trajectory planning unit 200 sends a parameter query command. After parsing the command, the parameter storage unit 500 locates the address from the index table, reads the parameter value and check code, and sends it back to the trajectory planning unit 200.

[0231] Update process: The ground station sends a parameter update packet (containing new values, timestamps, and checksums) via USB / CAN. After the parameter storage unit passes the 500 check, it overwrites the original data and updates the index table version number.

[0232] Example 2

[0233] As shown in Figure 2, this embodiment also provides an intelligent trajectory planning and energy management method for long-endurance fixed-wing UAVs. Based on the intelligent trajectory planning and energy management system for long-endurance fixed-wing UAVs in Embodiment 1, it includes the following steps:

[0234] S100: Through the multi-sensor fusion technology of the environmental perception unit, real-time data on terrain, weather conditions, obstacle distribution and threat areas are collected. After spatiotemporal registration, feature fusion and decision fusion processing, environmental situation assessment results are generated.

[0235] S200: Based on environmental situation assessment results and UAV performance parameters, offline trajectory pre-planning is carried out using a combination of genetic algorithm and simulated annealing algorithm. Online dynamic adjustment is achieved through Dijkstra algorithm. The feasibility of the trajectory is verified by turning radius constraint, energy consumption prediction and risk assessment.

[0236] S300 collects battery parameters and estimates remaining capacity through the energy monitoring module based on real-time flight track and energy status, uses a long short-term memory network to predict energy consumption for each flight segment, optimizes energy allocation using a dynamic programming algorithm, dynamically switches between primary and backup battery charging and discharging based on flight track scenarios, and achieves energy reuse by relying on flight redundancy energy storage and power threshold control.

[0237] The S400 inputs environmental perception, trajectory planning, and energy management data into the decision control unit. After processing by an adaptive weighted fusion algorithm, it generates a global mission strategy using a Markov decision process. The system then outputs flight control commands through model predictive control and PID regulation to drive the actuators to complete the mission.

[0238] The S500 stores and updates UAV performance parameters (such as maximum flight speed and fuel consumption rate) in real time through a parameter storage unit, which can be dynamically called by the trajectory planning unit during the trajectory planning process of the S200.

[0239] Those skilled in the art will understand that the process of implementing all or part of the steps of the above embodiments can be carried out by hardware or by a program instructing the relevant hardware.

[0240] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles, characterized in that, include: An environmental perception unit (100) is used to acquire UAV flight environment information in real time. Based on multi-sensor fusion technology, it integrates a data acquisition module (110) to collect data on terrain, meteorological conditions, obstacle distribution, and potential threat areas, and transmits the data to a trajectory planning unit (200) and an energy management unit (300). The trajectory planning unit (200) plans a flight trajectory that meets mission requirements, safety requirements, and energy constraints based on the UAV flight environment information acquired by the environmental perception unit (100) and pre-stored UAV performance parameters. It uses a genetic algorithm combined with a simulated annealing algorithm for offline pre-planning and uses Dijkstra's algorithm for online dynamic adjustment. The planning results are transmitted to the energy management unit (300) and a decision control unit (400). The energy management unit (300) determines the flight path based on the UAV's real-time energy status, The planning results of the flight mission and trajectory planning unit (200) realize dynamic energy management and optimization. Based on energy monitoring, energy prediction, energy allocation and energy optimization, combined with the energy recovery and reuse mechanism, the management data is fed back to the decision control unit (400). The decision control unit (400) is used to receive data from the environmental perception unit (100), the trajectory planning unit (200) and the energy management unit (300), and perform information fusion based on the adaptive weighted fusion algorithm. Based on the fused information, a hierarchical decision strategy is adopted to generate flight status control commands and mission execution control commands. Through the flight status control submodule (431) and the mission execution control submodule (432), PID regulation technology is used to drive the UAV actuator and mission execution equipment. The parameter storage unit (500) is used to store UAV performance parameters for the trajectory planning unit (200) to call.

2. The intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles according to claim 1, characterized in that, The data acquisition module (110) includes a terrain and landform acquisition submodule (111), a meteorological parameter acquisition submodule (112), an obstacle detection submodule (113), and a threat perception submodule (114), wherein: the terrain and landform acquisition submodule (111) uses lidar and optical camera data fusion technology to generate a three-dimensional terrain model; the meteorological parameter acquisition submodule (112) includes a micro weather station and a turbulence sensor to collect wind speed, air pressure, temperature and humidity, and turbulence intensity data in real time; the obstacle detection submodule (113) uses millimeter-wave radar and binocular vision multi-sensor fusion to realize dynamic obstacle detection within a preset range; the threat perception submodule (114) uses electronic reconnaissance equipment and infrared thermal imager to detect the location and intensity of electromagnetic radiation sources and heat source targets.

