Urgent danger avoiding method based on unmanned aerial vehicle investigation

Abstract

The invention belongs to the technical field of automatic control, and particularly relates to an urgent danger avoiding method based on unmanned aerial vehicle investigation, and the method comprises the steps: 1, taking a target unmanned aerial vehicle as a sphere center, taking a set value as a radius, and defining a spherical unmanned aerial vehicle danger avoiding space range; constructing an environmental threat vector; 2, screening out corresponding positions with collision probabilities greater than or equal to a set collision threshold value through the environmental threat vector in combination with the position vector and flight parameters of the target unmanned aerial vehicle, and reserving corresponding positions with collision probabilities lower than the set collision threshold value as candidate positions; and step 3, according to the position vector and the flight parameters of the target unmanned aerial vehicle, calculating the flight distance of the target unmanned aerial vehicle passing through each candidate position, and calculating an optimal control instruction, so that the target unmanned aerial vehicle passes through the candidate positions. The method can ensure that the unmanned aerial vehicle always selects the safest and efficient flight path in a complex environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of automatic control, and particularly relates to an emergency avoidance method based on UAV reconnaissance. Background Art

[0002] With the rapid development of UAV technology, UAVs have been widely used in many fields such as military reconnaissance, logistics distribution, urban patrol, environmental monitoring, and agricultural operations. However, since UAVs usually need to fly at low altitudes in complex environments, especially in urban built-up areas, forest areas, or dynamic environments with multiple obstacles, the collision risk increases significantly. To ensure the flight safety of UAVs and improve the mission completion rate, autonomous obstacle avoidance technology has become a key research direction in UAV systems. The existing autonomous obstacle avoidance methods mainly rely on technical means such as multi-sensor fusion, path planning algorithms, and environmental threat assessment. However, in high-dynamic environments, the existing technologies still have significant limitations and technical bottlenecks.

[0003] LiDAR has become an important tool for UAV autonomous obstacle avoidance due to its high-precision distance measurement ability and strong environmental adaptability. By collecting the spatial information of the surrounding environment in real time through LiDAR, the UAV can construct a high-precision three-dimensional map, thereby accurately detecting the position and shape of obstacles. In the existing technology, LiDAR obstacle avoidance algorithms usually use path planning methods based on grid maps or topological maps to find the shortest and collision-free flight path for the UAV. However, this method depends on the field of view range and scanning frequency of LiDAR. For high-speed moving dynamic obstacles, the radar may not be able to capture the latest state of the obstacle source in time, resulting in a lag in obstacle avoidance response. In addition, the high cost and large volume of LiDAR devices limit their application on small UAVs. Vision sensors (such as monocular, binocular, or depth cameras) have the advantages of small size, low cost, and rich environmental information acquisition, and are another commonly used obstacle avoidance means. Through vision processing algorithms, the UAV can identify the type, position, and movement trajectory of the obstacles in front, and achieve the obstacle avoidance function. However, the vision obstacle avoidance system has high requirements for lighting conditions and computing resources, and its performance deteriorates in strong light, shadow, or night environments. In addition, the existing vision obstacle avoidance algorithms usually adopt two-dimensional image processing technology, which is difficult to accurately predict the time dynamics of multiple obstacle sources moving at high speed in three-dimensional space, resulting in inflexible obstacle avoidance strategies. Summary of the Invention

[0004] In view of this, the main object of the present invention is to provide an emergency avoidance method based on UAV reconnaissance. By constructing a dynamic avoidance space, introducing an environmental threat vector, a time-dimensional risk prediction model, and a multi-objective comprehensive optimization strategy, the dynamic threat of obstacle sources is evaluated in real time and the optimal avoidance path is planned. Compared with the prior art, the present invention can significantly improve the autonomous avoidance ability of UAVs in the presence of multiple obstacle sources and high-dynamic environments, achieve a dynamic balance between flight distance and collision probability, ensure that the UAV always selects the safest and most efficient flight path in complex environments, and has higher real-time performance, robustness, and scalability.

[0005] The technical solution adopted by the present invention is as follows:

[0006] An emergency avoidance method based on UAV reconnaissance, the method comprising:

[0007] Step 1: Taking the target UAV as the center of the sphere and a set value as the radius, define the spherical UAV avoidance space range; within the UAV avoidance space range, construct an environmental threat vector through the position vectors and motion parameters of each obstacle source.

[0008] Step 2: Through the environmental threat vector, combined with the position vector and flight parameters of the target UAV, calculate the collision probability of each position within the spherical UAV avoidance space range; the collision probability describes the probability of the target UAV colliding with an obstacle source when passing through this position; screen out the positions where the collision probability is greater than or equal to the set collision threshold, and retain the positions where the collision probability is lower than the set collision threshold as candidate positions.

[0009] Step 3: Through the position vector and flight parameters of the target UAV, calculate the flight distance of the target UAV passing through each candidate position, and calculate the optimal control command so that when the target UAV passes through the candidate position, the flight distance and collision probability meet the set constraint conditions.

[0010] Further, the motion parameters of the obstacle source include: obstacle source speed, obstacle source pointing angle, and obstacle source mass; the flight parameters of the target UAV include: target UAV mass, target UAV speed, and target UAV pointing angle; the environmental threat vector is defined by the following formula: Let be the position vector of the target UAV; let R>0 be the set radius; the UAV avoidance space range is defined as represents the real number space; x is the position point in the UAV avoidance space range; ∥·∥ is the norm operation.

[0011] Further, let the position vector of the i-th obstacle source be Let be the obstacle source speed of the i-th obstacle source, and its modulus is denoted as vi ; Let θ i ∈[0, 2π] be the obstacle source pointing angle of the i-th obstacle source; Let m i > 0 be the obstacle source mass of the i-th obstacle source; Let be the velocity vector of the target UAV; Let r i = p t - p i be the relative position between the target UAV and the i-th obstacle source; d i = ∥r i ∥ be the relative distance between the target UAV and the i-th obstacle source; In the risk avoidance space S, calculate the threat coefficient Φ i of the i-th obstacle source:

[0012]

[0013] where r i,y is the Y-axis component of r i ; r i,x is the X-axis component of r i .

[0014] Furthermore, through the following formula, define the threat vector T i of the i-th obstacle source to the target UAV:

[0015]

[0016] where × is the vector cross product operation; Calculate the environmental threat vector as where N is the number of obstacle sources.

[0017] Furthermore, for any position point x in the risk avoidance space, satisfying ∥x - p t ∥ ≤ R, define its collision probability as P c (x):

[0018]

[0019] where Φ i (x) is the i-th obstacle source, its risk contribution to the position point x; ω i is the weight of the i-th obstacle source; α is the probability mapping steepness parameter, with a value range of 0.5 ≤ α ≤ 3.0; β is the probability mapping bias parameter, with a value range of -2.0 ≤ β ≤ 2.0.

[0020] Furthermore, the risk contribution Φ i (x) of the i-th obstacle source to the position point x is calculated using the following formula:

[0021]

[0022] where a i is the acceleration of the i-th obstacle source; t is the time variable; T is the upper limit of the set time range; Σ i (t) -1 is the uncertainty covariance matrix; where I is the 3×3 identity matrix; σ 0,i is the initial uncertainty parameter, with a value range from 0.1 to 0.5; σ 1,i is the uncertainty growth rate, with a value range from 0.01 to 0.1; Tr is the transpose operation of the vector.

