Unmanned aerial vehicle operation data management method and system based on flight safety
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING COMM INST OF TECH
- Filing Date
- 2026-03-16
- Publication Date
- 2026-06-19
AI Technical Summary
Existing drone return-to-home strategies suffer from high collision risks, inaccurate energy consumption estimates, and an inability to comprehensively assess risk probabilities and energy efficiency in complex environments, resulting in poor robustness and adaptability.
By collecting historical logs, environmental data, and image data, a three-dimensional mesh space is constructed. The YOLO object detection algorithm is used to identify obstacles. Combined with the weighted average method and predicted power, safe and feasible return routes are calculated and screened. The efficiency index is used to make the optimal decision.
It enables safe, efficient, and autonomous return of drones in complex environments, dynamically adapts to energy consumption predictions, avoids collision risks, and achieves a balance between safety and economy.
Smart Images
Figure CN122239738A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of unmanned aerial vehicle (UAV) control technology, specifically to a method and system for managing UAV operational data based on flight safety. Background Technology
[0002] With the large-scale application of drones in logistics, inspection, agriculture, and other fields, their operating environments are becoming increasingly complex, placing higher demands on flight safety and autonomy. Especially in offline states where signals are interrupted, ensuring the safe return of drones has become a key technological bottleneck restricting their reliability and application scope.
[0003] Currently, existing technologies primarily rely on pre-set fixed routes and simple obstacle avoidance for return-to-home operations, which has significant drawbacks. First, return-to-home path planning is often based on two-dimensional maps or simple elevation models, ignoring three-dimensional obstacles in the real environment, especially in complex urban airspace or areas with undulating terrain, leading to a high risk of collisions during actual flight. Second, energy consumption estimation mostly uses static theoretical models, failing to fully consider the additional power consumption caused by real-time wind conditions, airflow changes, and flight attitude adjustments, making dynamic correction impossible and prone to errors in power consumption prediction, potentially even leading to forced landings or loss of contact. Finally, path selection typically uses only straight-line distance or estimated flight time as a single optimization objective, without comprehensively assessing multiple dimensions such as risk probability, energy efficiency, and environmental uncertainties. Therefore, it is impossible to achieve an optimal trade-off between safety and economy, resulting in poor robustness and adaptability of the overall return-to-home strategy, making it difficult to cope with unexpected situations and changing conditions in real-world scenarios. Summary of the Invention
[0004] The purpose of this invention is to provide a method and system for managing unmanned aerial vehicle (UAV) operation data based on flight safety, so as to solve the problems mentioned in the background art.
[0005] To address the aforementioned technical problems, this invention provides a method for managing unmanned aerial vehicle (UAV) operational data based on flight safety, comprising:
[0006] After the S100 drone goes offline, it collects historical logs, operational data, environmental data, and image data.
[0007] "Offline" refers to a break in the signal connection between the drone and the ground control unit. Historical logs contain data records for each flight, specifically operational and environmental data.
[0008] Operational data includes battery capacity, launch location, battery level sequence, location sequence, and attitude log.
[0009] The takeoff position refers to the spatial location of the drone at takeoff. The battery sequence consists of the remaining battery power at different times. The position sequence consists of the three-dimensional coordinates at different times.
[0010] Attitude logs are used to record the drone's flight attitude, flight maneuvers, and duration at different times.
[0011] Flight attitude refers to the angular states of pitch, roll, and yaw. Flight maneuvers include at least level flight, climb, descent, hovering, and turning.
[0012] Attitude logs discretize continuous flight attitudes into specific flight actions, providing structured data for subsequent energy consumption analysis based on actions.
[0013] Environmental data refers to various types of meteorological indicators around the drone collected at different times through various airborne sensors. The types of meteorological indicators include at least temperature, humidity, air density, and wind vector.
[0014] Image data refers to video images of the drone's surroundings captured by the onboard camera from the start of takeoff.
[0015] Multimodal data fusion of state, environment, and vision is a prerequisite for subsequent intelligent environment modeling and accurate energy consumption prediction.
[0016] This established a complete and reliable offline data foundation for the entire autonomous return decision-making system.
[0017] Ensuring that all necessary analytical data is available after the drone loses contact with the ground station enables data autonomy in decision-making.
[0018] S200 analyzes the data records contained in the historical logs, calculates the weight index of each data record by combining operational and environmental data, and calculates the reference power of various flight maneuvers. A weighted average method is then used to calculate the predicted power of various flight maneuvers. Specifically, this includes:
[0019] S201. Retrieve past duration from historical logs. All data records are analyzed, and the environmental data in each record is compared with the current environmental data to calculate the difference index for each record. Specifically:
[0020] Obtain the values of various meteorological indicators from the current environmental data. Preset weighting coefficients for various meteorological indicators Substitute the values into the formula to calculate the difference index for each data record. :
[0021] ;
[0022] In the formula, As a preset constant, The number of all meteorological indicator types. For the environmental data in the data record Values for various meteorological indicators.
[0023] By calculating the difference index It quantifies the similarity between historical experience and the current scenario, thus achieving environmental perception.
[0024] S202. Mark data records where the difference index is less than the threshold, according to the formula: Calculate the weight index for each labeled data record. .in, The index representing the largest difference among all labeled data records.
[0025] Through weight index This approach assigns higher weight to historical data that is more similar to the current environment, reflecting the recency effect and the principle of environmental adaptation.