3. The intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles according to claim 1, characterized in that, The environmental perception unit (100) further includes a data fusion module (120), which includes a spatiotemporal registration submodule (121), a feature fusion submodule (122), and a decision fusion submodule (123). The spatiotemporal registration submodule (121) achieves accurate alignment of multi-sensor data in time and space through hardware time synchronization and homogeneous coordinate transformation. The feature fusion submodule (122) uses deep learning algorithms to extract point cloud and image features and establish cross-modal data correlation. The decision fusion submodule (123) fuses multi-source decision information based on evidence theory to generate an environmental situation assessment result that includes threat level and response strategy.

4. The intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles according to claim 1, characterized in that, The trajectory planning unit (200) includes a task parsing module (210), a path generation module (220), and a performance evaluation module (230), wherein: the task parsing module (210) is used to convert the task objective into an ordered sequence of waypoints and determine the priority of each waypoint, including the following steps: S210.1, generating a waypoint sequence using an improved objective decomposition algorithm. The calculation formula is: ;in, Represent the three-dimensional coordinates and time constraints of the i-th waypoint; S210.2, establish the task constraint set C based on the improved genetic algorithm, and use the priority weight formula... Calculate the priority of each waypoint; among which, As a task importance factor, As a factor of time urgency, As a resource consumption factor, + =1 and [0,1]; The path generation module (220) is used to construct an environmental cost map and plan a flight path, including the following steps: S220.1, constructing a three-dimensional environmental cost map. Its total cost function is: ;in, coordinates The environmental costs resulting from topographical factors; Representing coordinates The environmental costs arising from location and meteorological conditions; coordinates The environmental costs caused by obstacles; coordinates Environmental costs caused by regional threat factors; S220.2, global path planning using an improved genetic algorithm combined with simulated annealing algorithm, wherein the improved genetic algorithm has an adaptive crossover probability. and mutation probability The calculation formula is: ； ;in, 、 These are the maximum and average fitness of the population, respectively. Let f be the fitness of the crossover individual, and let f be the fitness of the variant individual. [0,1];S220.3、Use the improved Dijkstra algorithm for online dynamic adjustment, through heuristic functions To speed up the search; among them From the starting point to the node The actual cost, Let n be the estimated cost from node n to the target point. [0.5, 1.5] are weighting coefficients; the performance evaluation module (230) is used to verify the feasibility of the flight path and predict energy consumption and risk, including the following steps: S230.1, based on the improved dynamic constraint verification algorithm, through the formula Verify whether the turning radius of the flight path meets the minimum limit. ,in For flight speed, It is the acceleration due to gravity. S230.

2. An improved genetic algorithm based on the energy gradient field is used to predict the trajectory energy consumption E, and the calculation formula is as follows: ;in, Number of track segments; For segmented indexes; For the first The starting time of the flight path; For the first The end time of the flight path; For instantaneous power; S230.3, update formula for risk index based on Dijkstra's algorithm. Quantify flight path risks; among them, Total risk of the flight path; For a moment The location of the drone; For position The risk density function at the location enables dynamic assessment of risk throughout the entire flight path.

5. The intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles according to claim 4, characterized in that, The environmental cost map of the path generation module (220) A layered modeling strategy is adopted, specifically including: a static cost layer. It is constructed using the octree algorithm, integrating terrain elevation cost and obstacle cost, and the calculation formula is as follows: ;in, For the sake of terrain elevation, Cost of obstacles; weighting coefficient 、 satisfy + =1 and [0,1]; Dynamic cost layer Based on real-time updates using a meteorological field interpolation algorithm, and integrating the dynamic effects of headwind and turbulence, the calculation formula is as follows: ;in, Paying the price for headwind resistance, Cost of turbulence effects; weighting coefficient 、 satisfy and [0,1]; Threat Cost Layer The threat region is described based on a Gaussian mixture model, and the calculation formula is as follows: ;in, For the first The strength of each threat source, For position To the Euclidean distance of each threat source For the first The impact range parameter of each threat source This represents the total number of threat sources; the total cost of the environmental cost map is the sum of the costs from each layer, i.e.: = + + This is used to provide a quantitative basis for environmental risk in flight path planning; among which, obstacle costs have been approved. Integrate into the static cost layer middle.

6. The intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles according to claim 4, characterized in that, The performance evaluation module (230) includes a feasibility verification submodule (231), an energy consumption prediction submodule (232), and a risk assessment submodule (233), wherein: the feasibility verification submodule (231) uses the geometric constraint method to verify whether the turning radius of the UAV trajectory meets the minimum turning radius requirement of the UAV, and to verify whether the climb rate and descent rate are within the allowable range; the energy consumption prediction submodule (232) estimates the energy consumption of the entire trajectory based on the energy consumption model of the segmented trajectory, combined with flight speed, altitude, and weather conditions; the risk assessment submodule (233) uses the risk weight field algorithm to convert risk factors into path costs and calculate the comprehensive risk index of the trajectory.