[0023] Furthermore, the weight of the i-th obstacle source is calculated using the following formula:

[0024]

[0025] where L is the longest diameter of the target UAV.

[0026] Furthermore, the flight distance D from the target UAV to each candidate position is calculated through the position vector of the target UAV and the position vectors of each candidate position; the flight distance D should satisfy the following set constraints:

[0027]

[0028] where r is the distance integration variable; J 1 (·) is the first-order Bessel curve function; M is the mass of the target UAV; R oll is the rotor speed of the target UAV; F max is the maximum lift of the target UAV; F min is the minimum lift of the target UAV; z is the angle of attack of the target UAV.

[0029] Furthermore, the candidate position corresponding to the minimum product of the flight distance D and the collision probability P c (x) is used as the target position of the UAV, and the UAV is controlled to fly towards the target position to achieve emergency avoidance.

[0030] Adopting the above technical solutions, the present invention has the following beneficial effects: First of all, by defining a dynamic risk avoidance space centered on the target UAV, the present invention constructs a multi-dimensional space model that can reflect environmental threats in real time. This risk avoidance space can be dynamically updated with the position of the target UAV, enabling the UAV to always be in an efficient perception of the surrounding environment during flight. The construction of this dynamic risk avoidance space enables the system to accurately identify potential threats in a complex environment and lays a solid foundation for subsequent risk prediction and path planning. Compared with the traditional fixed-distance threshold method, the dynamic risk avoidance space of the present invention significantly improves the flexibility and accuracy of threat perception, especially in dynamic scenarios with multiple obstacles. Secondly, by introducing the environmental threat vector, the present invention realizes the comprehensive quantification of threats from multiple obstacle sources. The environmental threat vector not only considers the static characteristics of the obstacle source (such as position and size), but also combines the dynamic characteristics of the obstacle source (such as speed and direction), providing the system with comprehensive and accurate threat assessment capabilities. Through multi-dimensional parameter fusion and non-linear dynamic correction, this threat vector can reflect the comprehensive threat level of the obstacle source in real time. Through this method, the UAV can better judge the movement trend of the obstacle source and its impact on flight safety, avoiding misjudgment or omission caused by single-factor judgment in traditional methods. The present invention introduces a risk assessment method with a time dimension through the risk contribution model. The obstacle avoidance methods in the prior art usually only perform static assessment of threats at the current moment, while the present invention predicts the dynamic change trend of the obstacle source in the future period of time through time integration, and accurately calculates the cumulative risk of the obstacle source to the target position. This risk prediction with a time dimension greatly improves the forward-looking nature of the risk avoidance system, enabling the UAV to avoid high-risk areas that are about to appear in advance and avoiding the reaction delay problem caused by passive obstacle avoidance. Especially in a high-speed movement environment, this time prediction model buys precious reaction time for the UAV and improves the effectiveness of the risk avoidance strategy. Description of the Drawings

[0031] Figure 1 It is a schematic flowchart of a method for emergency risk avoidance based on UAV reconnaissance provided by an embodiment of the present invention. Detailed Embodiments

[0032] All features disclosed in this specification, or all steps in the disclosed methods or processes, except for mutually exclusive features and / or steps, can be combined in any manner.

[0033] Any feature disclosed in this specification (including any additional claims, abstract) can be replaced by other equivalent or similar-purpose alternative features unless specifically stated. That is, unless specifically stated, each feature is only an example of a series of equivalent or similar features.

[0034] Embodiment 1: Refer to Figure 1 , an emergency avoidance method based on UAV reconnaissance, the method comprising:

[0035] Step 1: With the target UAV as the center of the sphere and a set value as the radius, define the spherical UAV avoidance space range; within the UAV avoidance space range, construct an environmental threat vector through the position vectors and motion parameters of each obstacle source;

[0036] The target UAV uses an on-board or ground collaborative positioning system and sensor module to first determine its precise position and attitude information in three-dimensional space, and then uses this position as the center of the sphere to scan or detect the surrounding environment with a radius of the set value to ensure that the volume range of obstacle objects that are sufficient to threaten normal flight is included. In this way, on the one hand, it can ensure that no high-speed moving obstacle source that may approach or is about to enter the threat radius is missed, and on the other hand, it can maintain high-frequency dynamic monitoring of the local danger area while reducing redundant data, thus laying a solid foundation for subsequent collision probability assessment and control instruction planning. Within this spherical area, the present invention collects the relative position vectors of each obstacle source in real time through multi-sensor collaboration or the reconnaissance ability of the UAV itself, and combines its kinematic parameters (such as speed, acceleration, direction, mass characteristics, etc.) for centralized processing and calibration to generate a unified threat information set. In this way, the system can not only quickly judge the relative position relationship between obstacle sources, but also evaluate their potential influence on the flight safety of the UAV according to the respective motion states of the obstacle sources, and accordingly construct an environmental threat vector reflecting the overall risk situation. During this process, the position and attitude of the UAV are continuously updated, and the center of the sphere will naturally move in real time, so that the spherical avoidance space can closely fit the flight path and speed of the target UAV, making it fully reflect the flexible and autonomous perception characteristics in a complex or high-dynamic environment. Compared with most of the existing avoidance means that rely on simple fixed distances or single lidar ranging methods, the present invention realizes the comprehensive control of the accuracy and quantity of obstacle sources through the effective construction of the spherical space at the initial stage, which not only reduces the overall computational complexity of the system, but also significantly improves the risk attention to key areas, and provides high-quality data support for obtaining the collision probability and formulating the optimal flight path in a shorter time in the subsequent steps. Especially when the UAV is performing low-altitude flight, jungle patrol or urban reconnaissance and other scenarios, various obstacle sources may invade the flight space of the target UAV along different angles and speeds. The present invention can timely capture the existence of the obstacle source before it directly interferes with the flight path through the definition of this spherical area, and strive for valuable response time for the subsequent avoidance strategy planning.

[0037] Step 2: Calculate the collision probability of each position within the spherical UAV risk avoidance space by combining the environmental threat vector with the position vector and flight parameters of the target UAV; the collision probability describes the probability of the target UAV colliding with an obstacle source when passing through this position; screen out the positions where the collision probability is greater than or equal to the set collision threshold, and retain the positions where the collision probability is lower than the set collision threshold as candidate positions.

[0038] In Step 2, first obtain the position vector and flight parameters of the target UAV through its detection system, including the current speed, direction, attitude, etc. These parameters reflect the immediate flight state of the target UAV and are the basis for subsequent collision risk calculations. At the same time, the obstacle source data within the risk avoidance space is also obtained through multi-sensor fusion technology, including the spatial position vector, speed, acceleration, and possible size and mass characteristics of the obstacle source. The data of these obstacle sources is integrated into the environmental threat vector to quantitatively describe the overall distribution of potential threats within the risk avoidance space. Next, according to the relative position relationship between the target UAV and the obstacle source and the combination of flight parameters, the present invention uses a dynamic risk assessment model to calculate the collision probability of each position within the risk avoidance space. Specifically, the collision probability reflects the possibility of the target UAV colliding with the obstacle source within the risk avoidance space when passing through a specific spatial point, which is a specific manifestation of the probability distribution in a multi-dimensional dynamic system.