[0026] S203. Analyze the power sequence and attitude log in each marked data record, analyze the duration of each flight maneuver within each preset unit of power, and calculate the reference power for each flight maneuver in each marked data record. Specifically, this includes:
[0027] S2031. Obtain the marked data record. The power consumption sequence and attitude log are analyzed according to a preset unit. The time period corresponding to each unit of power consumed in the power consumption sequence is used as the sampling area, and the number of sampling areas is counted. .
[0028] S2032, Analyze the number of flight maneuver types included in the attitude log. Each flight maneuver and its duration are mapped to a sampling area, and a power range is preset for each flight maneuver.
[0029] S2033, Establishment There are several plans, each containing... The power is randomly selected from different power ranges, and the corresponding flight maneuver is also selected. The power of all flight maneuvers in different schemes is not exactly the same.
[0030] S2034, Obtaining the Solution The power of each flight maneuver was analyzed in the sampling area. The included flight maneuvers and their corresponding durations are substituted into the formula to calculate the solution. downsampling area Theoretical energy consumption :
[0031] ;
[0032] In the formula, The preset basic energy consumption, Sampling area The number of types of flight maneuvers included.
[0033] Sampling area The included first Types of flight maneuvers in the plan The corresponding power, For the first Types of flight maneuvers in the sampling area The duration corresponding to the time interval within the period.
[0034] S2035, Calculate the schemes separately The theoretical energy consumption of each sampling area is summed to form the scheme. The theoretical total energy consumption is calculated. Similarly, the theoretical total energy consumption of each scheme is calculated separately.
[0035] S2036. Analyze the electrical energy corresponding to each unit of battery capacity based on the battery capacity, and multiply by... The total electrical energy was then obtained. Calculate the theoretical total energy consumption and total electrical energy for each scheme. The absolute values between them are used to select the power values of each item in the scheme with the smallest absolute value as the reference power for the corresponding flight maneuver.
[0036] The preset unit is pre-set, typically using milliampere-hours (mAh), the basic unit of battery capacity, as a multiple. The multiple is determined based on the rate of power consumption during actual use.
[0037] By generating random schemes within a preset power range, and calculating the difference between the total theoretical energy consumption and the actual total electrical energy under each scheme, the most suitable scheme is selected to deduce the power of each action.
[0038] By using data fitting and optimization of the search process, the difficulty of building accurate physical models for complex actions is avoided, and power values are learned directly from historical data.
[0039] S204. Using the weighted average method, the predicted power of the flight maneuver is calculated based on the reference power of the flight maneuver in each marked data record and the corresponding weight index. The predicted power of each flight maneuver is calculated separately.
[0040] By using a data-driven approach, energy consumption prediction models can automatically adapt to specific environmental conditions, rather than relying on fixed theoretical values, greatly improving the reliability and personalization of predictions. This enables dynamic and accurate modeling of energy consumption for different UAV flight maneuvers.
[0041] S300, construct a three-dimensional mesh space and identify the mesh passage status based on image data, establish A number of return routes are selected, and the estimated total energy consumption of each route is calculated based on the flight maneuvers involved and their predicted power, thereby enabling route selection. Specifically, this includes:
[0042] S301. A three-dimensional mesh space is constructed using the UAV's current position and takeoff position as diagonals. After temporally aligning the image data and position sequence, obstacles in the image are identified and detected, and passable and impassable meshes are obtained through analysis. Specifically, this includes:
[0043] S3011. Obtain the three-dimensional coordinates of the drone's current position and takeoff position, take the straight line between the two points as the diagonal, and set a preset safe passage distance. Using the basic side length, construct a three-dimensional boundary space that completely includes the diagonal.
[0044] S3012, A three-dimensional mesh discretization algorithm is used to discretize the three-dimensional boundary space along the three dimensions of longitude, latitude, and elevation. The step size is uniformly divided to generate a series of continuous cubic mesh units, thus forming a three-dimensional mesh space.
[0045] A three-dimensional mesh space is constructed with safe passage distance as the granularity, achieving a balance between computational complexity and environmental resolution.
[0046] S3013. Extract image data and location sequences, and deframe the video images in the image data. Analyze the timestamp corresponding to each frame, and use a linear interpolation algorithm to resample the three-dimensional coordinates at the corresponding time in the location sequence.
[0047] S3014, For each frame image With each three-dimensional coordinate Perform temporal alignment to obtain a spatiotemporally aligned sequence. .for For each data set in the sequence, perform traffic status analysis:
[0048] First, Input a pre-trained model based on the YOLO object detection algorithm to identify and box out all obstacles in the image.
[0049] Then, with Using the viewpoint center, the three-dimensional spatial occupancy of each identified obstacle is calculated based on camera intrinsic parameters and flight attitude.
[0050] Finally, determine which cubic mesh cells in the 3D mesh space are occupied by obstacles at that viewpoint. Occupied cubic mesh cells are designated as impassable meshes, while unoccupied cubic mesh cells are designated as passable meshes.
[0051] By aligning image frames with GPS / IMU location data in a time sequence and using the YOLO target detection algorithm, obstacles in 2D images are mapped onto a 3D mesh in real time, marking them as passable / impassable, thus dynamically constructing a digital twin of the environment.
