Low-altitude unmanned aerial vehicle differential safety distance calculation method based on multi-parameter dynamic correction

Abstract

The invention discloses a low-altitude unmanned aerial vehicle differential safety distance calculation method based on multi-parameter dynamic correction, and aims to solve the problem that the safety distance of an existing unmanned aerial vehicle is interference or insufficient due to the characteristics of small size, weak interference resistance and no physiological delay, and cannot meet the requirements of high-precision automatic driving operation scenes such as electric power inspection, logistics distribution and the like. The medium-low altitude unmanned aerial vehicle differential safety distance calculation method comprises the steps that after a basic instruction and operation background data are determined, multi-source data are collected and processed in real time, and the collected multi-source data are input into an established multi-dimensional optimization model to obtain a corresponding weight coefficient and a corresponding correction coefficient; and calculating a safe distance according to the weight coefficient and the correction coefficient.

Description

Technical Field

[0001] Specifically, this invention relates to a multi-parameter dynamically corrected method for calculating the differentiated safe distance of low-altitude unmanned aerial vehicles (UAVs). Background Technology

[0002] With the rapid development of the low-altitude economy, drones have been widely used in aerial surveying, power line inspection, logistics delivery, urban security, and other fields. Flight safety has become a core bottleneck restricting the industry's large-scale development. Drone safety margin, as a key indicator for measuring safe distances from obstacles, other aircraft, and airspace boundaries during flight, directly determines the safety and reliability of drone flight through the scientific validity and accuracy of its calculation methods. Currently, drone safety distance calculations are often simplistic and fragmented, making it difficult to adapt to the complex and ever-changing flight environments and dynamic flight conditions. The fundamental reason is that early safety distance calculations primarily focused on the structural safety of the drone itself. Traditional drone structural safety factor formulas only consider structural and aerodynamic parameters such as gust loads, air density, and wing loading to assess the safety margin of the airframe's structural strength, completely ignoring the impact of environmental factors and dynamic flight parameters on the actual safe distance. This makes them unsuitable for flight safety assessments in complex outdoor scenarios. In the field of civilian drones, existing safety distance calculation formulas have been further simplified; for example, the safety distance formula used in Allumee drone performances... The current method for calculating safe distance by multiplying the drone's maximum speed and maximum tailwind speed by a fixed reaction time is simple, but it fails to consider the loss of effective speed due to the drone's tilted or turning flight attitude, and also neglects environmental variables such as visibility and obstacle distribution. This leads to significant deviations in calculations under low visibility and complex terrain conditions, increasing the risk of collisions. Furthermore, the proposed formula for safe flight in rainy weather only uses visibility, wind speed, and temperature coefficients to construct flight feasibility assessment indicators, which is a qualitative safety assessment method. It cannot quantify and calculate specific safe distance values, making it difficult to guide precise flight planning for drones.

[0003] In urban drone applications, while relevant standards incorporate collision zone radius and flight error parameters into their safety distance formulas, these formulas remain limited to static parameter superposition calculations. They fail to adapt to dynamic changes in drone airspeed and wind speed. Consequently, the calculated safety distances cannot meet actual flight safety requirements when drones are flying at high speeds or experiencing strong winds. Furthermore, existing technologies do not systematically couple the drone's dimensional characteristics with environmental obstacle compensation distances and visibility correction coefficients, resulting in insufficient comprehensiveness and accuracy in safety margin calculations. This becomes a key technological weakness hindering the safe operation of drones in complex environments such as densely populated urban areas and mountainous regions.

[0004] Currently, the technical implementation of safe distance control for low-altitude drones mainly focuses on two core solutions: the first is the traditional experience threshold mode, which relies on the operator's subjective judgment based on flight scenarios such as urban areas, suburbs, low altitude, and high altitude to preset a fixed safe distance threshold. For example, consumer drones are uniformly set with a 5m safety margin, while industrial drones are set with a 10m safety margin. During flight, the preset distance is maintained with obstacles through manual control or a simple obstacle avoidance switch. The second is the first-generation fixed parameter calculation model, which is based on simplified kinematic formulas and only considers two core parameters: the simple superposition of drone flight speed and wind speed and the basic reaction time. Through linear calculation logic, the initial quantification of safe distance is achieved. The commonality between these two approaches lies in their reliance on simplified assumptions and fixed parameters. Traditional experience-based threshold models are limited by individual operator experience and scenario judgment. In complex low-altitude environments such as gusts and low visibility, preset thresholds can easily lead to either over-threshold waste of airspace resources or under-threshold collision risks. Furthermore, they exhibit delayed response to sudden obstacles in remotely controlled scenarios. While the initial fixed-parameter calculation model broke through the dependence on pure manual experience and achieved preliminary quantitative calculations, it failed to consider multi-dimensional interference factors such as gust wind speed, crosswind angle, and electromagnetic interference intensity during low-altitude flight. In low-altitude applications, such as electric VTOL for manned aircraft, it also ignored the differences in perception-decision-execution link delays between UAVs and manned aircraft, failing to integrate key aspects such as sensor detection delays and image processing time. It only uses a uniform fixed reaction time parameter, resulting in calculation results that are disconnected from the actual safety requirements of UAV flight. Essentially, it still hasn't broken through the inherent framework of static simplified models. The main drawbacks are: First, the traditional model, which uses fixed values ​​to define safety distances, suffers from significant airspace resource waste. Currently, in order to cover the safety requirements of most scenarios, fixed safety distances are often set with redundancy values ​​based on the most extreme environments of strong winds and low visibility. Such configurations in extreme environments will generate a lot of unnecessary safety redundancy in normal flight environments.

[0005] Secondly, the simplistic formula for calculating safe distances suffers from fatal flaws due to its failure to adequately consider complex environmental interference, resulting in insufficient safe distance calculations and a significantly increased risk of collisions. Existing simple formulas often only consider a few parameters such as drone flight speed and basic reaction time, ignoring dynamic interference factors commonly present in low-altitude flight, such as gusts, crosswinds, electromagnetic interference, and low visibility. In real-world complex scenarios, these unconsidered interferences can cause the drone's actual flight trajectory to deviate from the expected path. The originally calculated safe distance becomes too small due to the lack of interference redundancy, greatly increasing the risk of collisions. For example, in a suburban inspection scenario with gusts of 5 m / s and mild electromagnetic interference, a simple formula will underestimate the drone's deceleration lag due to wind resistance and response delays caused by interference with flight control signals. The calculated safe distance will fail to meet actual obstacle avoidance requirements, potentially leading to collisions between the drone and buildings, trees, or other aircraft. This risk is further amplified, especially in high-density application scenarios such as drone swarm flights or low-altitude logistics delivery.