[0039] To achieve this goal, Step 2 adopts a probability modeling method that combines relative motion and geometric analysis. The calculation of the collision probability not only needs to consider the spatial distance between the UAV and the obstacle source, but also needs to introduce factors such as relative speed and motion direction for comprehensive judgment. The relative motion between the obstacle source and the target UAV can be used to predict whether the two will enter the same spatial area in the future. By performing projection analysis on the relative position vector and speed vector, it can be determined whether the obstacle source is moving towards the target UAV or whether there is an intersecting trajectory. In addition, to further improve the calculation accuracy, the present invention introduces a dynamic prediction model. By performing time integration on the current speed, acceleration, and motion direction of the obstacle source, predict its trajectory in the future time, and perform spatial intersection analysis with the motion path of the target UAV. In this process, the future position of the obstacle source is described in the form of a probability distribution, usually quantified by a Gaussian distribution or other suitable statistical models to quantify the uncertainty of the obstacle source position. Combining these prediction data, the collision probability of the target UAV colliding with the obstacle source at a specific position can be calculated more accurately.

[0040] In actual calculations, the present invention discretizes the avoidance space into a number of spatial points in order to evaluate the collision probability at each point. The collision probability of each spatial point depends on multiple factors, including the relative distance from all obstacle sources when the target UAV passes through this point, the movement direction, the degree of speed matching, and the threat weight of the obstacle sources, etc. The threat weight of the obstacle sources can be dynamically adjusted according to their speed, mass, and position uncertainty. For example, obstacle sources with a larger mass may have a higher threat weight because they pose a greater potential hazard to the UAV. Obstacle sources with a higher speed or moving towards the target UAV will also have an increased threat level accordingly. By weighted accumulation of the threat contributions of each obstacle source, the comprehensive collision probability at a specific position is finally obtained. This collision probability can be regarded as a field function, and its distribution within the avoidance space directly reflects the risk levels of different regions. Next, the present invention screens all position points within the avoidance space by setting a threshold for the collision probability. Any points with a collision probability greater than or equal to the threshold will be regarded as high-risk regions and excluded from the candidate paths; while points with a collision probability lower than the set threshold are retained as safe candidate positions that the target UAV can choose. This screening process ensures that in subsequent avoidance path planning, the target UAV will preferentially select those paths that are considered safe after risk assessment, thereby greatly reducing the possibility of collisions during the avoidance process.

[0041] Step 3: Calculate the flight distance of the target UAV passing through each candidate position based on the position vector and flight parameters of the target UAV, and calculate the optimal control instruction so that when the target UAV passes through the candidate position, the flight distance and the collision probability meet the set constraint conditions.

[0042] In this step, it is first necessary to use the current position vector and flight parameters of the target UAV to conduct a detailed analysis of each candidate position. These candidate positions are the results of collision probability screening in Step 2 and represent relatively safe areas within the current risk avoidance space. However, the assessment of collision probability is only the first step in determining the safety of these positions. The actual risk avoidance strategy needs to consider the feasibility of the flight path and the optimality of energy consumption on this basis. Therefore, the system must further calculate the flight distance from the current position of the target UAV to each candidate position as a key indicator for quantifying the path cost. The flight distance is not only a simple geometric distance measurement problem but also requires dynamic estimation in combination with the actual flight trajectory and flight dynamics model of the UAV. For a UAV, the flight path is usually not in a straight line form but is affected by factors such as the aircraft's heading change ability, speed limit, and real-time wind field. Therefore, calculating the flight distance must be based on a dynamic trajectory prediction model and implemented by combining multi-segment curves or optimal trajectory planning algorithms. In this process, the system will estimate the possible flight paths that the UAV may take to reach each candidate position according to the UAV's current velocity vector, heading angle, and maximum acceleration, thereby generating one or more potential trajectories and performing numerical integration or path optimization calculations on the actual length of each trajectory to obtain the corresponding flight distance value. After completing the calculation of the flight distance for each candidate position, the system will comprehensively compare this distance value with the collision probability obtained in Step 2 to construct a cost function incorporating the idea of multi-objective optimization as the core basis for generating the optimal control instruction. This cost function takes into account two key factors, namely the flight distance and the collision risk, and flexibly adjusts the importance of the two by setting different weight coefficients. In some mission scenarios, safety may be the primary consideration, and in this case, the cost function will assign a higher weight to the collision probability. In other emergency missions or energy consumption-limited situations, the flight distance may become the core optimization goal, thereby reducing energy consumption or shortening the flight time. Specifically, the system will use a non-linear optimization method to minimize the cost function, finally select the candidate position with the lowest cost, and generate the corresponding optimal control instruction based on this position.

[0043] In the process of generating the optimal control command, the system will convert the candidate position into a series of specific control inputs according to the dynamic constraints and flight characteristics of the UAV, such as the target heading angle, acceleration command, or motor speed command, etc., to ensure that the UAV can smoothly reach the selected candidate position along the optimal path. At the same time, in order to improve the control accuracy and real-time performance, the present invention adopts a method combining predictive control and feedback correction, that is, while generating the optimal control command, the system will continuously monitor the actual flight state of the UAV and dynamically correct the control command according to the data real-time feedback by the sensor. This method can effectively avoid path deviation caused by external interference or uncertain factors (such as sudden wind field changes, obstacle source acceleration, etc.), so as to ensure that the UAV is always on the preset optimal flight path. In addition, in the process of generating the control command, the present invention also introduces the concepts of energy management and dynamic resource allocation. By real-time monitoring and evaluating the flight energy consumption of the UAV, the intensity of the control command is accurately adjusted, avoiding the energy waste caused by excessive maneuvering in the traditional method, and ensuring the smoothness and efficiency of the avoidance action.

[0044] Embodiment 2: The motion parameters of the obstacle source include: obstacle source speed, obstacle source pointing angle, and obstacle source mass; the flight parameters of the target UAV include: target UAV mass, target UAV speed, and target UAV pointing angle; the environmental threat vector is defined by the following formula: Let be the position vector of the target UAV; let R>0 be the set radius; the UAV avoidance space range is defined as denotes the real number space; x is the position point in the UAV avoidance space range; ∥·∥ is the norm operation.

[0045] Specifically, in the avoidance method of the present invention, in order to ensure that the target UAV can accurately perceive the surrounding potential threats and make dynamic avoidance, it is first necessary to define a core area suitable for threat assessment in space - that is, the UAV avoidance space range. This space takes the current position of the target UAV as the center of the sphere and the preset radius R as the boundary, forming a three-dimensional spherical area to define the space range that the UAV needs to focus on during the current time period. In the formula, the position vector of the target UAV is set as Its specific form is three-dimensional coordinates (x t , y t , z t ), which describes the current accurate position of the UAV in the three-dimensional space. This position vector is the core basis for constructing the avoidance space, because it determines the center position of the avoidance space, enabling the entire avoidance space to be dynamically adjusted in real time as the target UAV moves, so as to always maintain accurate coverage of the potential threats around the flight path.

[0046] The radius R>0 of the avoidance space is a set positive real number, usually set according to the flight speed, reaction time of the UAV and the complexity of the surrounding environment. A larger radius R can provide earlier threat warnings for the UAV and is suitable for high-speed flight missions in open environments, while a smaller radius is more suitable for missions in densely built urban areas or narrow spaces to reduce redundant data collection and improve computational efficiency. Through this radius value, the formula defines a three-dimensional space set All points x in this set are regarded as potential passing positions within the avoidance space of the UAV. Among them, ||x - p t || represents the Euclidean distance between the position vector x and the target UAV position vector p t Here, the norm operator ||·|| is used for calculation, ensuring the accuracy and consistency of distance measurement. Through this set S, the system can clearly define the spatial range that the target UAV needs to focus on at the current moment, providing clear boundary conditions for subsequent threat analysis.