[0052] S302. In the three-dimensional grid space, starting from the current position of the UAV and ending at the departure position, establish a new location after avoiding impassable grids. The study analyzed the flight maneuvers required for each return route and their duration.
[0053] S303. The predicted energy consumption is obtained by multiplying the predicted power of each flight maneuver by the duration. The predicted total energy consumption is obtained by summing the predicted energy consumption of all flight maneuvers in a single return route. The predicted total energy consumption is calculated for each return route.
[0054] S304, Preset safety factor The remaining battery level is obtained based on the battery power sequence, and the corresponding remaining energy is analyzed in conjunction with the battery capacity before being multiplied by [the remaining energy]. Obtain the energy threshold. Select return routes whose total expected energy consumption is less than the energy threshold.
[0055] The safety factor ranges from 0 to 1 and measures the degree of redundancy in battery power for the drone during actual operation. A higher value indicates less redundancy, and vice versa.
[0056] In path planning, not only is geometric obstacle avoidance considered, but a safety factor is also introduced to set the power threshold and screen the route for energy consumption.
[0057] Ensuring that all candidate routes are energy-conservatively feasible eliminates risky routes that cannot be completed for the final decision, transforming the multi-objective optimization problem into a problem of finding the best solution in the feasible solution set.
[0058] The continuous physical flight space is discretized into a computable digital model, and a preliminary set of physically feasible and energy-safe return routes is planned. This achieves dual verification of environmental digitization and path feasibility.
[0059] S400: Analyze the traffic conditions of all grids traversed by each return route, and calculate the efficiency index based on the estimated total energy consumption. Select the optimal return route based on the efficiency index, and control the UAV to perform the autonomous return mission according to the optimal route. Specifically, this includes:
[0060] S401. Analyze the grids traversed by each return route and count the total number of grids traversed. and the number of passable grid cells within them. The efficiency index of each return route was calculated. The specific calculation formula is as follows:
[0061] ;
[0062] In the formula, and It is a constant. The electrical energy threshold, This represents the estimated total energy consumption.
[0063] This measure assesses energy economy; a higher value means more redundant power remaining after the flight, and thus greater safety.
[0064] This measure assesses the accessibility of a route; a higher value means that the route passes through a higher proportion of passable areas and the lower the risk of encountering unknown or dynamic obstacles.
[0065] constant and It is a trade-off parameter that allows for adjustments to the emphasis on energy consumption and safety based on actual needs.
[0066] By calculating and comparing all candidate return routes The system automatically selects the route with the highest overall score.
[0067] This quantitative decision-making mechanism enables drones to mimic comprehensive thinking in complex offline environments and execute energy-efficient and reliable return routes.
[0068] S402. Select the return route with the highest efficiency index as the optimal return route, and control the drone to perform the autonomous return mission according to the selected optimal return route.
[0069] Among multiple feasible return routes, an optimal decision is made based on a comprehensive quantitative indicator, driving the drone to execute the route. This achieves an intelligent trade-off between safety and economy, ultimately selecting the best route from the feasible ones.
[0070] The present invention also provides a flight safety-based UAV operation data management system, including a data acquisition module, a flight analysis module, a safety management module, and a return-to-home execution module.
[0071] The data acquisition module is used to collect historical logs, runtime data, environmental data, and image data.
[0072] When the drone loses signal with the ground control terminal, it will comprehensively collect historical logs as well as operational data, environmental data and image data of this flight.
[0073] The operational data specifically includes battery capacity, launch location, battery level sequence, location sequence, and attitude log.
[0074] Environmental data refers to meteorological indicators such as temperature, humidity, air density, and wind vectors collected by airborne sensors.
[0075] The image data consists of videos continuously collected by the airborne camera during this flight.
[0076] It provides a multimodal, temporal data foundation for all subsequent analyses, especially including real-time images for environmental perception, ensuring data integrity for return-to-home decisions in offline conditions.
[0077] The flight analysis module is used to parse the data records contained in the historical logs, combine the operational data and environmental data to calculate the weight index of each data record, and thus calculate the predicted power of various flight maneuvers.
[0078] The system analyzes historical logs, filters similar historical records by comparing the differences between current and historical environmental data, and calculates weight indices for these records.
[0079] By analyzing the power consumption and flight duration in historical records, a scheme based on a preset power range with random combinations and fitting of total power is used to reverse-calculate the historical reference power for each flight maneuver.
[0080] Finally, the weighted average method was used to calculate the predicted power for various flight maneuvers applicable to the current environment.
[0081] This enables adaptive and accurate prediction of energy consumption for different UAV maneuvers, allowing energy consumption estimates to dynamically adapt to specific meteorological environments and improving the reliability of subsequent route energy consumption assessments.
[0082] The safety management module is used to construct a three-dimensional mesh space, identify image data, and establish different return routes. Based on the flight maneuvers involved and their predicted power, the estimated total energy consumption of each return route is calculated, and the return routes are then selected.
[0083] First, a three-dimensional grid space is constructed using the drone's current position and takeoff position as diagonals and a preset safe passage distance as the basic side length.
[0084] Next, the image frames are time-aligned with the UAV position sequence, obstacles are identified using the YOLO target detection algorithm, and mapped onto a 3D mesh, marking each mesh as passable or impassable.
[0085] Based on this, starting from the current location and ending at the departure location, multiple return routes are planned to avoid impassable grids. Combined with the predicted power provided by the flight analysis module, the total energy consumption of each route is calculated.
