Low-altitude air route risk map construction method and system based on hexagonal grid cells

Abstract

Disclosed in the present invention are a low-altitude air route risk map construction method and system based on hexagonal grid cells. The method comprises: on the basis of management and control requirements for different types of airspace, forming low-altitude airspace three-dimensional hexagonal hierarchical grids by means of multi-scale partitioning of an airspace horizontal plane and multi-scale partitioning of an airspace vertical plane; constructing an unmanned aerial vehicle flight risk assessment indicator system for the low-altitude airspace three-dimensional hexagonal hierarchical grids, and after optimization, constructing an unmanned aerial vehicle operational risk assessment model based on a dynamic Bayesian network, so as to generate multi-scale grid risk values; and combining the multi-scale grid risk values with geographic location information of a region, designing a hexagonal grid code index, and generating a dynamic multi-scale low-altitude air route three-dimensional hexagonal risk map. The present invention can improve the accuracy and real-time performance of unmanned aerial vehicle flight risk assessment of grids, quickly obtain low-risk grids of an unmanned aerial vehicle in a target region, and improve the safety of air route planning.

Description

Low-altitude air route risk map construction method and system based on cellular grid unit TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle route planning, in particular to a low-altitude air route risk map construction method and system based on cellular grid unit. BACKGROUND

[0002] In the face of the accelerating urbanization process, the intensifying ground traffic congestion and the limited development of underground space, developing low-altitude resources and low-altitude economy will bring unlimited possibilities to the future of urban transportation. In recent years, unmanned aerial vehicles have been widely used in military, civilian and other fields. For example, in the military field, rescue search, surveillance and reconnaissance, and military attack tasks; in the civilian field, city air transport, forest protection, environmental monitoring and other application analysis. With the deep application of unmanned aerial vehicles in various fields, the scale of air traffic has increased dramatically, and the route planning and operation safety control of unmanned aerial vehicles have become a hot topic.

[0003] Most of the existing unmanned aerial vehicle route planning methods are based on latitude and longitude, and the method of solving three-dimensional curve equation with safety margin is used to realize the route planning of unmanned aerial vehicles. The calculation complexity is high, and it has been unable to adapt to the increasing demand for low-altitude traffic control of aircraft. The invention patent "path planning method, device, electronic equipment and route setting method (application number 202310752065.0)" proposes a three-dimensional path planning method for unmanned aerial vehicles based on quadrilateral three-dimensional Beidou grid, which can effectively reduce the calculation complexity. However, this method also has some shortcomings. First of all, it does not comprehensively consider the influence of meteorological conditions, unmanned aerial vehicle conditions, pilot qualifications and flight restrictions in laws and regulations on the safety of unmanned aerial vehicle flight, resulting in insufficient feasibility and scientificity of the planned route. Secondly, the discretization degree of the quadrilateral three-dimensional grid is not high, and the regional coverage efficiency is low, which limits the accuracy and real-time performance of unmanned aerial vehicle flight risk assessment.

[0004] SUMMARY

[0005] Objectives: In view of the shortcomings of the prior art, the present application proposes a low-altitude air route risk map construction method and system based on cellular grid unit, which uses hexagonal cellular grid to grid the geographical surface and constructs a cellular three-dimensional layered grid to quantitatively analyze the unmanned aerial vehicle flight risk of each grid. Finally, a dynamic multi-scale low-altitude air route three-dimensional cellular risk map is generated to improve the accuracy and real-time performance of risk assessment and enhance the safety of route planning.

[0006] Technical solution: In a first aspect, a low-altitude air route risk map construction method based on cellular grid unit includes the following steps:

[0007] Based on different types of airspace control requirements, through the horizontal plane multi-scale division and vertical plane multi-scale division of airspace, a low-altitude airspace three-dimensional cellular hierarchical grid is formed;

[0008] According to the unmanned aerial vehicle flight related data set, an unmanned aerial vehicle flight risk assessment index system for the low-altitude airspace three-dimensional cellular hierarchical grid is constructed;

[0009] Through the feature selection algorithm, the risk factors of the unmanned aerial vehicle flight risk assessment index system are optimized;

[0010] According to the corresponding data set of the optimized risk factors, the unmanned aerial vehicle operation risk assessment model based on dynamic Bayesian network is used to infer the probability of unmanned aerial vehicle operation accident, and the flight risk value of each cellular grid is obtained;

[0011] The sliding average method is used to compensate the error of the flight risk value of each cellular grid, and the weight of each cellular grid is calculated to calculate the regional risk weighted index. The flight risk value of each cellular grid after compensation and the regional risk weighted index form a multi-scale grid risk value;

[0012] The multi-scale grid risk value is combined with the geographic location information, and according to the cellular grid coding index, a dynamic multi-scale low-altitude air route three-dimensional cellular risk map is generated.