[0006] Secondly, existing technologies suffer from homogenized parameter considerations and inaccurate link delay calculations. On the one hand, existing technologies fail to distinguish the fundamental differences between UAVs and manned aircraft, directly applying safety distance model parameters for manned aircraft. This results in either excessive redundancy or insufficient safety distances, making it difficult to balance both. On the other hand, the lack of integration of critical link delays such as sensor detection latency and image processing time leads to a disconnect between the calculated results and the actual safety requirements of UAV flight, further exacerbating the deviation in safety distance calculations.

[0007] Finally, the existing technologies have significant shortcomings in engineering practicality. Traditional experience-based threshold models rely on the professional skills of operators, resulting in lag in response during remote control or autonomous flight scenarios and an inability to handle sudden obstacles. On the other hand, simplified formulas and fixed threshold models cannot be dynamically adapted to the type of drone, flight scenario, and intensity of environmental interference, leading to low calculation accuracy and poor robustness, which fails to meet the safety requirements of high-precision operation scenarios such as power line inspection, logistics distribution, and emergency rescue.

[0008] In summary, current calculations of safe distance for drones fail to integrate the drone's own attributes, dynamic flight status, environmental factors, and external risks. These issues collectively highlight the systemic deficiencies of existing technologies in terms of parameter completeness, scenario adaptability, calculation accuracy, and resource utilization efficiency. The current process for handling safe distances suffers from a lack of simplistic and refined formulas, making it difficult to accurately assess the flight safety of drones in complex scenarios. Summary of the Invention

[0009] This invention provides a multi-parameter dynamically corrected method for calculating differentiated safety distances of low-altitude unmanned aerial vehicles (UAVs) to solve the above-mentioned problems.

[0010] A multi-parameter dynamically corrected method for calculating the differentiated safety distance of low-altitude UAVs is proposed. The method involves determining the basic instructions and operational background data, performing real-time acquisition and processing of multi-source data, inputting the acquired multi-source data into an established multi-dimensional optimization model to obtain the corresponding weight coefficients and correction coefficients, and calculating the safety distance based on the weight coefficients and correction coefficients.

[0011] As a preferred option, the multi-dimensional optimization model is as follows: (1)

[0012] In the above formula, G represents the total safe distance; G is the geometric redundancy module. The module is based on the composite term of basic velocity and time. For dynamic interference correction module; A module for adapting to both environment and obstacles.

[0013] As a preferred option, the calculation process for the geometric redundancy module G is as follows: The minimum physical safety gap between the UAV and the target is determined by the geometric redundancy module G, and its value is the sum of the UAV's own safety radius and the target's safety radius. The calculation formula is as follows:

[0014] (2)

[0015] In the above formula, The target safety radius is defined when the target is another drone. This is the sum of the maximum fuselage size of the drone and the positioning error; when the target is a static obstacle, The equivalent safety radius of the obstacle; The safe radius of this drone is the sum of the maximum dimensions of the drone's fuselage and its own positioning error.

[0016] As a preferred option: Basic speed and time composite module The calculation process is as follows: Basic velocity and time composite term module It is a dynamic benchmark for safe distance, a composite module of basic speed and time. The distance the drone travels at its current airspeed within the time required for the obstacle avoidance system to complete the entire process of perception, decision-making, and execution is the basis for subsequent environmental interference correction. The calculation formula is as follows:

[0017] (3)

[0018] In the above formula, Airspeed, or the speed of a drone relative to the surrounding air, is the speed at which the drone flies. The value range is typically 4~10m / s for consumer-grade multirotor drones and 10~20m / s for industrial-grade drones;

[0019] The total response time of an obstacle avoidance system is the total time from when the UAV senses the target / interference to when the actuator completes the obstacle avoidance action. It includes perception delay, decision delay, and execution delay. This applies to visual obstacle avoidance systems. The value range is 0.5~2s; the lidar obstacle avoidance system The value range is 0.2~1s.

[0020] As a preferred option: Dynamic interference correction module The calculation process is as follows: Dynamic interference correction module Adjusting the base velocity and time composite term module using correction factors The value of is chosen to ensure that the safe distance is suitable for complex flight environments. The calculation formula is as follows:

[0021] (4)

[0022] In the above formula, This is a wind speed interference correction submodule. The electromagnetic interference correction factor has a value range of 1.0 to 1.5, and is calculated using the following formula:

[0023] (5)

[0024] In the above formula, This is a submodule for crosswind angle correction. This is the wind speed weighting coefficient, under normal operating conditions. The value ranges from 0.6 to 0.9 under strong wind conditions. The value ranges from 0.1 to 0.3, and the calculation formula is as follows:

[0025] (6)

[0026] In the above formula, The ambient wind speed is collected in real time by meteorological sensors mounted on the drone, with positive values ​​taken when the wind is downwind and negative values ​​taken when the wind is upwind; through The relative proportions of wind speed and airspeed are used to represent the interference of wind speed on the actual flight speed of drones in a dimensionless manner, adapting to the airspeed differences of different drone models. This is the crosswind weighting coefficient, with a value ranging from 0.2 to 0.5; The crosswind angle, which is the angle between the direction of the wind and the direction of the drone's flight, ranges from 0° to 180°. It is calculated by fusing the drone's attitude sensor and meteorological sensor data. The matching process establishes a pattern where crosswind interference is strongest and headwind / tailwind interference is weakest, thereby accurately adjusting the safe distance under crosswind conditions.

[0027] As a preferred option: Environment and obstacle matching module The calculation process is as follows:

[0028] (7)

[0029] (8)

[0030] (9)

[0031] In the above formula, This is a visibility correction submodule. The basic visibility safety redundancy ranges from 1 to 2 meters.

[0032] Visibility coefficient, The value ranges from 0 to 1. It is set to 0 when visibility is greater than 200m, and to [value missing] when visibility is between 50 and 200m. When visibility is below 50, the value is 0; Obstacle type adaptation submodule;

[0033] This is a correction value for obstacle type. The value ranges from 0.5 to 1m for static obstacles and from 1 to 2m for dynamic obstacles.