[0047] In this avoidance space, each position point x may become a potential flight path point of the target UAV. Therefore, it is necessary to quantitatively evaluate its safety in combination with the motion parameters of the obstacle source. The motion parameters of the obstacle source include speed v i , pointing angle θ i , and mass m i . These parameters jointly determine the dynamic characteristics and threat level of the obstacle source. Specifically, the obstacle source speed v i reflects its moving rate per unit time. The greater the speed, the more significant the threat of the obstacle source to the UAV; the pointing angle θ i describes the moving direction of the obstacle source. The relative angle with respect to the target UAV can determine whether the obstacle is approaching or moving away from the UAV, thus assisting in calculating the collision probability; while the mass m i is used to measure the physical impact intensity or damage degree that the obstacle may cause to the UAV. Obviously, an obstacle with a larger mass has a higher risk weight. The flight parameters of the target UAV include mass m t , speed v t , and pointing angle θ t . These parameters determine the maneuverability of the target UAV and its response speed to environmental threats. The mass m t is related to flight energy consumption and inertia. A larger mass means that the UAV needs to consume more energy when performing avoidance maneuvers, and it will also increase the response delay in emergency avoidance actions; the speed v t is the current flight speed of the UAV. The higher the speed, the earlier the UAV needs to perceive and avoid potential threats to ensure sufficient reaction time to complete maneuvering operations; the pointing angle θ tThe heading angle of the UAV is described, and by combining with the movement angle of the obstacle source, the cross - risk of the trajectories of both sides can be better judged. Based on the above - mentioned movement parameters, the environmental threat vector is comprehensively calculated through a mathematical model, which reflects the overall threat level of the obstacle source to the target UAV in the risk - avoidance space. This threat vector is not a simple superposition of the characteristics of a single obstacle, but a comprehensive quantification result through non - linear weight combination and dynamic prediction correction. The core of the calculation of this vector lies in projecting the movement state of each obstacle source to each position point x within the range of the risk - avoidance space, so as to judge whether there is a high collision probability at these points. Through this dynamic threat vector, the present invention can globally evaluate all potential flight path points in the risk - avoidance space and provide a quantitative risk basis for subsequent steps, ensuring that the UAV can always select the safest and optimal flight path in a complex environment.

[0048] Embodiment 3: Let the position vector of the i - th obstacle source be Let be the velocity vector of the i - th obstacle source, and its modulus is denoted as v i ; Let θ i ∈[0, 2π] be the pointing angle of the i - th obstacle source; Let m i >0 be the mass of the i - th obstacle source; Let be the velocity vector of the target UAV; Let r i =p t -p i be the relative position between the target UAV and the i - th obstacle source; Let d i =∥r i ∥ be the relative distance between the target UAV and the i - th obstacle source; In the risk - avoidance space S, calculate the threat coefficient Φ i :

[0049]

[0050] where, r i,y is the Y - axis component of r i ; r i,x is the X - axis component of r i .

[0051] Specifically, in the emergency risk - avoidance method based on UAV reconnaissance proposed by the present invention, in Embodiment 3, by constructing the threat coefficient Φ iTo quantify the risk level of each obstacle source to the target UAV, providing a basis for high-precision risk assessment for the UAV's risk avoidance strategy. The threat coefficient formula integrates multi-dimensional parameters such as the position, speed, direction, and mass of the obstacle source to reflect the dynamic threat level of the obstacle source in the UAV's risk avoidance space. Different from traditional methods based only on distance or simple collision detection, the present invention organically integrates multiple physical factors through a complex mathematical model, making the threat assessment more accurate and flexible, especially suitable for real-time risk avoidance in a multi-obstacle source and high-dynamic environment.

[0052] First, the core idea of the threat coefficient formula is to quantify the potential threat of the obstacle source to the UAV by calculating the relative distance, speed difference, and direction relationship between the target UAV and the obstacle source. Let the current position vector of the target UAV be The position vector of the i-th obstacle source is The relative position vector between the two is r i = p t - p i , and its modulus length d i = ∥r i ∥ represents the Euclidean distance between the UAV and the obstacle source. Obviously, the smaller the distance, the greater the threat, so the distance is an important basis for the threat coefficient. However, simple linear distance attenuation cannot accurately reflect the complex dynamic environment. The present invention enhances the accuracy and dynamic response ability of the threat assessment by introducing a non-linear transformation to the distance. In the threat coefficient formula, first, through This term comprehensively weighs the mass m i of the obstacle source and the relative distance d i . The mass of the obstacle source reflects the potential intensity of the physical impact. The greater the mass, the higher its threat coefficient; while the distance d i controls the threat attenuation through the square reciprocal relationship, and at the same time introduces a smoothing factor of 10 -5 to avoid the zero-distance singularity. This distance attenuation mode makes the threat coefficient increase rapidly at close range, reflecting the high-risk characteristics of the obstacle source approaching the UAV in the real scene, and rapidly attenuating at long range, reducing the interference of low-risk obstacle sources far from the UAV to the system judgment. In addition, to further improve the ability of the threat coefficient to reflect the dynamic characteristics of the obstacle source, the present invention introduces a Gaussian attenuation term This term limits the spatial range of the threat influence through the Gaussian function, emphasizing the local threat environment around the UAV. When the distance d i is small, the value of the Gaussian function is close to 1, indicating that the obstacle source is within the high-risk range; while when the distance increases, this value rapidly decays to close to zero, effectively filtering the low-risk influence of distant obstacles. This Gaussian-shaped attenuation not only reduces the unnecessary computational burden but also improves the sensitivity of the risk avoidance system to high-threat obstacle sources at close range.

[0053] To more accurately capture the relative motion state between the obstacle source and the UAV, the present invention further introduces a relative velocity term (v i -v t )·r i , that is, the dot product of the obstacle source velocity vector and the relative position vector, which reflects the movement trend of the obstacle source in the relative position direction. When the obstacle source approaches the target UAV, this value is positive, indicating a high collision risk; when the obstacle source moves away, this value is negative, and the collision risk is significantly reduced. This velocity term is smoothed by the hyperbolic tangent function tanh to keep it within a finite interval at all times, avoiding numerical anomalies caused by excessive speed differences. This smoothing process can provide a more sensitive dynamic response while maintaining computational stability, enabling threat assessment to maintain high accuracy under different speed conditions. In the last part of the threat coefficient, the present invention quantifies the angular relationship between the movement direction of the obstacle source and the relative position direction of the target UAV through the angle correction term cos(θ i -arctan 2(r i,y ,r i,x ). The pointing angle θ i of the obstacle source represents its movement direction on the horizontal plane, while the polar angle arctan 2(r i,y ,r i,x ) of the relative position vector describes the relative azimuth angle between the obstacle source and the target UAV on the horizontal plane. By calculating the cosine value between the two, the present invention can determine whether the obstacle source is approaching the relative position direction of the target UAV. When the movement direction of the obstacle source is consistent with the relative position direction, the cosine value is close to 1, indicating a high threat; when the two directions are opposite, the cosine value tends to 0, indicating a low threat. This angle correction greatly improves the spatial direction resolution ability of the threat coefficient. Especially in a multi-obstacle source environment, it can effectively identify obstacles in high-risk directions and help the UAV avoid the most threatening targets first.