[0086] Finally, based on the current remaining power and the energy threshold set by the safety factor, a feasible route with safe energy consumption is selected.
[0087] It enables three-dimensional digital modeling of the flight environment and dynamic obstacle perception, and completes preliminary path planning and energy safety screening, providing a set of physically feasible and energy-safe candidate solutions for the final route decision.
[0088] The return-to-home execution module analyzes the traffic conditions of the return route traversing the grid and calculates the efficiency index based on the estimated total energy consumption. Based on the efficiency index, it selects the optimal return route and executes the autonomous return-to-home mission.
[0089] For each feasible return route selected by the safety management module, further analysis is performed on all the grids it passes through, and the proportion of passable grids is statistically analyzed.
[0090] Simultaneously, combining the estimated total energy consumption and power threshold of the route, an efficiency index for each route is calculated using a comprehensive formula. Finally, the route with the highest efficiency index is selected as the optimal return route, and the drone is controlled to execute it.
[0091] By comprehensively weighing the energy efficiency and traffic safety of the return route using quantitative indicators, the optimal solution is intelligently selected from many feasible routes, ultimately enabling the drone to return safely and efficiently in an offline state.
[0092] Compared with the prior art, the beneficial effects achieved by the present invention are:
[0093] Adaptive Energy Consumption Prediction: This solution analyzes historical flight logs, compares current environmental meteorological data with historical records, filters out flight data under similar scenarios, and calculates the predicted power for different flight maneuvers accordingly. This overcomes the shortcomings of existing technologies that use fixed energy consumption models, enabling return-to-base energy consumption estimation to dynamically adapt to complex environmental factors such as actual wind conditions and temperature, significantly improving the accuracy and reliability of power consumption prediction.
[0094] Safe Path Generation: This solution utilizes real-time images from an onboard camera, identifies obstacles through target detection technology, and combines this with the UAV's position and attitude information to construct a 3D grid space that marks passable and impassable areas. Based on this, a return route is planned, ensuring the physical feasibility of the path in geometric space. This solves the fundamental problem of traditional straight-line return routes or simple waypoint planning ignoring three-dimensional obstacles, thus avoiding collision risks from the outset.
[0095] Multi-objective intelligent decision-making: After screening multiple energy-feasible return routes, this scheme does not select the shortest path, but instead makes a decision based on a comprehensive efficiency index. This index simultaneously considers the expected energy redundancy of the route and the accessibility of the areas traversed by the route, thus achieving an automatic trade-off between successful return and stable flight. This overcomes the limitations of existing single-objective optimization methods, achieving a balance between safety and economy, and selecting the globally optimal return route. Attached Figure Description
[0096] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:
[0097] Figure 1 This is a flowchart illustrating the UAV operation data management method based on flight safety according to the present invention.
[0098] Figure 2 This is a schematic diagram of the structure of the UAV operation data management system based on flight safety according to the present invention. Detailed Implementation
[0099] 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 skilled in the art without creative effort are within the scope of protection of the present invention.
[0100] Example 1: Please refer to Figure 1 This invention provides a method for managing unmanned aerial vehicle (UAV) operational data based on flight safety, including:
[0101] After the S100 drone goes offline, it collects historical logs, operational data, environmental data, and image data.
[0102] In practice, "offline" refers to the interruption of the signal connection between the drone and the ground control terminal. Historical logs contain data records for each flight, specifically operational and environmental data.
[0103] Operational data includes battery capacity, launch location, battery level sequence, location sequence, and attitude log.
[0104] The takeoff position refers to the spatial location of the drone at takeoff. The battery sequence consists of the remaining battery power at different times. The position sequence consists of the three-dimensional coordinates at different times.
[0105] Attitude logs are used to record the drone's flight attitude, flight maneuvers, and duration at different times.
[0106] Flight attitude refers to the angular states of pitch, roll, and yaw. Flight maneuvers include at least level flight, climb, descent, hovering, and turning.
[0107] Attitude logs discretize continuous flight attitudes into specific flight maneuvers (such as climb and turn), providing structured data for subsequent energy consumption analysis based on maneuvers.
[0108] Environmental data refers to various types of meteorological indicators around the drone collected at different times through various airborne sensors. The types of meteorological indicators include at least temperature, humidity, air density, and wind vector.
[0109] Image data refers to video images of the drone's surroundings captured by the onboard camera from the start of takeoff.
[0110] In the specific implementation process, the fusion of multimodal data of state, environment and vision is the prerequisite for subsequent intelligent environment modeling and accurate energy consumption prediction.
[0111] This established a complete and reliable offline data foundation for the entire autonomous return decision-making system.
[0112] This ensures that all necessary analytical data (historical experience, current status, and real-time environment) are available after the drone loses contact with the ground station, achieving data autonomy for decision-making.
[0113] S200 analyzes the data records contained in the historical logs, calculates the weight index of each data record by combining operational and environmental data, and calculates the reference power of various flight maneuvers. A weighted average method is then used to calculate the predicted power of various flight maneuvers. Specifically, this includes:
[0114] S201. Retrieve past duration from historical logs. All data records are analyzed, and the environmental data in each record is compared with the current environmental data to calculate the difference index for each record. Specifically:
[0115] Obtain the values of various meteorological indicators from the current environmental data. Preset weighting coefficients for various meteorological indicators Substitute the values into the formula to calculate the difference index for each data record. :
[0116] ;
[0117] In the formula, As a preset constant, The number of all meteorological indicator types. For the environmental data in the data record Values for various meteorological indicators.