[0034] As a preferred approach: Before the final decision output and obstacle avoidance execution, the multi-dimensional optimization model undergoes three optimization processes, which are as follows:

[0035] Optimize geometric redundancy handling, and reduce the original Change to ,in For the radius of obstacles or other drones, This refers to the radius of the machine body;

[0036] Dynamic obstacle buffering, dynamic items This is the visibility margin coefficient. As a visibility correction factor, static As a dynamic item and Comprehensive item, static Suitable for low-visibility scenarios; Additional margin for unexpected obstacles, through Complete the process of distinguishing different types of obstacles, such as birds and tall buildings;

[0037] Quantizing multi-source latency processing to fix the response time Replace with end-to-end multi-source latency The process is broken down into four stages: sensing and detection, image processing, decision calculation, and execution response. This ensures the accuracy of calculations in the time dimension while yielding a complete optimized safety distance model. The final calculation formula for the multi-dimensional optimized model, obtained after these three optimization processes, is as follows:

[0038]

[0039] In the above formula, The target safety radius is defined when the target is another drone. This is the sum of the maximum fuselage size of the drone and the positioning error; when the target is a static obstacle, The equivalent safety radius of the obstacle; For multi-source delay; Based on the basic safety margin, the corresponding value range for consumer-grade products is 2~3m, and the value range for industry-grade products is 3~5m. This is the visibility margin coefficient; This is a visibility correction factor; For sudden obstacles, the additional margin is 1~3m for birds and 3~5m for tall buildings.

[0040] The end-to-end latency of drone obstacle avoidance is broken down into four stages. The quantification formula is:

[0041] (11)

[0042] In the above formula, the delay of each link is strongly related to the hardware performance of the UAV. When the corresponding LiDAR detection is used, the delay time is 0.01s, and when the corresponding visual sensor detection is used, the delay time is 0.1s. The image processing time ranges from 0.05 to 0.2s. The decision calculation time ranges from 0.05 to 0.1s. The execution response time ranges from 0.05 to 0.1s.

[0043] As a preferred option, the calculation formula for the modified model that incorporates environmental interference is:

[0044] (12)

[0045] In the above formula, This refers to the actual airspeed, including gusts and crosswinds. This refers to the gust coefficient; This is the crosswind linear correction factor; The fixed response time is 0.5s.

[0046] As a preferred option: the basic command is the pre-planned flight path, the operational background data is the scene parameters, and the multi-source data refers to the UAV's own parameters, environmental parameters, and obstacle parameters.

[0047] As a preferred approach: After calculating the safe distance, a consistency check is performed on the safe distance. Based on the successful check results, the final decision is output and obstacle avoidance is executed. The calculation process corresponding to the consistency check includes multi-level checks, and the multi-level check process is as follows:

[0048] The primary test is a model bias test, and the calculation process for the primary test is as follows:

[0049] Calculate the safety distance of the basic model The general formula for the safe collision distance of drones is:

[0050] (13)

[0051] Calculate the safe distance of the optimization model By combining formulas (10) and (12) and substituting them into the calculation of real-time multi-source data in the same scenario, the corresponding calculation deviation rate formula is:

[0052] (14)

[0053] After completing the above calculations, a judgment is made according to the standard. When the percentage is less than or equal to 5%, it indicates that the model calculations are consistent, and the test proceeds to the second stage; when... When the deviation exceeds 5%, multi-source data should be collected again to calibrate the wind speed and electromagnetic interference intensity sensors. The calculation should be repeated until the deviation meets the standard.

[0054] The secondary test is a real-time data reliability test, and the calculation process for the secondary test is as follows:

[0055] Validate the collected multi-source data:

[0056] The validity verification process for its own parameters is as follows: the airspeed fluctuation range is within ±0.5m / s, and the values ​​are compared with the hardware calibration values ​​for 5 consecutive sampling cycles to ensure that the total deviation is within ±0.1s.

[0057] The validity verification process for environmental parameters is as follows: the wind sampling variance is less than or equal to 3m. ² / s ² It is evident that the difference between three consecutive samples is less than or equal to 0.1;

[0058] The validity verification process for obstacle parameters is as follows: The recognition error is within ±0.1m, and the confidence level for obstacle type recognition is greater than or equal to 90%.

[0059] When all the above parameters meet the reliability requirements, the corresponding inspection standard is met; when any of the above parameters fails to meet the standard, the sensor redundancy switch is triggered and the above verification process is repeated until all parameters meet the reliability requirements.

[0060] Compared with existing technologies, this invention provides a multi-parameter dynamically corrected method for calculating differentiated safety distances of low-altitude UAVs, which has the following advantages:

[0061] The differentiated safety distance calculation method for low-altitude UAVs of this invention can be adapted to the process of accurately calculating and configuring safety distances in different scenarios, improving the flight safety assessment capability of UAVs in complex scenarios. In particular, it can accurately complete the calculation of safety distances under various flight attitudes of UAVs, providing an accurate data foundation for UAVs to be adapted to safe flight in various scenarios. The specific advantages of this invention are:

[0062] The nonlinear correction term for the crosswind angle proposed in the calculation process of this invention, in order to Replacing the traditional linear factor, it fundamentally solves the calculation overflow problem when the crosswind angle approaches 90°, and the corrected logic is more in line with the physical law of UAV crosswind deviation, significantly improving obstacle avoidance reliability in complex wind conditions.

[0063] The calculation method of this invention constructs a differentiated parameter system for consumer-grade and industrial-grade drones. It formulates hierarchical value rules based on core parameters such as fuselage size and basic safety margin, and combines redundant configurations for obstacle types and sudden risks to achieve full coverage of diverse low-altitude operation scenarios, greatly enhancing the model's generalization ability.

[0064] The computational model established during the calculation process of this invention is a multi-dimensional optimization model, which, after three optimization processes, can decompose the traditional fixed response time into end-to-end latency. It accurately matches the actual response process of the obstacle avoidance system, effectively eliminates the decision lag problem caused by the deviation of the delay assumption, and greatly improves the accuracy of safe distance calculation.

[0065] This invention adopts a lightweight computing architecture, with formulas containing only basic arithmetic operations and no complex iterations or matrix calculations. It can be directly embedded into the UAV flight control embedded module to achieve millisecond-level real-time computing, eliminating dependence on cloud computing power and ensuring the timeliness and reliability of obstacle avoidance decisions in complex environments. Attached Figure Description

[0066] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0067] 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.

[0068] Specific implementation method one: Combining Figure 1 This embodiment describes a method for calculating the differentiated safe distance of low-altitude UAVs. The method involves determining the basic instructions and operational background data, then performing real-time multi-source data acquisition and processing. The acquired multi-source data is input into an established multi-dimensional optimization model to obtain the corresponding weight coefficients and correction coefficients. The safe distance is then calculated based on the weight coefficients and correction coefficients.

[0069] In this implementation, the basic instruction is the pre-planned flight path, and the background data is the scenario parameters.