[0054] Embodiment 4: Define the threat vector T i of the i-th obstacle source to the target UAV through the following formula:

[0055]

[0056] where × is the vector cross product operation; the calculated environmental threat vector is where N is the number of obstacle sources.

[0057] Specifically, the construction of the threat vector is based on two main parts: the radial threat and the lateral rotation effect. First, the radial threat term reflects the main threat intensity of the obstacle source to the UAV along the relative position direction. The relative position vector between the target UAV and the i-th obstacle source is r i = p t - p i , and d i = ∥r i ∥ represents the distance between the two. To capture the threat level of the obstacle source to the UAV, the first part of this threat vector uses the threat coefficient Φ i multiplied by the unitized relative position vector to form a radial component that always points to the obstacle source. The magnitude of this component is directly related to the threat coefficient Φ i , and the threat coefficient combines various factors such as the mass, speed, direction of the obstacle source, and the relative distance from the UAV, comprehensively reflecting the comprehensive threat level of the obstacle source. The role of this part can be understood as the radial attraction effect: when the obstacle source approaches the UAV, the magnitude of this vector rapidly increases, indicating a high risk of direct collision for the UAV; when the obstacle source moves away, this component tends to decrease or even approach zero, indicating a significant reduction in threat. Through this radial description, the UAV can accurately identify the spatial direction of the threat source, so as to preferentially avoid high-threat areas when planning an evasion path. However, relying solely on the radial threat is not sufficient to comprehensively describe the dynamic impact of the obstacle source. Therefore, the present invention further introduces the lateral rotation effect into the threat vector to capture the lateral impact brought by the relative speed change between the obstacle source and the UAV. This part is achieved through the vector cross product operation, and the specific form is r i ×(v i - v t ), that is, the cross product result of the relative position vector and the velocity difference vector. Geometrically, the vector cross product can be understood as describing the normal vector of the plane formed by two vectors, and the result is a vector perpendicular to the plane where the relative position and the velocity difference are located. This lateral component is used to quantify the lateral threat effect of the obstacle source, especially in an environment with multiple obstacle sources. For example, when the obstacle source moves along a direction parallel but slightly deviated from the trajectory of the target UAV, the radial threat may be small, but the lateral rotation effect will increase rapidly, prompting the system to pay attention to the potential lateral collision risk. To balance the intensity of this component, the present invention introduces as a normalization term in the formula to prevent the lateral threat from increasing infinitely at close range, while ensuring the numerical stability and calculation reliability.

[0058] After adding the above two parts, the comprehensive threat vector T i, this vector not only describes the threat intensity of the obstacle source to the target UAV, but also clearly points out the specific direction of the threat, enabling the UAV to make more reasonable maneuvering decisions based on the overall threat situation when avoiding risks. In the risk avoidance space, multiple obstacle sources may simultaneously affect the target UAV. Therefore, in the present invention, by superimposing the threat vectors of all obstacle sources, an environmental threat vector is constructed. where N is the number of obstacle sources. This environmental threat vector is a comprehensive description of the overall threat situation in the current risk avoidance space. It not only quantifies the overall threat level but also provides a reference for the UAV on the direction of the overall threat. In practical applications, the direction of the environmental threat vector E often points to the area where the threats are most concentrated, and its magnitude reflects the size of the overall risk. The UAV can plan the optimal maneuvering path according to the reverse direction of this vector, thereby effectively avoiding high-risk areas. The threat vector method of the present invention significantly improves the dynamic risk avoidance ability of the UAV in complex environments. Compared with traditional risk avoidance methods based on single parameters or scalar threat values, the threat vector can not only provide a more comprehensive and accurate threat assessment but also provide strong support for the path planning of the UAV through spatial direction information, especially having outstanding advantages in multi-obstacle source and high-dynamic environments. In addition, the introduction of the threat vector brings a high degree of flexibility and scalability to the UAV's autonomous risk avoidance strategy. The system can adapt to different mission requirements and environmental changes by adjusting the parameter weights in the threat coefficient, thus achieving efficient autonomous risk avoidance in various scenarios such as urban reconnaissance, low-altitude flight, and complex terrain patrol.

[0059] Example 5: For any position point x in the risk avoidance space, satisfying ∥x - p t ∥ ≤ R, its collision probability is defined as P c (x):

[0060]

[0061] where, Φ i (x) is the i-th obstacle source, its risk contribution to the position point x; ω i is the weight of the i-th obstacle source; α is the steepness parameter of the probability mapping, and its value range is 0.5 ≤ α ≤ 3.0; β is the bias parameter of the probability mapping, and its value range is -2.0 ≤ β ≤ 2.0.

[0062] Specifically, in the formula of Example 5, the collision probability P c (x) is the risk assessment value corresponding to any position point x of the target UAV in the risk avoidance space . To accurately reflect the threat degree of the obstacle source to this position, the system first analyzes all obstacle sources one by one and calculates the risk contribution Φ i(x). This risk contribution is not only affected by the relative distance between the obstacle source and the location point, but also closely related to various dynamic parameters of the obstacle source, such as its motion state, speed, direction, mass, etc. In other words, Φ i (x) can be regarded as describing the risk field exerted by the obstacle source on the target location point in space. The closer the obstacle source is to the location point and the more dangerous its motion state is, the higher Φ i (x) is, indicating a higher risk at that location point. However, the overall risk faced by the UAV cannot be accurately described solely by the risk contributions of each obstacle source. Therefore, the present invention introduces the obstacle source weight ω i , which is used to weight the risk contributions of different obstacle sources. The design of the weight ω i fully considers the individual characteristics of the obstacle source and its potential impact on the UAV, such as the mass, dynamic characteristics of the obstacle source, and its priority within the avoidance space. Through this weighting method, the system can flexibly adjust the degree of attention paid to various obstacle sources in a multi-obstacle environment, ensuring preferential avoidance of high-risk obstacles and avoiding unnecessary avoidance actions in low-risk areas. The setting of the weight ω i provides the present invention with a high degree of scalability and adaptability, enabling it to be flexibly applied in various flight scenarios (such as low-altitude urban flight, forest cruising, etc.).