[0013] Further, the horizontal plane multi-scale division method is as follows:

[0014] Based on the control of the airspace range, a point at the top left corner of the corresponding area horizontal plane is determined as the earth plane coordinate origin on the map. The coordinates of this point are the origin coordinates o(x0, y0). At the same time, the maximum length of the airspace range in the horizontal and vertical directions is estimated, and a plane cellular grid with n x m x l specifications is established, where n is the number of plane cellular grids on the horizontal plane, m is the number of plane cellular grids on the vertical plane, and l is the minimum circumscribed circle diameter of a single plane cellular grid.

[0015] In the plane cellular grid map coordinate system established with o(x0, y0) as the origin and n x m x l specifications, an index G(v) and g(x, y) is established for each plane cellular grid and the vertex of the grid, respectively, where v is the plane cellular grid number, G(v) represents the vth plane cellular grid, and g(x, y) represents the vertex index value of the plane cellular grid. And n x l is greater than the maximum length of the airspace area in the horizontal direction, (n+1) x l is less than the maximum length of the airspace area in the horizontal direction, m x l is greater than the maximum length of the airspace area in the vertical direction, and (m+1) x l is less than the maximum length of the airspace area in the vertical direction;

[0016] The vertical plane multi-scale division method is as follows:

[0017] After the horizontal plane honeycomb grid is divided, six prismatic shapes are formed by extending the six vertices of each plane honeycomb grid upward by a distance h, and an index g(x, y, z) of the vertices of the regular hexagonal prism is established;

[0018] By estimating the maximum height of the controlled airspace, a three-dimensional honeycomb grid coordinate system with a size of mxnxh is established, where h is the height of a single three-dimensional honeycomb grid, and mxh is greater than the maximum height of the controlled airspace, and (m+1)xh is less than the maximum height of the controlled airspace.

[0019] Further, according to the unmanned aerial vehicle flight related data set, an unmanned aerial vehicle flight risk assessment index system for the three-dimensional honeycomb layered grid of the low-altitude airspace is constructed, including: obtaining unmanned aerial vehicle related data from the meteorological department, the unmanned aerial vehicle flight control system, the unmanned aerial vehicle registration system, the unmanned aerial vehicle comprehensive supervision platform and the unmanned aerial vehicle obstacle detection and obstacle avoidance system, eliminating outliers and filling in missing values to obtain an unmanned aerial vehicle flight risk assessment data set;

[0020] According to the unmanned aerial vehicle flight risk assessment data set, the risk factors affecting the safety of unmanned aerial vehicle flight are determined to form an unmanned aerial vehicle flight risk assessment index system, which includes eight types of indexes of meteorological factors, flight restrictions, air risks, unmanned aerial vehicle conditions, obstacle conditions, ground conditions, pilots and laws and regulations. Each type of index contains one or more risk factors.

[0021] Further, the unmanned aerial vehicle operation risk assessment model based on dynamic Bayesian network is used to infer the probability of unmanned aerial vehicle operation accident, and the flight risk value of each honeycomb grid is obtained, including:

[0022] Pearson correlation analysis is performed on the selected risk factors, and two risk factors with a Pearson correlation coefficient greater than a specified threshold value are identified as having a correlation between them, the causal relationship between them is analyzed, and the causal relationship is connected using a directed arrow to obtain a static Bayesian network;

[0023] The static Bayesian network is extended on the time axis to generate a dynamic Bayesian network containing 2 time slices, the risk factors that change over time are used as dynamic variables, and each dynamic variable is connected to the adjacent time slice nodes using a directed edge. According to the Bayesian formula, the joint prediction probability of the unmanned aerial vehicle flight safety risk across time slices is obtained;

[0024] For each three-dimensional honeycomb grid, the corresponding unmanned aerial vehicle flight safety risk joint prediction probability is obtained according to the actual situation of its risk factors by the above calculation method, and the probability value is the unmanned aerial vehicle flight risk value of the honeycomb grid.