[0070] The specific implementation process of the differentiated safety distance calculation method for low-altitude UAVs in this embodiment can be as follows:

[0071] First, pre-planned flight paths and scenario parameters are imported to clarify the basic instructions and operational background of the mission. Then, real-time multi-source data acquisition is initiated, simultaneously acquiring the UAV's own parameters, environmental parameters, and obstacle parameters to provide multi-dimensional input for subsequent calculations. Next, in the coefficient calibration and model calculation phase, weight coefficients w1, w2, and w3, as well as correction coefficient k, are obtained based on environmental parameters. These are then substituted into the formula to calculate the safe distance S, and a consistency check is performed to ensure the rationality of the calculation results. Finally, decision output and obstacle avoidance execution are completed. Based on the aforementioned calculation results, flight control commands are generated to guide the UAV in autonomous obstacle avoidance and mission execution.

[0072] Specific Implementation Method Two: This implementation method is a further limitation of Specific Implementation Method One. The multi-dimensional optimization model in this implementation method is as follows:

[0073] (1)

[0074] In the above formula, This is the total safe distance. For geometric redundancy modules; The module is based on the composite term of basic velocity and time. For dynamic interference correction module; A module for adapting to both environment and obstacles.

[0075] This multi-dimensional optimization model compensates for the environmental adaptation defects of traditional static models. It also takes into account the attenuation effect of electromagnetic interference on UAVs, which essentially interferes with the three core links of UAV flight control communication link, sensor data acquisition, and power system control, resulting in a decrease in the actual flight performance of UAV speed, stability, and response accuracy. In addition, it solves the defects of unreasonable geometric redundancy terms and static obstacle buffering in traditional models.

[0076] Specific Implementation Method Three: This implementation method is a further limitation of Specific Implementation Method One or Two. In this implementation method, the calculation process of the geometric redundancy module G is as follows:

[0077] The minimum physical safety clearance between the UAV and the target is determined by the geometric redundancy module G, which is the sum of the UAV's own safety radius and the target's safety radius. The calculation formula is as follows:

[0078] (2)

[0079] In the above formula, The target safety radius is defined when the target is another drone. This is the sum of the maximum fuselage size of the drone and the positioning error; when the target is a static obstacle, The equivalent safety radius of the obstacle; The safe radius of this drone is the sum of the maximum dimensions of the drone's fuselage and its own positioning error.

[0080] In this embodiment, when the target is a multi-rotor UAV with a wheelbase of 0.5m, and the positioning error is ±0.1m, then the corresponding... .

[0081] Specific Implementation Method Four: This implementation method is a further limitation of Specific Implementation Methods One, Two, or Three. In this implementation method, the basic speed and time composite module... To provide a physical baseline guarantee for safe distance, ensure that there is at least a physical gap of length G between the drone and the target to avoid direct collision.

[0082] Specifically, the basic speed and time composite module It can also be expressed as a base speed-time term, which is the dynamic benchmark for the safe distance in this invention. It represents the distance the UAV flies at the current airspeed within the time required for the obstacle avoidance system to complete the entire process of "perception-decision-execution". It is the basic carrier for subsequent environmental interference correction.

[0083] The basic speed and time composite module in this embodiment The calculation process is as follows: Basic velocity and time composite term module It is a dynamic benchmark for safe distance, a composite module of basic speed and time. The distance the drone travels at its current airspeed within the time required for the obstacle avoidance system to complete the entire process of perception, decision-making, and execution is the basis for subsequent environmental interference correction. The calculation formula is as follows:

[0084] (3)

[0085] In the above formula, Airspeed, or the speed of a drone relative to the surrounding air, is the speed at which the drone flies. The value range is typically 4~10m / s for consumer-grade multi-rotor drones and 10~20m / s for industrial-grade drones. It is collected in real time by the drone flight control system and is specifically calculated through the fusion of airspeed tube, inertial measurement unit (IMU) and GPS.

[0086] in addition, The total response time of an obstacle avoidance system is the total time from when the UAV senses a target or interference to when the actuator completes the obstacle avoidance action. It includes perception delay, decision delay, and execution delay. This applies to visual obstacle avoidance systems. The value range is 0.5~2s; the lidar obstacle avoidance system The value range is 0.2~1s. Among them, the actuator is a motor, the perception delay is the image acquisition and recognition time of the vision sensor, the decision delay is the calculation time of the trajectory planning algorithm, and the execution delay is the motor speed adjustment time.

[0087] Specific Implementation Method Five: This implementation method is a further limitation of Specific Implementation Methods One, Two, Three, or Four. In this implementation method, the dynamic interference correction module... The calculation process is as follows:

[0088] The dynamic interference correction module is used to quantify the impact of three types of dynamic environmental factors—electromagnetic interference, wind speed, and crosswind angle—on the safe distance. By adjusting the value of the base speed-time term through correction factors, the safe distance is adapted to complex flight environments.

[0089] The calculation formula is:

[0090] (4)

[0091] In the above formula, This is a wind speed interference correction submodule. This is an electromagnetic interference correction factor, with a value ranging from 1.0 to 1.5. The stronger the electromagnetic interference, the lower the value. The larger the value, the better when there is no electromagnetic interference. =1.0, under the condition of high-voltage transmission lines with strong electromagnetic interference, =1.3, the calculation formula is:

[0092] (5)

[0093] In the above formula, This is a submodule for crosswind angle correction. This is the wind speed weighting coefficient, used to control the influence of wind speed on the effective velocity. Under normal operating conditions, The value ranges from 0.6 to 0.9 under strong wind conditions. The value ranges from 0.1 to 0.3, and the calculation formula is as follows:

[0094] (6)

[0095] In the above formula, The ambient wind speed is collected in real time by meteorological sensors mounted on the drone, with positive values ​​taken when the wind is downwind and negative values ​​taken when the wind is upwind; through The relative proportions of wind speed and airspeed are used to represent the interference of wind speed on the actual flight speed of drones in a dimensionless manner, adapting to the airspeed differences of different drone models. This is the crosswind weighting coefficient, with a value ranging from 0.2 to 0.5; The crosswind angle, which is the angle between the direction of the wind and the direction of the drone's flight, ranges from 0° to 180°. It is calculated by fusing the drone's attitude sensor and meteorological sensor data. The matching process establishes a pattern where crosswind interference is strongest and headwind / tailwind interference is weakest, thereby accurately adjusting the safe distance under crosswind conditions.

[0096] Specific Implementation Method Six: This implementation method is a further limitation of Specific Implementation Methods One, Two, Three, Four, or Five. In this implementation method, the environment and obstacle matching module... It transforms dynamic environmental disturbances into quantifiable correction factors, enabling the safety distance to adapt in real time to changes in electromagnetic interference, wind speed, and crosswind angle, thereby improving safety robustness in complex environments. This is achieved through an environment and obstacle adaptation module. This refers to the environment-barrier complex matching module. The calculation process is as follows:

[0097] (7)

[0098] (8)

[0099] (9)

[0100] In the above formula, This is a visibility correction submodule. The basic visibility safety redundancy ranges from 1 to 2 meters.