[0063] After calculating the weighted risk values of all obstacle sources at the location point x, the system performs a probability mapping on this cumulative risk value through a non-linear function, converting it into a collision probability value between 0 and 1. Specifically, the non-linear mapping function adopts a form similar to the Logistic function, and its expression is It can smoothly map any cumulative risk value into a finite interval. This non-linear mapping has significant advantages. Firstly, it can provide a more sensitive response in both high-risk and low-risk regions, avoiding the saturation problem of traditional linear functions in extreme risk states. Secondly, it can flexibly change the shape of the mapping curve by adjusting the steepness parameter α and the bias parameter β, so as to adapt to different risk assessment requirements. The steepness parameter α is the key factor determining the sensitivity of the probability mapping function, and its value range is usually between 0.5 ≤ α ≤ 3.0. A smaller α value will make the mapping curve smoother, which is suitable for scenarios that require a wide range of assessment of potential risks; while a larger α value will make the curve steeper, and the collision probability will quickly jump to a high value when the risk accumulation value approaches a certain critical point, which is suitable for high-sensitivity scenarios. For example, when the UAV is in a narrow space or an area with dense obstacle sources, setting a higher α value can enable the system to respond quickly and avoid missing key avoidance opportunities. On the other hand, the bias parameter β is used to adjust the horizontal offset of the risk mapping curve, and its value range is -2.0 ≤ β ≤ 2.0. By changing the value of β, the starting value and trigger critical point of the collision probability can be adjusted. For example, when β > 0, a significant collision probability will only be triggered when the risk accumulation value is relatively high, while when β < 0, the system will also be sensitive to smaller risk values. This highly flexible non-linear probability mapping model not only improves the adaptability of the present invention in various complex scenarios, but also can effectively avoid the phenomena of high-frequency misjudgment or missed judgment in traditional avoidance methods. Especially in an environment with multiple obstacle sources, by smoothly mapping the cumulative risk value, the system can accurately distinguish high-risk and medium-low-risk regions, thus helping the UAV to make a more accurate path selection during the avoidance process. Finally, the collision probability P c (x) provides an intuitive and quantifiable risk assessment index for each position point, facilitating the system to directly use this probability value as a constraint condition or optimization goal in subsequent optimal path planning and control instruction generation, ensuring that the UAV is always on the safest flight path.

[0064] Example 6: The risk contribution Φ i (x) of the i-th obstacle source to the position point x is calculated using the following formula:

[0065]

[0066] Among them, a i is the acceleration of the i-th obstacle source; t is the time variable; T is the upper limit of the set time range; Σ i (t) -1 is the uncertainty covariance matrix; Among them, I is a 3×3 identity matrix; σ 0,iis the initial uncertainty parameter, with a value range of 0.1 to 0.5; σ 1,i is the uncertainty growth rate, with a value range of 0.01 to 0.1; Tr is the transpose operation of a vector.

[0067] Specifically, in this model, the position of the obstacle source is not just a fixed point, but a predicted trajectory that changes dynamically over time. The position of the obstacle source at a future time is represented by , which reflects the quadratic motion trajectory of the obstacle source in three-dimensional space. This trajectory formula not only considers the initial position p i and velocity v i , but also further introduces the influence of acceleration a i , so as to more realistically simulate the motion state of the obstacle source in different complex environments. The introduction of acceleration enables the model to adapt to non-uniform motion situations such as approaching rapidly, decelerating and turning, significantly improving the accuracy and flexibility of predicting the future position of the obstacle source. However, there is always a certain degree of uncertainty in the future position of the obstacle source, and this uncertainty may come from sensor measurement errors, environmental interference or the irregular motion of the obstacle source itself. The present invention uses the uncertainty covariance matrix to describe the evolution characteristics of this uncertainty over time. The initial uncertainty parameter σ 0,i in the covariance matrix reflects the measurement accuracy of the obstacle source position at the initial time t = 0, while the uncertainty growth rate σ 1,i describes how this uncertainty gradually increases over time. This design fully considers the uncertainty diffusion effect of the obstacle source motion, enabling risk assessment to not only focus on the threats at the current moment, but also dynamically predict the potential impact of the obstacle source on the UAV in a future period of time.

[0068] To accurately evaluate the risk contribution of the obstacle source at the position point x, the present invention uses a three-dimensional Gaussian distribution as the probability density function to describe the probability that the obstacle source appears near the position point x at a future time t. The exponential term (x - p i (t)) T Σ i (t) -1 (x - p i (t)) in the Gaussian distribution calculates the distance between the position point x and the predicted position p iThe Mahalanobis distance between (t) not only considers the geometric distance between the two points, but also combines the covariance matrix to weight the uncertainty, which can more accurately measure the risk intensity of the obstacle source in different directions. For example, if the uncertainty in a certain direction is large, the risk contribution of the Mahalanobis distance in that direction will be reduced accordingly, so that the risk assessment is more in line with the actual situation. In order to obtain the comprehensive risk contribution of the obstacle source in the entire time range, the present invention integrates the Gaussian distribution over the time interval [0, T] and accumulates the risk contribution at each moment t. This integration operation can not only capture the instantaneous risk of the obstacle source to the location point at multiple moments, but also accurately reflect the cumulative effect of the risk. For example, in the case where the obstacle source continues to approach the target location, the time integration will significantly increase the risk value of the location point, prompting the system to give priority to avoiding this high-risk area. On the contrary, if the obstacle source is close to the location point only for a short time and then moves away, the cumulative risk is relatively low, and the system can more flexibly choose a risk avoidance strategy. This time integration model has significant advantages in complex dynamic environments. Traditional static distance or speed threshold judgment methods can usually only provide risk assessment at the current moment and cannot effectively predict the threats that may be posed by the future movement of obstacles. The present invention, by introducing risk integration in the time dimension, can capture future high-risk scenarios in advance and gain valuable risk avoidance time for the drone. Especially in a multi-obstacle source environment, different obstacle sources may have an impact on the target position in different time periods. Through time integration, the system can accurately quantify the risk contribution of each obstacle source on different time scales to form a more comprehensive environmental risk situation awareness. In addition, the risk contribution formula in the present invention is highly adaptable and extensible. By adjusting the time range T and the parameter σ in the covariance matrix 0,i and σ 1,i , the system can flexibly adjust the risk assessment strategy according to different mission requirements and environmental complexity. For example, in an open environment, a longer time range T and a smaller uncertainty parameter can be selected to more accurately predict the threat of distant obstacle sources; in a highly dynamic environment such as low-altitude flight in a city, the time range can be shortened and the uncertainty parameter can be appropriately increased to enhance the system's sensitivity to short-term high-risk scenarios. This flexible adjustment capability enables the present invention to adapt to a variety of complex scenarios and provides strong technical support for autonomous risk avoidance of drones in different environments.

[0069] Embodiment 7: The weight of the i-th obstacle source is calculated using the following formula:

[0070]

[0071] Where L is the longest diameter of the target UAV.

[0072] Specifically, the main parameters in the weight formula represents the square of the relative distance between the target UAV and the i-th obstacle source. This term plays a basic distance attenuation role, ensuring that the influence of the obstacle source on the UAV gradually decreases as the distance increases, thus avoiding excessive interference of low-risk obstacle sources at long distances on the system decision-making. However, in order to better reflect the actual threats in a dynamic environment, the present invention introduces a variety of dynamic correction terms on the basis of distance, making the weight calculation not only depend on the distance, but also be able to dynamically adapt to the relative motion state and spatial position relationship of the obstacle source. The hyperbolic tangent function in the formula is a key dynamic correction term, used to reflect the influence of the relative velocity between the obstacle source and the target UAV on the weight. Here, (v i -v t )·r i represents the projection value of the velocity difference in the direction of the relative position vector, describing the movement trend of the obstacle source along the direction of the UAV's relative position. If the projection value is positive, it means that the obstacle source is approaching the target UAV, and the collision risk is relatively high; if the projection value is negative, it indicates that the obstacle source is moving away from the target UAV, and the risk is relatively low. The hyperbolic tangent function plays a smoothing role here, mapping the velocity projection value to a finite interval, avoiding unstable calculations caused by too large or too small numerical values. At the same time, by adjusting the sensitivity of the function through the coefficient 0.3, the system can dynamically adjust the response intensity to the relative velocity according to the environment, ensuring timely response in high-risk scenarios.