[0025] Further, the moving average method compensates for the error of the flight risk value, including:

[0026] Taking each three-dimensional honeycomb grid as a central hexagonal prism, the following calculations are sequentially performed: taking 6 hexagonal prisms sharing a side face with the central hexagonal prism grid and 2 hexagonal prism grids sharing a bottom face and a top face with the central hexagonal prism grid as neighborhoods, the risk average of the neighborhood grids is calculated and the central hexagonal prism grid is assigned.

[0027] Further, the calculation formula of the regional risk weighted index is as follows:

[0028] In the formula, w n is the risk weight of the nth three-dimensional honeycomb grid, s n is the risk value of the nth three-dimensional honeycomb grid, which is the joint prediction probability of the UAV flight risk obtained by the dynamic Bayesian network, N is the number of three-dimensional honeycomb grids in the region, and V is the weighted risk index of the region.

[0029] Further, the honeycomb grid coding index is composed of 25 bits of coding, in order: 6 bits of administrative region code, 8 bits of latitude and longitude coding, 2 bits of honeycomb grid height, 2 bits of honeycomb grid diameter, 2 bits of multi-scale grid risk value, and 5 bits of serial number.

[0030] In a second aspect, a low-altitude air route risk map construction system based on a honeycomb grid unit includes an airspace honeycomb grid construction module, a grid flight risk assessment module, an air route risk map construction module,

[0031] The airspace honeycomb grid construction module is configured to form a low-altitude airspace three-dimensional honeycomb layered grid through airspace horizontal plane multi-scale division and airspace vertical plane multi-scale division based on different types of airspace control requirements.

[0032] The grid flight risk assessment module includes a risk assessment index system construction unit and a multi-scale grid risk assessment unit. The risk assessment index system construction unit is configured to construct a UAV flight risk assessment index system for the low-altitude airspace three-dimensional honeycomb layered grid according to a UAV flight related data set. The multi-scale grid risk assessment unit is configured to optimize risk factors of the UAV flight risk assessment index system through a feature selection algorithm, infer the UAV operation accident occurrence probability using a UAV operation risk assessment model based on a dynamic Bayesian network according to the corresponding data set of the optimized risk factors, obtain the flight risk value of each honeycomb grid, compensate the flight risk value of each honeycomb grid using a moving average method, assign a weight to each honeycomb grid to calculate a regional risk weighted index, and form a multi-scale grid risk value from the compensated flight risk value of each honeycomb grid and the regional risk weighted index.

[0033] The air route risk map construction module is configured to combine the multi-scale grid risk values with geographical position information, and generate a dynamic multi-scale low-altitude air route three-dimensional cellular risk map according to a cellular grid coding index.

[0034] In a third aspect, a computer device includes a memory storing one or more programs, and a processor communicatively coupled to the memory and configured to execute the programs, which when executed by the processor implement the steps of the cellular grid cell based low-altitude air route risk map construction method according to the first aspect of the application.

[0035] In a fourth aspect, a non-transitory processor-readable storage medium has stored thereon processor-executable instructions, which when executed by a processor of a mobile terminal implement the steps of the cellular grid cell based low-altitude air route risk map construction method according to the first aspect of the application.

[0036] Advantages: Compared with the prior art, the application has the following advantages:

[0037] 1. A hexagonal grid cell construction method is designed. The hexagonal grid is a polygon that can seamlessly fill a plane and has the maximum number of edges connected at each vertex. It has the advantages of consistent neighborhood, isotropy, compactness, high sampling rate, and more flexible adaptation to terrain changes. Compared with the existing quadrate grid based partitioning algorithm in the unmanned aerial vehicle operation risk assessment algorithm, the accuracy and real-time performance of risk assessment are higher.

[0038] 2. According to different flight safety control related standards and requirements of different types of airspace in the country, a multi-scale layered space grid setting method is designed to meet the needs of overall risk situation awareness and micro risk feature insight of the region.

[0039] 3. Based on the national unmanned aerial vehicle operation management standards and regulations system, combined with the actual operation of unmanned aerial vehicles, a multi-level unmanned aerial vehicle flight risk assessment system including meteorological conditions, unmanned aerial vehicle conditions, obstacles and pilot qualifications is established to provide support for more scientific and feasible assessment of unmanned aerial vehicle operation risks.