[0101] Visibility coefficient, The value ranges from 0 to 1. It is set to 0 when visibility is greater than 200m, and to [value missing] when visibility is between 50 and 200m. When visibility is below 50, the value is 0; Obstacle type adaptation submodule;

[0102] This is a correction value for the obstacle type. The value ranges from 0.5 to 1m for static obstacles and from 1 to 2m for dynamic obstacles such as pedestrians and vehicles.

[0103] Specific Implementation Method Seven: This implementation method is a further limitation of Specific Implementation Methods One, Two, Three, Four, Five, or Six. In this implementation method, the essence of the UAV's safe distance is to quantify the safe buffer space from the UAV's perception of an obstacle to completing an obstacle avoidance maneuver. The initial form of the basic safe distance model in this implementation method considers that when collision avoidance occurs, the safe distance should be calculated from the geometric center of the UAV. Therefore, it is necessary to calculate the radius of the current UAV and the UAV at risk of collision, and also consider the safety threshold under environmental influences, as well as leaving a safety margin. Therefore, the most basic safe distance only considers the geometric dimensions of the UAV and the obstacle, and a fixed buffer, and is defined as:

[0104] ;

[0105] in: To provide safety redundancy for the drone itself and avoid collisions with the fuselage; To establish a fixed safety margin, which is an empirical value, for example, 2~3m for consumer drones; For fixed obstacle buffers, only the presence or absence of obstacles is distinguished.

[0106] As can be seen from the above formula, the initial model simplifies the safety distance to the superposition of static geometric quantities, which is only applicable to low-speed scenarios with no wind and no interference. However, in actual flight, gusts and crosswinds can cause the drone's trajectory to deviate. The static model with fixed values ​​cannot adapt to complex scenarios.

[0107] Introducing an environmental disturbance correction model is necessary because, in actual flight, gusts can alter the airspeed of the drone, and crosswinds can cause lateral drift. Therefore, environmental disturbance terms need to be incorporated into the basic model, defining a safe distance with environmental correction. For example, if a drone traveling normally is affected by gusts, or tilts to turn or avoid obstacles, the effective component of its airspeed in the horizontal direction will decrease. This can be calculated mathematically as follows: Therefore, it needs to be divided by Only in this way can the airspeed in the tilted state be restored to the actual effective speed in the horizontal direction, ensuring that the calculation is of the horizontal motion effect, thereby correcting the "direction deviation of the speed".

[0108] Meanwhile, the actual speed of the drone is affected by wind speed, therefore the weighting coefficient for wind interference is set to... ,use The magnitude of wind interference is quantified by the ratio of wind speed to airspeed, thereby correcting for the wind's interference with airspeed. Therefore, the formula can be further optimized to obtain:

[0109]

[0110] In the above formula:

[0111] This refers to the actual airspeed, including gusts and crosswinds.

[0112] The response time is fixed, specifically an empirical value, which can be taken as 0.5s.

[0113] In the above formula, w1 is the gust coefficient. To correct for crosswind linearity, the above model compensates for the environmental adaptation deficiencies of the static model, but two problems still exist: one is the crosswind angle. exist When the angle is equal to 90°, it will approach 0, causing the entire fraction to approach infinity, thus causing the computer to overflow; secondly, the effect of electromagnetic interference on the speed reduction of the drone is not considered.

[0114] In addition, an optimization model that incorporates multiple interferences and delays is also included. Since the attenuation effect of electromagnetic interference on UAVs is essentially due to interference with the three core links of UAV flight control communication link, sensor data acquisition, and power system control, it leads to a decrease in the actual speed, stability, and response accuracy of the UAV's flight performance. Therefore, attenuation corrections are needed for airspeed and response efficiency in the safe distance model.

[0115] To address the issues of computational overflow and missing electromagnetic interference, two improvements were implemented: one was to linearly correct the crosswind. Replace with nonlinear terms ,use The boundedness of the values ​​completely avoids numerical overflow;

[0116] Another improvement is the introduction of an electromagnetic interference deceleration coefficient. The effect of electromagnetic interference on the attenuation of the actual airspeed of UAVs was quantified.

[0117] The optimized model is as follows:

[0118]

[0119] In the above formula, The electromagnetic interference reduction factor is 1.0 for no interference, 0.7 for slight interference, and 0.5~0.6 for severe interference.

[0120] The meanings of the remaining parameters are the same as those of the second formula in this embodiment.

[0121] The model at this stage has solved the problems of environmental disturbance quantification and computational stability, but it still has shortcomings such as unreasonable geometric redundancy terms and static barrier buffering. For example The geometric redundancy design is flawed, only considering collision avoidance between drones without taking into account differences in model and size, or fixed or moving obstacles such as birds encountered during flight. It fails to match the actual size differences between the drone and the obstacle, thus failing to accurately cover the geometric safety requirements of different obstacle types. Secondly, The static approach to obstacle buffering is not adapted to differences in obstacle type and environmental visibility. Firstly, there are differences in obstacle type: different obstacles have different motion characteristics and levels of danger. Specifically, static obstacles such as trees and tall buildings, which have no active displacement, require less buffering; while dynamic, sudden obstacles such as birds and low-flying aircraft, which have unpredictable trajectories, require additional reaction buffering. Secondly, there are differences in environmental visibility: low visibility environments such as fog or sandstorms reduce sensor detection accuracy, for example, increasing lidar ranging errors from centimeter-level to decimeter-level, requiring increased buffering to offset these errors; however, in clear weather, excessively large fixed buffers can affect flight maneuverability.

[0122] To further improve the model's engineering adaptability, three optimization processes are performed on the multi-dimensional optimization model before the final decision output and obstacle avoidance execution. These three optimization processes include optimizing geometric redundancy, dynamically buffering obstacles, and quantifying multi-source delays. The specific processes are as follows:

[0123] Optimize geometric redundancy handling, and reduce the original Change to ,in For the radius of obstacles or other drones, This refers to the radius of the machine body;

[0124] Dynamic obstacle buffering, dynamic items This is the visibility margin coefficient. As a visibility correction factor, static As a dynamic item and Comprehensive item, static Suitable for low-visibility scenarios; Additional margin for unexpected obstacles, through Complete the process of distinguishing different types of obstacles, such as birds and tall buildings;

[0125] Quantizing multi-source latency processing to fix the response time Replace with end-to-end multi-source latency The process is broken down into four stages: sensing and detection, image processing, decision calculation, and execution response. This ensures the accuracy of calculations in the time dimension while yielding a complete optimized safety distance model. The final calculation formula for the multi-dimensional optimized model, obtained after these three optimization processes, is as follows:

[0126]

[0127] In the above formula, The target safety radius is defined when the target is another drone. This is the sum of the maximum fuselage size of the drone and the positioning error; when the target is a static obstacle, The equivalent safety radius of the obstacle; For multi-source delay; Based on the basic safety margin, the corresponding value range for consumer-grade products is 2~3m, and the value range for industry-grade products is 3~5m. This is the visibility margin coefficient; This is a visibility correction factor; For sudden obstacles, the additional margin is 1~3m for birds and 3~5m for tall buildings.