[0073] In addition, in order to further improve the spatial direction resolution ability of the weight calculation, an angle correction term cos(θ i -arctan 2(r i,y ,r i,x )) is also introduced in the formula. This term calculates the movement direction θ iThe angle between the polar angle of the relative position vector quantifies the directional threat of the obstacle source. When the moving direction of the obstacle source is consistent with the relative position direction of the target UAV, the angle is close to zero and the cosine value tends to 1, indicating a higher threat; while when their moving directions are opposite, the cosine value tends to zero, indicating a lower threat. This angular correction term can effectively distinguish obstacle sources with different moving directions, helping the UAV to preferentially avoid high-risk targets approaching head-on rather than low-risk obstacles far from the path. To make the weight calculation more in line with the actual spatial scale, the formula also introduces the geometric parameter L of the target UAV, that is, the longest diameter of the UAV. This parameter is used to describe the spatial size of the UAV itself, ensuring that the actual occupied space of the UAV is considered in the weight calculation, thereby avoiding the collision risk that may be caused by ignoring its geometric characteristics. In practical applications, the introduction of L enables the collision avoidance strategy to be adaptively adjusted for UAVs of different types and sizes. For example, larger UAVs have a higher collision probability in complex environments, so the weight calculation will be more sensitive, while smaller UAVs can flexibly avoid in a narrower space. The vector cross product operation ∥r i ×(v i -v t )∥ in the formula further enhances the spatial dynamic characteristics of the weight calculation. The geometric meaning of the vector cross product is to calculate the normal vector of the plane formed by two vectors, and its modulus reflects the magnitude of the vertical component between the two vectors. In this formula, the cross product result of the vector r i and the velocity difference vector v i -v t is used to quantify the lateral movement trend of the obstacle source relative to the target UAV. When the cross product value is large, it indicates that the obstacle source has strong dynamic changes in the lateral direction and may pose a lateral threat to the UAV, so its weight needs to be increased; while when the cross product value is close to zero, it indicates that the obstacle source has little change in the lateral direction and the threat level is relatively low. Through this lateral correction term, the system can more comprehensively capture the dynamic changes of the obstacle source in multi-dimensional space and avoid misjudgment caused by simply relying on the radial distance. Finally, in the weight calculation formula, multiplying by the collision avoidance space radius R as a normalization factor ensures the consistency of the weight calculation results at different spatial scales. The radius R of the collision avoidance space is a parameter set according to the mission requirements and environmental complexity. A larger R value is suitable for long-distance monitoring in open areas, while a smaller R value is suitable for short-distance precise collision avoidance in complex environments such as cities or forests. Through this normalization process, the system can maintain consistent threat assessment accuracy in different scenarios, ensuring that the weight calculation results have high comparability and robustness.

[0074] Example 8: Calculate the flight distance D from the target UAV to each candidate position through the position vector of the target UAV and the position vectors of each candidate position; the flight distance D should satisfy the following set constraints:

[0075]

[0076] where r is the distance integration variable; J 1 (·) is the first-order Bessel curve function; M is the mass of the target UAV; R oll is the rotor speed of the target UAV; F max is the maximum lift of the target UAV; F min is the minimum lift of the target UAV; z is the angle of attack of the target UAV.

[0077] Specifically, the flight distance D must be restricted by the power output ability of the target UAV, and the power output ability is mainly determined by the lift range F max -F min and the rotor speed R oll When the UAV performs emergency evasion, it needs to quickly adjust the flight attitude and speed within a short time, so as to change the original flight path. However, this rapid maneuver is not unlimited. Especially when making a large-angle turn or climbing, if the flight distance exceeds the range allowed by the physical performance, it may lead to insufficient rotor power or UAV out of control. The present invention accurately models the vertical lift change of the UAV in the flight distance constraint formula by introducing the lift parameters F max and F min , so as to ensure that the path planning does not exceed the lift limit and avoid the UAV falling due to insufficient lift. At the same time, the flight distance D is also dynamically affected by the mass M and the angle of attack Z of the target UAV. The mass M directly affects the lift and power output required by the UAV during maneuvering. The larger the mass, the more obvious the inertial effect of the UAV during high-speed turning or climbing, thus significantly increasing the path length and power consumption. In order to accurately consider this inertial effect in the flight path selection, the present invention dynamically adjusts the upper limit value of the flight distance through the exponential function , so that when the UAV with a large mass performs a maneuver, the path length is reasonably compressed, thereby reducing the additional risks caused by inertia. At the same time, the introduction of the angle of attack Z can simulate the change of the aerodynamic characteristics of the UAV in different flight postures, ensuring that the path constraint can be flexibly adjusted in different flight states. An excessive angle of attack may lead to lift loss and increase the stall risk, while an insufficient angle of attack may affect the maneuverability. Therefore, the path planning must be carried out within the safe range of the angle of attack.

[0078] It should be noted that the constraint of the flight distance D is not simply restricted by a fixed threshold, but is dynamically generated by path integration. This integration operation gradually integrates the distance variable r from 0 to R, simulating the change of the path length of the UAV during actual flight. The sine function term It plays a role in periodic correction in the integral expression, describing the fluctuation characteristics and periodic changes of the path curve. The introduction of this term makes the path length calculation more in line with the actual flight trajectory. Especially in the scenarios of low-altitude flight in cities or complex terrains, drones often cannot fly along a completely straight path, but need to bypass obstacles, presenting a certain arc or fluctuating trajectory. The sine function term can effectively correct the periodic changes of the path length, avoid over-long or over-short path estimations on long paths, and ensure more accurate path planning. In addition, the Bessel function is introduced to further improve the ability to capture local detailed changes in the path. During the actual flight process, the path may be affected by the wind field, airflow disturbances or obstacle source interferences, resulting in local minor path deviations. Bessel functions are often used to describe such oscillations and detailed changes. Its role in the integral formula is to enhance the ability to perceive small fluctuations in the path, ensuring that the path planning can reflect the real flight conditions, rather than just being based on idealized linear assumptions. This sensitivity to local path fluctuations is particularly important in dynamic risk avoidance scenarios, which can help drones better cope with short-term high-frequency disturbances and thus maintain flight stability in complex environments. To make the flight distance constraint have stronger physical significance and adaptive ability, the present invention multiplies the maximum outer diameter L of the target drone and the risk avoidance space radius R as the normalization factor to ensure the consistency of the path constraint at different spatial scales. The outer diameter L of the drone determines its maneuverability in narrow spaces. Smaller drones have higher maneuverability in complex environments, so the path constraint can be relatively relaxed; while larger drones require more stringent path constraints to prevent collision risks due to insufficient space. The risk avoidance space radius R reflects the spatial scope of the current task. A larger radius is suitable for long-distance monitoring in open areas, while a smaller radius is suitable for close-range precise risk avoidance in complex environments such as cities or jungles.

[0079] Example 9: The candidate position corresponding to the minimum product of the flight distance D and the collision probability P c (x) is taken as the target position of the drone, and the drone is controlled to fly towards the target position to achieve emergency risk avoidance.