[0118] In the specific implementation process, the difference index is calculated. (Using a weighted logarithmic formula to compare the historical and current environments) the similarity between historical experience and the current scene is quantified, thus achieving environmental perception.
[0119] S202. Mark data records where the difference index is less than the threshold, according to the formula: Calculate the weight index for each labeled data record. .in, The index representing the largest difference among all labeled data records.
[0120] Through weight index This approach assigns higher weight to historical data that is more similar to the current environment, reflecting the recency effect and the principle of environmental adaptation.
[0121] S203. Analyze the power sequence and attitude log in each marked data record, analyze the duration of each flight maneuver within each preset unit of power, and calculate the reference power for each flight maneuver in each marked data record. Specifically, this includes:
[0122] S2031. Obtain the marked data record. The power consumption sequence and attitude log are analyzed according to a preset unit. The time period corresponding to each unit of power consumed in the power consumption sequence is used as the sampling area, and the number of sampling areas is counted. .
[0123] S2032, Analyze the number of flight maneuver types included in the attitude log. Each flight maneuver and its duration are mapped to a sampling area, and a power range is preset for each flight maneuver.
[0124] S2033, Establishment There are several plans, each containing... The power is randomly selected from different power ranges, and the corresponding flight maneuver is also selected. The power of all flight maneuvers in different schemes is not exactly the same.
[0125] S2034, Obtaining the Solution The power of each flight maneuver was analyzed in the sampling area. The included flight maneuvers and their corresponding durations are substituted into the formula to calculate the solution. downsampling area Theoretical energy consumption :
[0126] ;
[0127] In the formula, The preset basic energy consumption, Sampling area The number of types of flight maneuvers included.
[0128] Sampling area The included first Types of flight maneuvers in the plan The corresponding power, For the first Types of flight maneuvers in the sampling area The duration corresponding to the time interval within the period.
[0129] S2035, Calculate the schemes separately The theoretical energy consumption of each sampling area is summed to form the scheme. The theoretical total energy consumption is calculated. Similarly, the theoretical total energy consumption of each scheme is calculated separately.
[0130] S2036. Analyze the electrical energy corresponding to each unit of battery capacity based on the battery capacity, and multiply by... The total electrical energy was then obtained. Calculate the theoretical total energy consumption and total electrical energy for each scheme. The absolute values between them are used to select the power values of each item in the scheme with the smallest absolute value as the reference power for the corresponding flight maneuver.
[0131] In practice, the preset unit is set in advance, typically using milliampere-hours (mAh), the basic unit of battery capacity, as a multiple. The multiple is determined based on the rate of power consumption during actual use.
[0132] By generating random schemes within a preset power range, and calculating the difference between the total theoretical energy consumption and the actual total electrical energy under each scheme, the most suitable scheme is selected to deduce the power of each action.
[0133] By using data fitting and optimization of the search process, the difficulty of building accurate physical models for complex actions is avoided, and power values are learned directly from historical data.
[0134] S204. Using the weighted average method, the predicted power of the flight maneuver is calculated based on the reference power of the flight maneuver in each marked data record and the corresponding weight index. The predicted power of each flight maneuver is calculated separately.
[0135] By using a data-driven approach, energy consumption prediction models can automatically adapt to specific environmental conditions (such as wind speed and air density) instead of using fixed theoretical values, greatly improving the reliability and personalization of predictions. This enables dynamic and accurate modeling of energy consumption for different drone flight maneuvers.
[0136] S300, construct a three-dimensional mesh space and identify the mesh passage status based on image data, establish A number of return routes are selected, and the estimated total energy consumption of each route is calculated based on the flight maneuvers involved and their predicted power, thereby enabling route selection. Specifically, this includes:
[0137] S301. A three-dimensional mesh space is constructed using the UAV's current position and takeoff position as diagonals. After temporally aligning the image data and position sequence, obstacles in the image are identified and detected, and passable and impassable meshes are obtained through analysis. Specifically, this includes:
[0138] S3011. Obtain the three-dimensional coordinates of the drone's current position and takeoff position, take the straight line between the two points as the diagonal, and set a preset safe passage distance. Using the basic side length, construct a three-dimensional boundary space that completely includes the diagonal.
[0139] S3012, A three-dimensional mesh discretization algorithm is used to discretize the three-dimensional boundary space along the three dimensions of longitude, latitude, and elevation. The step size is uniformly divided to generate a series of continuous cubic mesh units, thus forming a three-dimensional mesh space.
[0140] In the specific implementation process, a three-dimensional mesh space is constructed with safe passage distance as the granularity, achieving a balance between computational complexity and environmental resolution.
[0141] S3013. Extract image data and location sequences, and deframe the video images in the image data. Analyze the timestamp corresponding to each frame, and use a linear interpolation algorithm to resample the three-dimensional coordinates at the corresponding time in the location sequence.
[0142] S3014, For each frame image With each three-dimensional coordinate Perform temporal alignment to obtain a spatiotemporally aligned sequence. .for For each data set in the sequence, perform traffic status analysis:
[0143] First, Input a pre-trained model based on the YOLO object detection algorithm to identify and box out all obstacles in the image.