[0128] The end-to-end latency of drone obstacle avoidance is broken down into four stages. The quantification formula is:

[0129] (11);

[0130] In the above formula, the delay of each link is strongly related to the hardware performance of the UAV. When the corresponding LiDAR detection is used, the delay time is 0.01s, and when the corresponding visual sensor detection is used, the delay time is 0.1s. The image processing time ranges from 0.05 to 0.2s. The decision calculation time ranges from 0.05 to 0.1s. The execution response time ranges from 0.05 to 0.1s.

[0131] Specific Implementation Method Eight: This implementation method is a further limitation of Specific Implementation Method Seven. The calculation formula for the correction model that introduces environmental interference in this implementation method is as follows:

[0132] (12);

[0133] In the above formula, This refers to the actual airspeed, including gusts and crosswinds. This refers to the gust coefficient; This is the crosswind linear correction factor; The fixed response time is 0.5s.

[0134] The above process ensures that the final model gradually iterates from the static superposition of geometric quantities to a dynamic quantization model that integrates geometric adaptation, multiple environmental interferences, and end-to-end latency, which not only adapts to complex flight scenarios but also ensures computational stability and accuracy in engineering applications.

[0135] Specific Implementation Method Nine: This implementation method is a further limitation of Specific Implementation Method One, Two, Three, Four, Five, Six, Seven or Eight. The embedded module in this implementation method is a local computing hardware mounted on the UAV flight control system, specifically an STM32, which can run lightweight algorithms locally without relying on cloud computing power, ensuring the real-time performance of obstacle avoidance decisions.

[0136] The flight control system in this embodiment is specifically the core control unit of the UAV, which is responsible for receiving sensor data, running the safe distance calculation algorithm, and outputting obstacle avoidance control commands. It is the hardware carrier for the implementation of the technical solution.

[0137] The sensor fusion in this embodiment integrates and optimizes the detection data from multiple devices such as lidar, visual sensors, and anemometers to improve the accuracy of obstacle recognition and environmental parameter acquisition, and to provide reliable data input for safe distance calculation.

[0138] Specific Implementation Method Ten: This implementation method is a further limitation of Specific Implementation Methods One, Two, Three, Four, Five, Six, Seven, Eight, or Nine. Multi-source data refers to the UAV's own parameters, environmental parameters, and obstacle parameters.

[0139] In this embodiment, the UAV collects the following parameters in real time using devices such as lidar, visual sensors, and anemometers.

[0140] Intrinsic parameters: For the fuselage radius, Standard flight speed;

[0141] Environmental parameters: For the radius of obstacles / other drones, θ represents gust speed, θ represents crosswind angle, electromagnetic interference intensity, and visibility level;

[0142] Delay parameters: , etc. are used for calculation ;

[0143] Burst parameters; Additional margin for the type of sudden obstacle.

[0144] During the coefficient calibration and model calculation phase, correction coefficients are automatically matched based on environmental parameters: w1 is the gust coefficient. is the electromagnetic interference coefficient, w3 is the visibility coefficient, and k is the crosswind nonlinearity coefficient;

[0145] Finally, calculate the real-time total T value based on hardware performance; substitute it into the optimization model formula to calculate the safe distance. ; Perform consistency verification: Compare the initial formula 1 of the basic model with the results of the optimized model. If the deviation is >5%, re-collect parameters and recalibrate.

[0146] Specific Implementation Method Eleven: This implementation method is a further limitation of Specific Implementation Methods One, Two, Three, Four, Five, Six, Seven, Eight, Nine, or Ten. In this implementation method, after the calculated safety distance is obtained, the safety distance is subjected to a consistency check. The specific check process includes multi-level checks, and the multi-level check flow is as follows:

[0147] The primary test is a model bias test, and the calculation process for the primary test is as follows:

[0148] Calculate the safety distance of the basic model The general formula for the safe collision distance of drones is:

[0149] (13);

[0150] Calculate the safe distance of the optimization model By combining formulas (10) and (12) and substituting them into the calculation of real-time multi-source data in the same scenario, the corresponding calculation deviation rate formula is:

[0151] (14);

[0152] After completing the above calculations, a judgment is made according to the standard. When the percentage is less than or equal to 5%, it indicates that the model calculations are consistent, and the test proceeds to the second stage; when... When the deviation exceeds 5%, multi-source data should be collected again to calibrate the wind speed and electromagnetic interference intensity sensors. The calculation should be repeated until the deviation meets the standard.

[0153] The secondary test is a real-time data reliability test, and the calculation process for the secondary test is as follows:

[0154] Validate the collected multi-source data:

[0155] The validity verification process for its own parameters is as follows: the airspeed fluctuation range is within ±0.5m / s, and the values ​​are compared with the hardware calibration values ​​for 5 consecutive sampling cycles to ensure that the total deviation is within ±0.1s.

[0156] The validity verification process for environmental parameters is as follows: the wind speed sampling variance is less than or equal to 3 (m / s). ²It is evident that the difference between three consecutive samples is less than or equal to 0.1;

[0157] The validity verification process for obstacle parameters is as follows: R0 identification error within ±0.1m, and obstacle type identification confidence level greater than or equal to 90%;

[0158] When all the above parameters meet the reliability requirements, the corresponding inspection standard is met; when any of the above parameters fails to meet the standard, the sensor redundancy switch is triggered and the above verification process is repeated until all parameters meet the reliability requirements.

[0159] Specific Implementation Method Twelve: This implementation method is a further limitation of Specific Implementation Methods One, Two, Three, Four, Five, Six, Seven, Eight, Nine, Ten, or Eleven. It describes the process of final decision output and obstacle avoidance execution based on the qualified inspection results, i.e., the calculated... The data is sent to the flight control system to compare the real-time relative distance between the drone and obstacles or other drones, and is categorized into two cases:

[0160] When the real-time distance is greater than or equal to At that time, the current flight status should be maintained.