[0040] 4. The outstanding role of dynamic Bayesian networks in causal relationship uncertainty and correlation analysis is utilized to construct an unmanned aerial vehicle operation risk assessment model. The analysis results provide a strong basis for risk prevention and control management and air route planning of unmanned aerial vehicles.

[0041] 5. According to the actual air traffic control application needs of different coarse and fine resolutions, a regional weighted risk evaluation index calculation method is designed to highlight the unmanned aerial vehicle operation risk characteristics of different regions. The sliding average model is used to compensate for the error of grid risk outliers, and the safety of the planned air route is improved. BRIEF DESCRIPTION OF DRAWINGS

[0042] Fig. 1 is a schematic diagram of a low-altitude air route risk map construction method based on a cellular grid unit;

[0043] Fig. 2 is an example of a low-altitude airspace three-dimensional cellular grid;

[0044] Fig. 3 is a schematic diagram of horizontal plane hexagonal grid division coordinates;

[0045] Fig. 4 is a schematic diagram of vertical plane height layered hexagonal prism indexing;

[0046] Fig. 5 is the contribution and cumulative contribution of each risk assessment factor calculated by random forest;

[0047] Fig. 6 is a correlation heat map of the influence factors of the UAV flight safety event;

[0048] Fig. 7 is a static Bayesian network structure diagram for UAV flight risk assessment;

[0049] Fig. 8 is a dynamic Bayesian network structure diagram for UAV flight risk assessment;

[0050] Fig. 9 is a schematic diagram of grid risk assessment coding rules;

[0051] Fig. 10 is an example of a final generated three-dimensional cellular risk map. DETAILED DESCRIPTION

[0052] The technical solutions of the present application will be further described below in combination with the drawings.

[0053] The present application proposes a low-altitude air route risk map construction method based on a cellular grid unit, referring to Fig. 1, the method comprises the following steps:

[0054] Step 1: Construct a low-altitude airspace three-dimensional cellular grid.

[0055] In step 1 of the present application, based on different types of airspace control requirements, through horizontal plane multi-scale division and vertical plane multi-scale division of airspace, a low-altitude airspace three-dimensional cellular layered grid is formed.

[0056] According to the embodiment of the present application, through various remote sensing and surveying means such as space-based, air-based and ground-based, a multi-layer geographic information database of the target area is formed. At the same time of generating each layer of geographic information database, the data information is divided into appropriate scale plane cellular grid according to different regional restriction range in the "National Airspace Basic Classification Method", and the plane cellular grid is divided into different height levels according to the airspace restriction height in the "National Airspace Basic Classification Method", forming a low-altitude airspace three-dimensional cellular layered grid.

[0057] As an example, data of roads, terrain, elevation, slope, vegetation, rivers, grassland, farmland, digital elevation model (DEM), three-dimensional digital terrain model (DTM), and meteorological hydrology, etc. in the target area are acquired to form a multi-layer geographic information database of the controlled low-altitude airspace area. Based on the multi-layer geographic information database, the horizontal plane and vertical plane of the controlled low-altitude airspace are divided into multiple scales according to the control requirements of different types of airspace in the "National Airspace Basic Classification Method". For airspace with high operation demand and safety requirements, a finer grid resolution is adopted to provide more accurate flight safety control; otherwise, a coarser grid resolution is adopted. The correspondence between airspace types and airspace setting accuracy is shown in Table 1. The grid horizontal diameter refers to the diameter of the circumscribed circle of the regular hexagon.

[0058] Table 1: Airspace type and airspace setting accuracy

[0059] For example, a certain controlled low-altitude airspace contains three types of C, D, and G. According to Table 1, the formed three-dimensional honeycomb grid setting diagram is shown in FIG. 2. The grid in Table 1 refers to a spatial grid, also known as a three-dimensional honeycomb grid, which is a hexagonal prism shape, i.e., the concept of a honeycomb grid unit in the present application. Unless otherwise explicitly stated, the grid in the present application refers to a three-dimensional honeycomb grid. The three-dimensional honeycomb grid is formed by horizontal plane hexagonal division and vertical plane layering. For clarity of description, the horizontal plane hexagon is also referred to as a planar honeycomb grid or planar grid to distinguish.