[0144] Then, with Using the viewpoint center, the three-dimensional spatial occupancy of each identified obstacle is calculated based on camera intrinsic parameters and flight attitude.
[0145] Finally, determine which cubic mesh cells in the 3D mesh space are occupied by obstacles at that viewpoint. Occupied cubic mesh cells are designated as impassable meshes, while unoccupied cubic mesh cells are designated as passable meshes.
[0146] By aligning image frames with GPS / IMU location data in a time sequence and using the YOLO target detection algorithm, obstacles in 2D images are mapped onto a 3D mesh in real time, marking them as passable / impassable, thus dynamically constructing a digital twin of the environment.
[0147] S302. In the three-dimensional grid space, starting from the current position of the UAV and ending at the departure position, establish a new location after avoiding impassable grids. The study analyzed the flight maneuvers required for each return route and their duration.
[0148] S303. The predicted energy consumption is obtained by multiplying the predicted power of each flight maneuver by the duration. The predicted total energy consumption is obtained by summing the predicted energy consumption of all flight maneuvers in a single return route. The predicted total energy consumption is calculated for each return route.
[0149] S304, Preset safety factor The remaining battery level is obtained based on the battery power sequence, and the corresponding remaining energy is analyzed in conjunction with the battery capacity before being multiplied by [the remaining energy]. Obtain the energy threshold. Select return routes whose total expected energy consumption is less than the energy threshold.
[0150] In practice, the safety factor is set between 0 and 1, and is used to measure the degree of redundancy left for the drone's battery power during actual operation. The larger the value, the smaller the redundancy, and vice versa.
[0151] In path planning, not only is geometric obstacle avoidance considered, but a safety factor is also introduced to set the power threshold and screen the route for energy consumption.
[0152] Ensuring that all candidate routes are conservatively feasible in terms of energy eliminates risky routes that cannot be completed for the final decision, the multi-objective optimization problem (safety, energy consumption) is transformed into a problem of finding the best solution in the feasible solution set.
[0153] The continuous physical flight space is discretized into a computable digital model, and a preliminary set of physically feasible and energy-safe return routes is planned. This achieves dual verification of environmental digitization and path feasibility.
[0154] S400: Analyze the traffic conditions of all grids traversed by each return route, and calculate the efficiency index based on the estimated total energy consumption. Select the optimal return route based on the efficiency index, and control the UAV to perform the autonomous return mission according to the optimal route. Specifically, this includes:
[0155] S401. Analyze the grids traversed by each return route and count the total number of grids traversed. and the number of passable grid cells within them. The efficiency index of each return route was calculated. The specific calculation formula is as follows:
[0156] ;
[0157] In the formula, and It is a constant. The electrical energy threshold, This represents the estimated total energy consumption.
[0158] This measure assesses energy economy; a higher value means more redundant power remaining after the flight, and thus greater safety.
[0159] This measure assesses the accessibility of a route; a higher value means that the route passes through a higher proportion of passable areas and the lower the risk of encountering unknown or dynamic obstacles.
[0160] constant and It is a trade-off parameter that allows for adjustments to the emphasis on energy consumption and safety based on actual needs.
[0161] By calculating and comparing all candidate return routes The system automatically selects the route with the highest overall score.
[0162] In practice, this quantitative decision-making mechanism enables drones to mimic comprehensive thinking in complex offline environments and execute energy-efficient and reliable return routes.
[0163] S402. Select the return route with the highest efficiency index as the optimal return route, and control the drone to perform the autonomous return mission according to the selected optimal return route.
[0164] Among multiple feasible return routes, an optimal decision is made based on a comprehensive quantitative indicator, driving the drone to execute the route. This achieves an intelligent trade-off between safety and economy, ultimately selecting the best route from the feasible ones.
[0165] Example 2: Please refer to Figure 2 The present invention also provides a UAV operation data management system based on flight safety, including a data acquisition module, a flight analysis module, a safety management module, and a return-to-home execution module.
[0166] The data acquisition module is used to collect historical logs, runtime data, environmental data, and image data.
[0167] In the specific implementation process, when the signal between the UAV and the ground control terminal is interrupted (offline), the historical logs, as well as the operational data, environmental data and image data of this flight, will be collected in a comprehensive manner.
[0168] The operational data specifically includes battery capacity, launch location, battery level sequence, location sequence, and attitude log.
[0169] Environmental data refers to meteorological indicators such as temperature, humidity, air density, and wind vectors collected by airborne sensors.
[0170] The image data consists of videos continuously collected by the airborne camera during this flight.
[0171] It provides a multimodal, temporal data foundation for all subsequent analyses, especially including real-time images for environmental perception, ensuring data integrity for return-to-home decisions in offline conditions.
[0172] The flight analysis module is used to parse the data records contained in the historical logs, combine the operational data and environmental data to calculate the weight index of each data record, and thus calculate the predicted power of various flight maneuvers.
[0173] In the specific implementation process, historical logs are analyzed, and similar historical records are filtered by comparing the differences between the current and historical environmental data (calculating the difference index), and weight indexes are calculated for these records.
[0174] By analyzing the power consumption and flight duration in historical records, a scheme based on a preset power range with random combinations and fitting of total power is used to reverse-calculate the historical reference power for each flight maneuver.
[0175] Finally, the weighted average method was used to calculate the predicted power for various flight maneuvers applicable to the current environment.