[0161] When the real-time distance is less than When this occurs, obstacle avoidance actions are triggered to ensure the vehicle is in a hovering, circling, or rising / falling motion. The specific process is as follows:

[0162] Step 1: Calculate real-time performance based on hardware capabilities , To calculate the end-to-end latency of the obstacle avoidance system, it is necessary to break it down into the sum of the latency of four stages: sensing and detection, image processing, decision calculation, and execution response, based on the hardware performance of the UAV. Substituting these values ​​into formula (12) yields the result. .

[0163] Step 2: Substitute the values ​​into the optimization model formula to calculate the safe distance. The multi-source data collected in real time and the calculations in step one will be used to... Substituting into the optimization model formula 10 in the seventh specific implementation method, the safe distance is calculated. .

[0164] Step 3: Perform consistency verification, which involves comparing the results of the basic model and the optimized model. Using the basic model formulas (13) and (14) as the benchmark, compare their results with... The deviation, the specific calculation process and procedure are as follows:

[0165] First-level test: Model bias test, the specific calculation process is as follows:

[0166] Calculate the safety distance of the basic model The general formula for the safe collision distance of drones is:

[0167] (13)

[0168] Calculate the safe distance of the optimization model : Using formula (10) in claim 7, combined with formula (12) in claim 7, and substituting real-time multi-source data of the same scenario;

[0169] The formula for calculating the deviation rate is:

[0170] (14)

[0171] Judgment criteria: If If the value is less than or equal to 5%, the model calculation is consistent, and the process proceeds to the second-level test; if... If the deviation exceeds 5%, re-collect multi-source data, prioritizing the collection of data from calibrated wind speed and electromagnetic interference intensity sensors, and repeat the calculation until the deviation meets the standard.

[0172] Secondary verification: Real-time data reliability verification, which involves validating the collected multi-source data.

[0173] Self-parameters: Airspeed fluctuation range is within ±0.5m / s, specifically after 5 consecutive sampling cycles, compared with the hardware calibration value, the total deviation is within ±0.1s;

[0174] Environmental parameters: Wind speed sampling variance less than or equal to 3 (m / s) ² It is evident that the difference between three consecutive samples is less than or equal to 0.1;

[0175] Obstacle parameters: The R0 identification error in the lidar ranging accuracy is within ±0.1m, and the obstacle type identification confidence level is greater than or equal to 90%;

[0176] Judgment criteria: If all parameters meet the reliability requirements, the inspection criteria are met. If any parameter fails to meet the criteria, sensor redundancy switching is triggered. Specifically, this can be achieved by switching to LiDAR when the visual sensor fails.

[0177] Step 4: After completing the above process, finally execute the decision output and obstacle avoidance, and verify that the target has been met. The data is sent to the flight control system, which compares the real-time relative distance L between the drone and the obstacle and executes the corresponding action.

[0178] When L is greater than or equal to At the same time: maintain the current flight status, that is, the speed, heading and altitude are consistent with the pre-planned flight path, with errors of less than or equal to ±0.5m / s, ±2° and ±0.3m respectively.

[0179] When L is less than At the same time: Obstacle avoidance is triggered in the order of detour → hover → ascent / descend.

[0180] Detour: Using the obstacle as the center, Horizontal orbit with a radius of +1.0m, angular velocity less than or equal to 10° / s;

[0181] Insufficient detour space: Maintain current altitude and hover with an accuracy of less than or equal to ±0.2m, and simultaneously replan the flight path;

[0182] Hovering risk: Vertical ascent and descent at a speed of less than or equal to 2 m / s, resulting in a vertical distance greater than or equal to... The horizontal flight path was then replanned.

[0183] This invention can be adapted to different types of drones. To improve the model's generalization ability, this invention designs a differentiated parameter reference table, as shown in the table below:

[0184] Table 1

[0185] Table of safe distance calculations for different scenarios:

[0186] Table 2

[0187] As illustrated in Tables 1 and 2, this invention can be adapted to the process of accurately calculating and configuring safe distances in different scenarios, thereby improving the flight safety self-assessment performance of UAVs in complex scenarios, especially in accurately acquiring safe distances under various flight attitudes of UAVs.

Claims

1. A method for calculating differentiated safe distances for low-altitude unmanned aerial vehicles (UAVs) with multi-parameter dynamic correction, characterized in that: The method for calculating the differentiated safe distance of low-altitude UAVs involves determining the basic instructions and operational background data, then collecting and processing multi-source data in real time. The collected multi-source data is then input into an established multi-dimensional optimization model to obtain the corresponding weight coefficients and correction coefficients. The safe distance is then calculated based on the weight coefficients and correction coefficients.

2. The method for calculating differentiated safety distances of low-altitude UAVs with multi-parameter dynamic correction according to claim 1, characterized in that: The multi-dimensional optimization model is as follows: （1） In the above formula, G represents the total safe distance; G is the geometric redundancy module. The module for the basic velocity and time composite term; For dynamic interference correction module; A module for adapting to both environment and obstacles.

3. The method for calculating the differentiated safety distance of low-altitude UAVs with multi-parameter dynamic correction according to claim 2, characterized in that: The calculation process for the geometric redundancy module G is as follows: The minimum physical safety clearance between the UAV and the target is determined by the geometric redundancy module G, which is the sum of the UAV's own safety radius and the target's safety radius. The calculation formula is as follows: （2） In the above formula, The target safety radius is defined when the target is another drone. This is the sum of the maximum fuselage size and the positioning error of the drone; When the target is a static obstacle The equivalent safety radius of the obstacle; The safe radius of this drone is the sum of the maximum dimensions of the drone's fuselage and its own positioning error.

4. The method for calculating the differentiated safety distance of low-altitude UAVs with multi-parameter dynamic correction according to claim 3, characterized in that: Basic speed and time composite module The calculation process is as follows: Basic speed and time composite module It is a dynamic benchmark for safe distance, a composite module of basic speed and time. The distance the drone travels at its current airspeed within the time required for the obstacle avoidance system to complete the entire process of perception, decision-making, and execution is the basis for subsequent environmental interference correction. The calculation formula is as follows: （3） In the above formula, Airspeed, or the speed of a drone relative to the surrounding air, is the speed at which the drone flies. The value range is typically 4~10m / s for consumer-grade multirotor drones and 10~20m / s for industrial-grade drones; The total response time of an obstacle avoidance system is the total time from when the UAV senses the target / interference to when the actuator completes the obstacle avoidance action. It includes perception delay, decision delay, and execution delay. This applies to visual obstacle avoidance systems. The value range is 0.5~2s; the lidar obstacle avoidance system The value range is 0.2~1s.