[0080] Specifically, the candidate positions are the low-risk position points screened through the previous steps, and they all meet certain collision probability and flight performance constraints. However, these candidate positions are not exactly the same as the target positions in the optimal risk avoidance path. Only considering the shortest flight distance D can improve the maneuvering efficiency of the drone, but it may lead the drone to high-risk areas and increase the possibility of collision; while only considering the candidate position with the minimum collision probability may cause the drone to choose an overly conservative path, significantly prolonging the flight time and energy consumption, and even missing the risk avoidance opportunity in a dynamic environment. Therefore, the present invention proposes to use the flight distance D and the collision probability Pc The product of (x) is used as a comprehensive optimization index to ensure the flexible balance between the safety and efficiency of the flight path under different risk scenarios. Specifically, the flight distance D reflects the physical path length of the target UAV from the current position to the candidate position and is an important indicator for measuring the path cost. The collision probability P c (x) describes the probability that the target UAV encounters an obstacle source during flight and is the core parameter for safety assessment. By multiplying these two key parameters, a comprehensive cost function can be formed. This comprehensive cost function can not only dynamically reflect the overall risk level of path selection but also provide an accurate optimization basis in scenarios with multiple obstacle sources and complex paths. The process of selecting the target position is actually a process of minimizing this cost function, that is, among all candidate positions, finding the x that minimizes Cost(x) and using this position as the target position to guide the UAV's flight.

[0081] This optimization strategy has significant practical significance. First, when a certain path among the candidate positions has a long flight distance but an extremely low collision probability, the comprehensive cost of this path may still be small and thus be selected as the optimal path. This selection can effectively avoid high-risk areas and give priority to ensuring the flight safety of the UAV. On the contrary, if a certain candidate position has a short distance but a high collision probability, then the comprehensive cost of this position may increase and it will ultimately be excluded by the system. This multi-objective optimization method can adapt to different risk scenarios in various complex environments by automatically balancing the path length and safety, significantly enhancing the robustness and flexibility of the UAV's risk avoidance decision-making. At the control level, the present invention adopts a path tracking algorithm based on real-time feedback, taking the optimal target position as the flight control target of the UAV. The UAV uses on-board sensors to continuously monitor its own position and attitude and constantly corrects its flight trajectory to ensure that it always maneuvers towards the target position. To improve the control accuracy, the system introduces a strategy combining predictive control and adaptive adjustment. By continuously updating the dynamic model of the UAV, it corrects the deviation that may be caused by external interference or environmental changes to ensure that the flight path always remains within the optimal trajectory. At the same time, to prevent sudden changes in obstacle sources, the system continuously monitors environmental threats during flight. If a new high risk appears in the optimal path, it immediately recalculates the comprehensive cost function and updates the target position to achieve dynamic risk avoidance.

[0082] Although the specific embodiments of the present invention have been described above, those skilled in the art should understand that these specific embodiments are merely illustrative. Without departing from the principle and essence of the present invention, those skilled in the art can make various omissions, substitutions, and changes to the details of the above methods and systems. For example, combining the above method steps so as to perform substantially the same function in a substantially the same manner to achieve substantially the same result falls within the scope of the present invention. Therefore, the scope of the present invention is only defined by the appended claims.

Claims

1. An emergency risk avoidance method based on drone reconnaissance, characterized in that: The method comprises: Step 1: Define the spherical drone avoidance space range with the target drone as the sphere center and the set value as the radius; within the drone avoidance space range, construct the environmental threat vector through the position vector and motion parameters of each obstacle source; Step 2: Calculate the collision probability of each position within the spherical UAV risk avoidance space by combining the environmental threat vector with the position vector and flight parameters of the target UAV; the collision probability describes the probability of the target UAV colliding with the obstacle source when passing through the position; screen out the positions corresponding to the collision probability greater than or equal to the set collision threshold, and retain the positions corresponding to the collision probability lower than the set collision threshold as candidate positions; Step 3: Calculate the flight distance of the target UAV through each candidate position through the position vector and flight parameters of the target UAV, and calculate the optimal control instructions so that when the target UAV passes through the candidate position, the flight distance and collision probability meet the set constraints.

2. The emergency avoidance method based on drone reconnaissance as claimed in claim 1, characterized in that: The motion parameters of the obstacle source include: obstacle source speed, obstacle source pointing angle and obstacle source mass; the flight parameters of the target UAV include: target UAV mass, target UAV speed and target UAV pointing angle; the environmental threat vector is defined by the following formula: is the position vector of the target drone; let R>0 be the set radius; the drone's avoidance space range is defined as represents the real number space; x is the position point in the UAV's risk avoidance space; ∥·∥ is the norm operation.

3. The emergency avoidance method based on drone reconnaissance as claimed in claim 2, characterized in that: Assume the position vector of the i-th obstacle source is set up is the obstacle source velocity of the ith obstacle source, and its modulus is recorded as v i ; Let θ i ∈[0,2π] is the obstacle source pointing angle of the i-th obstacle source; let m i >0 is the obstacle source quality of the i-th obstacle source; let is the velocity vector of the target UAV; let r i =p t -p i is the relative position between the target UAV and the i-th obstacle source; d i =∥r i ∥ is the relative distance between the target UAV and the i-th obstacle source; in the risk avoidance space S, calculate the threat coefficient Φ of the i-th obstacle source i : Among them, r i,y For r i The Y-axis component of r i,x For r i The X-axis component of .

4. The emergency avoidance method based on drone reconnaissance as claimed in claim 3 is characterized in that: The threat vector T of the i-th obstacle source to the target drone is defined by the following formula: i : Where × is the vector cross multiplication operation; the calculated environmental threat vector is Where N is the number of obstacle sources.

5. The emergency avoidance method based on drone reconnaissance as claimed in claim 4, characterized in that: For any point x in the safe haven space, ∥xp t ∥≤R, define its collision probability as P c (x): Among them, Φ i (x) is the i-th obstacle source and its risk contribution to the location point x; ω i is the weight of the i-th obstacle source; α is the probability mapping steepness parameter, and its value range is 0.5≤α≤3.0; β is the probability mapping bias parameter, and its value range is -2.0≤β≤2.

0.

6. The emergency avoidance method based on drone reconnaissance as claimed in claim 4, characterized in that: The i-th obstacle source contributes Φ to the risk of the location point x i (x) is calculated using the following formula: in, a i is the acceleration of the ith obstacle source; t is the time variable; T is the upper limit of the set time range; Σ i (t) -1 is the uncertainty covariance matrix; Where I is the 3×3 identity matrix; σ 0,i is the initial uncertainty parameter, ranging from 0.1 to 0.5; σ 1,i is the uncertainty growth rate, ranging from 0.01 to 0.1; Tr is the transpose operation of the vector.

7. The emergency avoidance method based on drone reconnaissance as claimed in claim 6, characterized in that: The weight of the i-th obstacle source is calculated using the following formula: Where L is the longest diameter of the target UAV.

8. The emergency avoidance method based on drone reconnaissance as claimed in claim 7, characterized in that: The flight distance D from the target UAV to each candidate position is calculated by the position vector of the target UAV and the position vector of each candidate position. The flight distance D must meet the following constraints: Where r is the distance integral variable; J1(·) is the first-order Bezier curve function; M is the mass of the target UAV; R oll is the rotor speed of the target UAV; F max is the maximum lift of the target UAV; F min is the minimum lift of the target UAV; z is the angle of attack of the target UAV.

9. The emergency avoidance method based on drone reconnaissance as claimed in claim 8, characterized in that: The flight distance D and the collision probability P c The candidate position corresponding to the minimum product of (x) is used as the target position of the UAV, and the UAV is controlled to fly toward the target position to achieve emergency avoidance.

Images