[0176] This enables adaptive and accurate prediction of energy consumption for different UAV maneuvers, allowing energy consumption estimates to dynamically adapt to specific meteorological environments and improving the reliability of subsequent route energy consumption assessments.
[0177] The safety management module is used to construct a three-dimensional mesh space, identify image data, and establish different return routes. Based on the flight maneuvers involved and their predicted power, the estimated total energy consumption of each return route is calculated, and the return routes are then selected.
[0178] In the specific implementation process, a three-dimensional grid space is first constructed using the current position and take-off position of the drone as the diagonal and the preset safe passage distance as the basic side length.
[0179] Next, the image frames are time-aligned with the UAV position sequence, obstacles are identified using the YOLO target detection algorithm, and mapped onto a 3D mesh, marking each mesh as passable or impassable.
[0180] Based on this, starting from the current location and ending at the departure location, multiple return routes are planned to avoid impassable grids. Combined with the predicted power provided by the flight analysis module, the total energy consumption of each route is calculated.
[0181] Finally, based on the current remaining power and the energy threshold set by the safety factor, a feasible route with safe energy consumption is selected.
[0182] It enables three-dimensional digital modeling of the flight environment and dynamic obstacle perception, and completes preliminary path planning and energy safety screening, providing a set of physically feasible and energy-safe candidate solutions for the final route decision.
[0183] The return-to-home execution module analyzes the traffic conditions of the return route traversing the grid and calculates the efficiency index based on the estimated total energy consumption. Based on the efficiency index, it selects the optimal return route and executes the autonomous return-to-home mission.
[0184] In the specific implementation process, for each feasible return route selected by the safety management module, we further analyze all the grids it passes through and count the proportion of passable grids.
[0185] Simultaneously, combining the estimated total energy consumption and power threshold of the route, an efficiency index for each route is calculated using a comprehensive formula. Finally, the route with the highest efficiency index is selected as the optimal return route, and the drone is controlled to execute it.
[0186] By comprehensively weighing the energy efficiency and traffic safety of the return route using quantitative indicators (efficiency index), the optimal solution is intelligently selected from many feasible routes, ultimately enabling the drone to return safely and efficiently in an offline state.
[0187] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0188] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for managing unmanned aerial vehicle (UAV) operational data based on flight safety, characterized in that: The method includes: S100: After the drone goes offline, it collects historical logs, operational data, environmental data, and image data. S200 analyzes the data records contained in the historical logs, calculates the weight index of each data record by combining operational data and environmental data, and the reference power of various flight maneuvers, and uses the weighted average method to calculate the predicted power of various flight maneuvers. S300, construct a three-dimensional mesh space and identify the mesh passage status based on image data, establish The return route is calculated based on the flight maneuvers involved and their predicted power, and the total energy consumption of each return route is then selected. S400: Analyze the traffic status of all grids traversed by each return route, and calculate the efficiency index based on the estimated total energy consumption; select the optimal return route based on the efficiency index, and control the UAV to perform the autonomous return mission according to the optimal return route.
2. The UAV operation data management method based on flight safety according to claim 1, characterized in that: In S100, offline means that the signal connection between the UAV and the ground control terminal is interrupted. The historical log contains data records for each flight, specifically including operational data and environmental data. Operational data includes battery capacity, launch location, battery level sequence, location sequence, and attitude log; The takeoff position refers to the spatial location of the drone when it takes off; the battery sequence consists of the remaining battery power at different times; the position sequence consists of the three-dimensional coordinates at different times. Attitude logs are used to record the UAV's flight attitude, flight maneuvers, and duration at different times. Flight attitude refers to the angles of pitch, roll, and yaw; flight maneuvers include at least level flight, climb, descent, hovering, and turning; Environmental data refers to various types of meteorological indicators around the drone collected at different times through various airborne sensors. The types of meteorological indicators include at least temperature, humidity, air density, and wind vector. Image data refers to video images of the drone's surroundings captured by the onboard camera from the start of takeoff.
3. The UAV operation data management method based on flight safety according to claim 2, characterized in that: S200 includes: S201. Retrieve past duration from historical logs. All data records are analyzed, and the environmental data in each record is compared with the current environmental data to calculate the difference index for each record. ; S202. Mark data records where the difference index is less than the threshold, according to the formula: Calculate the weight index for each labeled data record. ;in, The index representing the largest difference among all labeled data records; S203. Analyze the power sequence and attitude log in each marked data record, analyze the duration of each flight action within each preset unit of power, and calculate the reference power of each flight action in each marked data record. S204. Using the weighted average method, the predicted power of the flight maneuver is calculated based on the reference power of the flight maneuver in each marked data record and the corresponding weight index. The predicted power of each flight maneuver is calculated separately.
4. The UAV operation data management method based on flight safety according to claim 3, characterized in that: In S201, the values of various meteorological indicators in the current environmental data are obtained. Preset weighting coefficients for each meteorological indicator Substitute the values into the formula to calculate the difference index for each data record. : ； In the formula, As a preset constant, The number of all meteorological indicator types. For the environmental data in the data record Values for various meteorological indicators.