5. The method for calculating the differentiated safety distance of a low-altitude UAV with multi-parameter dynamic correction according to claim 4, characterized in that: Dynamic interference correction module The calculation process is as follows: Dynamic interference correction module Adjusting the base velocity and time composite term module using correction factors The value of is chosen to ensure that the safe distance is suitable for complex flight environments. The calculation formula is as follows: （4） In the above formula, This is a wind speed interference correction submodule. The electromagnetic interference correction factor has a value range of 1.0 to 1.5, and is calculated using the following formula: （5） In the above formula, This is a submodule for crosswind angle correction. This is the wind speed weighting coefficient, under normal operating conditions. The value ranges from 0.6 to 0.9 under strong wind conditions. The value ranges from 0.1 to 0.3, and the calculation formula is as follows: （6） In the above formula, The ambient wind speed is collected in real time by meteorological sensors mounted on the drone, with positive values ​​taken when the wind is downwind and negative values ​​taken when the wind is upwind; through The relative proportions of wind speed and airspeed are used to represent the interference of wind speed on the actual flight speed of drones in a dimensionless manner, adapting to the airspeed differences of different drone models. This is the crosswind weighting coefficient, with a value ranging from 0.2 to 0.5; The crosswind angle, which is the angle between the direction of the wind and the direction of the drone's flight, ranges from 0° to 180°. It is calculated by fusing the drone's attitude sensor and meteorological sensor data. The matching process establishes a pattern where crosswind interference is strongest and headwind / tailwind interference is weakest, thereby accurately adjusting the safe distance under crosswind conditions.

6. The method for calculating the differentiated safety distance of a low-altitude UAV with multi-parameter dynamic correction according to claim 5, characterized in that: Environment and obstacle matching module The calculation process is as follows: （7） （8） （9） In the above formula, This is a visibility correction submodule. The basic visibility safety redundancy ranges from 1 to 2 meters. Visibility coefficient, The value ranges from 0 to 1. It is set to 0 when visibility is greater than 200m, and to [value missing] when visibility is between 50 and 200m. When visibility is below 50, the value is 0; Obstacle type adaptation submodule; This is a correction value for obstacle type. The value ranges from 0.5 to 1m for static obstacles and from 1 to 2m for dynamic obstacles.

7. A method for calculating differentiated safety distances for low-altitude unmanned aerial vehicles (UAVs) with multi-parameter dynamic correction according to any one of claims 1 to 6, characterized in that: Before the final decision output and obstacle avoidance execution, the multi-dimensional optimization model undergoes three optimization processes, which are as follows: Optimize geometric redundancy handling, and reduce the original Change to ,in For the radius of obstacles or other drones, This refers to the radius of the machine body; Dynamic obstacle buffering, dynamic items This is the visibility margin coefficient. As a visibility correction factor, static As a dynamic item and Comprehensive item, static Suitable for low-visibility scenarios; Additional margin for unexpected obstacles, through Complete the process of distinguishing different types of obstacles, such as birds and tall buildings; Quantizing multi-source latency processing to fix the response time Replace with end-to-end multi-source latency The process is broken down into four stages: sensing and detection, image processing, decision calculation, and execution response. This ensures the accuracy of calculations in the time dimension while yielding a complete optimized safety distance model. The final calculation formula for the multi-dimensional optimized model, obtained after these three optimization processes, is as follows: （10） In the above formula, The target safety radius is defined when the target is another drone. This is the sum of the maximum fuselage size and the positioning error of the drone; When the target is a static obstacle The equivalent safety radius of the obstacle; For multi-source delay; Based on the basic safety margin, the corresponding value range for consumer-grade products is 2~3m, and the value range for industry-grade products is 3~5m. This is the visibility margin coefficient; This is a visibility correction factor; For sudden obstacles, the additional margin is 1~3m for birds and 3~5m for tall buildings. The end-to-end latency of drone obstacle avoidance is broken down into four stages. The quantification formula is: （11） In the above formula, the delay of each link is strongly related to the hardware performance of the UAV. When the corresponding LiDAR detection is used, the delay time is 0.01s, and when the corresponding visual sensor detection is used, the delay time is 0.1s. The image processing time ranges from 0.05 to 0.2s. The decision calculation time ranges from 0.05 to 0.1s. The execution response time ranges from 0.05 to 0.1s.

8. The method for calculating the differentiated safety distance of a low-altitude UAV with multi-parameter dynamic correction according to claim 7, characterized in that: The calculation formula for the modified model that incorporates environmental disturbances is as follows: （12） In the above formula, This refers to the actual airspeed, including gusts and crosswinds. This refers to the gust coefficient; This is the crosswind linear correction factor; The fixed response time is 0.5s.

9. The method for calculating differentiated safety distances of low-altitude UAVs with multi-parameter dynamic correction according to claim 1, characterized in that: The basic command is the pre-planned flight path, and the background data for the operation is the scene parameters; multi-source data refers to the parameters of the UAV itself, environmental parameters, and obstacle parameters.

10. The method for calculating the differentiated safety distance of a low-altitude UAV with multi-parameter dynamic correction according to claim 1, characterized in that: After calculating the safe distance, a consistency check is performed on the safe distance. Based on the results of the successful check, the final decision is output and obstacle avoidance is executed. The calculation process corresponding to the consistency check includes multi-level checks, and the multi-level check process is as follows: The primary test is a model bias test, and the calculation process for the primary test is as follows: Calculate the safety distance of the basic model The general formula for the safe collision distance of drones is: （13） Calculate the safe distance of the optimization model By combining formulas (10) and (12) and substituting them into the calculation of real-time multi-source data in the same scenario, the corresponding calculation deviation rate formula is: （14） After completing the above calculations, a judgment is made according to the standard. When the percentage is less than or equal to 5%, it indicates that the model calculations are consistent, and the test proceeds to the second stage; when... When the deviation exceeds 5%, multi-source data should be collected again to calibrate the wind speed and electromagnetic interference intensity sensors. The calculation should be repeated until the deviation meets the standard. The secondary test is a real-time data reliability test, and the calculation process for the secondary test is as follows: Validate the collected multi-source data: The validity verification process for its own parameters is as follows: the airspeed fluctuation range is within ±0.5m / s, and the values ​​are compared with the hardware calibration values ​​for 5 consecutive sampling cycles to ensure that the total deviation is within ±0.1s. The validity verification process for environmental parameters is as follows: the wind speed sampling variance is less than or equal to 3 (m / s). ² It is evident that the difference between three consecutive samples is less than or equal to 0.1; The validity verification process for obstacle parameters is as follows: The recognition error is within ±0.1m, and the confidence level for obstacle type recognition is greater than or equal to 90%. When all the above parameters meet the reliability requirements, the corresponding inspection standard is met; when any of the above parameters fails to meet the standard, the sensor redundancy switch is triggered and the above verification process is repeated until all parameters meet the reliability requirements.

Images