[0060] Taking the division of C-class airspace (according to the "National Airspace Basic Classification Method", C-class airspace is set above a general aviation airport with a tower, usually a single ring structure with a radius of 5 kilometers and a runway pavement- airport elevation of 600 meters (including)) as an example, the division method is as follows:

[0061] Horizontal plane hexagonal grid (planar honeycomb grid) division method:

[0062] (1) After determining the controlled airspace range, the earth plane coordinates of a point at the upper left corner of the horizontal plane in the region are determined on the map, the point coordinates are taken as the origin, and the origin coordinates o(x0, y0) are established. At the same time, the maximum length of the airspace range in the horizontal and vertical directions is estimated, and a plane honeycomb grid of n x m x l specifications is established, that is, a regular hexagonal plane grid, wherein: n is the number of plane honeycomb grids on the horizontal plane horizontal axis, m is the number of plane honeycomb grids on the horizontal plane vertical axis, and l is the minimum circumscribed circle diameter of a single plane honeycomb grid, in meters, n x l needs to be greater than the maximum length of the airspace region in the horizontal direction, (n+1) x l needs to be less than the maximum length of the airspace region in the horizontal direction, m x l needs to be greater than the maximum length of the airspace region in the vertical direction, and (m+1) x l needs to be less than the maximum length of the airspace region in the vertical direction.

[0063] (2) In the plane grid map coordinate system established with o(x0, y0) as the origin and n x m x l specifications, an index G(v) and g(x, y) are respectively established for each plane honeycomb grid and the vertex of the grid, wherein: v is the plane honeycomb grid number, G(v) represents the vth plane honeycomb grid, and g(x, y) represents the vertex index value of the plane honeycomb grid. The plane honeycomb grid drawing schematic is shown in Figure 3. The vertex index value of each plane honeycomb grid can be obtained according to basic mathematical operations, which will not be described here.

[0064] Vertical plane height layering method:

[0065] After dividing the horizontal plane honeycomb grid, the six vertices of each plane honeycomb grid are extended upward by a distance h to form a regular hexagonal prism (referred to as hexagonal prism) shape, and an index g(x, y, z) of the vertex of the regular hexagonal prism is established. Taking the first regular hexagonal prism as an example, the index schematic is shown in Figure 4. The vertex index value of each hexagonal prism grid can be obtained according to simple mathematical operations, which will not be described here.

[0066] The maximum height of the controlled airspace is estimated, and a three-dimensional honeycomb grid coordinate system of m x n x h specifications is established, wherein: h is the height of a single three-dimensional honeycomb grid, in meters, wherein m x h needs to be greater than the maximum height of the controlled airspace, and (m+1) x h needs to be less than the maximum height of the controlled airspace.

[0067] The present application designs a honeycomb grid unit construction method based on a regular hexagon. The regular hexagonal grid is a polygon that can seamlessly fill a plane and has the most edges connected at each vertex. It has the advantages of consistent neighborhood, isotropy, compactness, high sampling rate, and more flexible adaptation to terrain changes. Compared with the existing unmanned aerial vehicle operation risk assessment algorithm based on a regular quadrilateral grid division algorithm, the accuracy and real-time performance of risk assessment are higher. According to different flight safety control standards of different types of airspace in the country, a multi-scale layered space grid setting method is designed to meet the needs of managers for overall risk situation awareness and micro risk feature insight in the region.

[0068] Step 2: Establishing a three-dimensional cellular grid UAV flight risk assessment index system.

[0069] According to the embodiments of the present application, a multi-level risk assessment system is established in the three-dimensional cellular grid, including indicators such as geographical restriction range, airspace restriction height, obstacles, weather conditions, aircraft parameters, and UAV pilot qualifications. Specifically, the following steps are included:

[0070] Step 2.1: Data collection and preprocessing. Collect data on weather, flight performance, flight control requirements, UAV pilot conditions, and obstacles from meteorological departments, UAV flight control systems, UAV registration systems, UAV comprehensive supervision platforms, and UAV obstacle detection and obstacle avoidance systems, etc. The weather data includes temperature, humidity, rainfall, wind speed, etc. The flight performance data includes maximum turning angle, flight speed, UAV size, UAV weight, etc. The flight control requirements include airspace type, height limit, time limit, speed limit, etc. The UAV pilot conditions include UAV pilot qualification level, flight duration, etc. The obstacle data includes obstacle proximity rate, obstacle size, obstacle speed, etc. Use box plots to detect outliers in the data set and fill in missing values to obtain the UAV flight risk assessment data set.