5. The UAV operation data management method based on flight safety according to claim 3, characterized in that: S203 includes: S2031. Obtain the marked data record. The power consumption sequence and attitude log are analyzed according to a preset unit. The time period corresponding to each unit of power consumed in the power consumption sequence is used as the sampling area, and the number of sampling areas is counted. ; S2032, Analyze the number of flight maneuver types included in the attitude log. Each flight maneuver and its duration are mapped to a sampling area, and a power range is preset for each flight maneuver. S2033, Establishment There are several options, each containing... The power is randomly selected from different power ranges, and the corresponding flight maneuver is not exactly the same for all flight maneuvers in different schemes; S2034, Obtaining the Solution The power of each flight maneuver was analyzed in the sampling area. The included flight maneuvers and their corresponding durations are substituted into the formula to calculate the solution. downsampling area Theoretical energy consumption : ； In the formula, The preset basic energy consumption, Sampling area The number of types of flight maneuvers included; Sampling area The included first Types of flight maneuvers in the plan The corresponding power, For the first Types of flight maneuvers in the sampling area The duration corresponding to the inner duration segment; S2035, Calculate the schemes separately The theoretical energy consumption of each sampling area is summed to form the scheme. The theoretical total energy consumption is calculated; and so on, the theoretical total energy consumption of each scheme is calculated separately. S2036. Analyze the electrical energy corresponding to each unit of battery capacity based on the battery capacity, and multiply by... The total electrical energy was then obtained. Calculate the theoretical total energy consumption and total electrical energy for each scheme. The absolute values between them are used to select the power values of each item in the scheme with the smallest absolute value as the reference power for the corresponding flight maneuver.
6. The UAV operation data management method based on flight safety according to claim 3, characterized in that: The S300 includes: S301. Construct a three-dimensional grid space with the current position and takeoff position of the UAV as the diagonal. After temporally aligning the image data and position sequence, identify and detect obstacles in the image, and analyze to obtain passable and impassable grids. S302. In the three-dimensional grid space, starting from the current position of the UAV and ending at the departure position, establish a new location after avoiding impassable grids. Analyze the return routes and all flight maneuvers required for each route, as well as their duration. S303. The estimated energy consumption is obtained by multiplying the predicted power of each flight maneuver by the duration. The estimated total energy consumption is obtained by summing the estimated energy consumption of all flight maneuvers in a single return route. The estimated total energy consumption of each return route is calculated separately. S304, Preset safety factor The remaining battery level is obtained based on the battery power sequence, and the corresponding remaining energy is analyzed in conjunction with the battery capacity before being multiplied by [the remaining energy]. Obtain the energy threshold; select return routes where the total expected energy consumption is less than the energy threshold.
7. The UAV operation data management method based on flight safety according to claim 6, characterized in that: S301 includes: S3011. Obtain the three-dimensional coordinates of the drone's current position and takeoff position, take the straight line between the two points as the diagonal, and set a preset safe passage distance. Using the basic side length, construct a three-dimensional boundary space that completely includes the diagonal; S3012, A three-dimensional mesh discretization algorithm is used to discretize the three-dimensional boundary space along the three dimensions of longitude, latitude, and elevation. The step size is uniformly divided to generate a series of continuous cubic mesh units, thus forming a three-dimensional mesh space; S3013. Extract image data and position sequence, and deframe the video images in the image data; analyze the timestamp corresponding to each frame image, and use a linear interpolation algorithm to resample the three-dimensional coordinates at the corresponding time in the position sequence; S3014, For each frame image With each three-dimensional coordinate Perform temporal alignment to obtain a spatiotemporally aligned sequence. ;for For each data set in the sequence, perform traffic status analysis: First, Input a pre-trained model based on the YOLO object detection algorithm to identify and box out all obstacles in the image; Then, with Using the viewpoint center, calculate the three-dimensional spatial occupancy of each identified obstacle based on camera intrinsic parameters and flight attitude; Finally, determine which cubic mesh cells in the 3D mesh space are occupied by obstacles at that field of view angle; the occupied cubic mesh cells are designated as impassable meshes, and the unoccupied cubic mesh cells are designated as passable meshes.
8. The UAV operation data management method based on flight safety according to claim 6, characterized in that: The S400 includes: S401. Analyze the grids traversed by each return route and count the total number of grids traversed. and the number of passable grid cells within them. The efficiency index of each return route is calculated. S402. Select the return route with the highest efficiency index as the optimal return route, and control the drone to perform the autonomous return mission according to the selected optimal return route.
9. The UAV operation data management method based on flight safety according to claim 8, characterized in that: Performance Index The calculation formula is as follows: ； In the formula, and It is a constant. The electrical energy threshold, This represents the estimated total energy consumption.
10. A flight safety-based UAV operation data management system, applied to the flight safety-based UAV operation data management method as described in claim 1, characterized in that: The system includes a data acquisition module, a flight analysis module, a safety management module, and a return-to-base execution module; The data acquisition module is used to collect historical logs, operational data, environmental data, and image data; The flight analysis module is used to parse the data records contained in the historical logs, combine the operational data and environmental data to calculate the weight index of each data record, and thus calculate the predicted power of various flight maneuvers. The safety management module is used to construct a three-dimensional mesh space, identify image data, and establish different return routes; it calculates the estimated total energy consumption of each return route based on the flight maneuvers involved and their predicted power, and then filters the return routes. The return-to-home execution module analyzes the traffic status of the return-to-home route as it passes through the grid, and calculates the efficiency index based on the estimated total energy consumption; it then selects the optimal return-to-home route based on the efficiency index and executes the autonomous return-to-home task.