7. The intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles according to claim 1, characterized in that: The energy management unit (300) includes an energy monitoring module (310), an energy prediction module (320), an energy allocation module (330), and an energy optimization module (340), wherein: the energy monitoring module (310) collects the state of charge, charging and discharging power, and temperature parameters of the battery pack in real time through voltage sensors, current sensors, and temperature sensors, and estimates the remaining battery capacity in combination with the Kalman filter algorithm; the energy prediction module (320) predicts the energy consumption distribution of each flight segment based on a long short-term memory network model trained on historical flight data, combined with the current flight path planning results, and generates a prediction curve of the remaining flight range; the energy allocation module (330) optimizes flight parameters using a dynamic programming algorithm according to task priority and real-time energy status, balancing cruise energy consumption and task completion efficiency; the energy optimization module (340) achieves efficient charging of solar panels through maximum power point tracking technology, and maximizes solar energy collection efficiency in combination with flight attitude adjustment algorithm.

8. The intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles according to claim 1, characterized in that, The energy management unit (300) energy recovery and reuse mechanism specifically includes: when the trajectory planning unit (200) outputs trajectory adjustment instructions, the energy management unit (300) monitors the charge status of the main and backup batteries in real time; during the cruise phase, the main battery prioritizes powering the UAV, and when the trajectory enters a recyclable energy scenario, it automatically switches to charging the backup battery to store redundant energy during flight; the charging and discharging switching is based on a simple power threshold preset by the flight controller to balance battery use and energy retention.

9. The intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles according to claim 1, characterized in that, The decision control unit (400) includes an information fusion module (410), a decision generation module (420), and an instruction execution module (430). The information fusion module (410) is used to purify multi-source data and dynamically allocate fusion weights. The information fusion module (410) includes a data preprocessing submodule (411) and a dynamic weight allocation submodule (412). The data preprocessing submodule (411) uses a Kalman filter algorithm to reduce noise in environmental perception data, aligns trajectory planning data using timestamp synchronization technology, and uses a threshold detection method to filter out abnormal values ​​in energy management data. The dynamic weight allocation submodule (412) constructs a confidence model based on the sensor's historical error rate and current operating status, and dynamically adjusts the fusion ratio in conjunction with the task phase weight matrix. The decision generation module (420) generates global strategies and control instructions based on the fused data and constructs a decision model using a Markov decision process. The system uses the UAV's position, remaining energy, and mission progress as state variables to generate a global strategy; and uses a model predictive control algorithm to continuously optimize flight parameters within a preset prediction time domain, and outputs a trajectory tracking command in combination with UAV dynamic constraints; the command execution module (430) is used to convert decision commands into controllable signals for the actuators. The command execution module (430) includes a flight state control submodule (431) and a mission execution control submodule (432), wherein: the flight state control submodule (431) adopts a PID+feedforward composite control structure, the feedforward part pre-calculates the ideal control quantity based on the UAV dynamic model, and the feedback part performs PID adjustment through real-time error, outputting a flight control signal containing attitude angle and throttle value; the mission execution control submodule (432) presets a device load safety threshold, and triggers a proportional decay strategy when the command exceeds the threshold and feeds it back to the decision generation module (420) to adjust the strategy.

10. A method for intelligent trajectory planning and energy management of long-endurance fixed-wing unmanned aerial vehicles, based on the intelligent trajectory planning and energy management system for long-endurance fixed-wing unmanned aerial vehicles as described in any one of claims 1-9, characterized in that, The process includes the following steps: S100, using the multi-sensor fusion technology of the environmental perception unit (100), real-time data on terrain, weather conditions, obstacle distribution, and threat areas are collected. After spatiotemporal registration, feature fusion, and decision fusion, an environmental situation assessment result is generated; S200, based on the environmental situation assessment result and UAV performance parameters, offline trajectory pre-planning is performed using a genetic algorithm combined with simulated annealing algorithm. Online dynamic adjustment is achieved through Dijkstra's algorithm. The feasibility of the trajectory is verified through turning radius constraints, energy consumption prediction, and risk assessment; S300, based on the real-time trajectory and energy status, battery parameters are collected through the energy monitoring module (310) and the remaining capacity is estimated. Long short-term memory is used to... The network predicts the energy consumption of each flight segment, uses dynamic programming algorithm to optimize energy allocation, and dynamically switches the charging and discharging of the main and backup batteries in combination with the flight path scenario. It relies on flight redundancy energy storage and power threshold control to realize energy reuse. In S400, environmental perception, flight path planning and energy management data are input into the decision control unit (400). After processing by the adaptive weighted fusion algorithm, the global task strategy is generated by the Markov decision process. The flight control command is output through model predictive control and PID adjustment to drive the actuator to complete the task execution. In S500, the UAV performance parameters are stored and updated in real time through the parameter storage unit (500) for dynamic calling by the flight path planning unit (200) in the flight path planning process in S200.

Images