[0071] Step 2.2: Determine the UAV flight risk assessment index system, including weather factors, flight restrictions, air risks, UAV conditions, obstacle conditions, ground conditions, pilot conditions, and legal regulations.

[0072] Weather factors: Weather conditions are an important factor affecting the safety of UAV flight. For example, strong winds, heavy rain, and other adverse weather conditions can cause the UAV to lose control or affect the flight effect.

[0073] Flight restrictions: Different airspaces have restrictions on UAV flight height, flight time period, and flight speed. If these restrictions are not followed, it may result in risks such as exceeding the wireless communication range, crossing with other vehicles, and aircraft crashes.

[0074] Air risks: During flight, UAVs are prone to collisions with obstacles, loss of control, and other flight accidents due to their high speed and maneuverability. In addition, electromagnetic interference can cause the UAV to lose communication and navigation capabilities.

[0075] UAV conditions: In actual flight, the maximum turning angle, flight speed, and UAV weight of the UAV, as well as battery failure and damaged propellers, can also affect the risk of UAV collisions.

[0076] Obstacles: During drone flight, obstacles such as birds or other vehicles may suddenly appear in localized areas. These unexpected events are difficult to predict and can significantly impact drone flight.

[0077] Ground conditions: If the drone goes out of control and the population density on the ground is high, it may cause serious injury.

[0078] Pilot qualifications: Pilots with different qualifications have different abilities to operate drones and respond to emergencies, which in turn affects the operational risks of drones.

[0079] Laws and regulations: Different regions have different laws and regulations restricting the flight paths and areas of drones. Risk assessments need to take these restrictions into account to evaluate the compliance of drone flights.

[0080] The UAV flight risk assessment index system established in this invention is shown in Table 2, which consists of eight categories and 22 risk factors.

[0081] Table 2. Unmanned Aerial Vehicle (UAV) Flight Risk Assessment Index System

[0082] Based on national standards and regulations for unmanned aerial vehicle (UAV) operation management, and combined with the actual operation of UAVs, this invention establishes a multi-level UAV flight risk assessment system that includes indicators such as meteorological conditions, UAV status, obstacles, and pilot qualifications, providing support for a more scientific and feasible assessment of UAV operation risks.

[0083] Step 3: Construct a UAV operation risk assessment model based on a dynamic Bayesian network to evaluate the risk value of a multi-scale three-dimensional cellular grid. This includes the following steps:

[0084] Step 3.1: Risk factor optimization.

[0085] The original high-dimensional drone flight risk sample dataset contains numerous factors, and excessive redundant factors may introduce unnecessary noise that could negatively impact model performance. Therefore, a random forest feature selection algorithm is employed to select the optimal set of factors from the original high-dimensional drone accident risk factor dataset, thereby improving the model's computational efficiency and fitting accuracy.

[0086] Random forest is an ensemble supervised learning model that uses decision trees as base learners. By calculating the contribution of each factor to each tree, the importance of factors to drone flight accidents can be assessed. The corresponding algorithm steps are as follows:

[0087] (1) Randomly draw K training datasets {D1, D2, ..., D...} with replacement from the collected original sample set. K} and K corresponding out-of-bag test datasets {B1, B2, ..., B K};

[0088] (2) Set k = 1, build the decision tree on the training set D k , use the decision tree to predict B k , and calculate the out-of-bag data unmanned aerial vehicle flight accident prediction accuracy, denoted as

[0089] (3) Apply random noise disturbance to the factors X k in B i , i = 1, 2, …, n, and calculate the out-of-bag data prediction accuracy

[0090] (4) Repeat steps (2) and (3) for k = 1, 2, …, K;

[0091] (5) Measure the importance of the factor X i using the following formula

[0092] (6) Sort the factor importance in descending order, and delete redundant factors according to the cumulative contribution rate to generate an optimal factor set.

[0093] According to the method described in step 3.1, 22 unmanned aerial vehicle flight safety risk factors are screened, and the factor contribution degree ranking and cumulative contribution degree are shown in FIG. 5. The cumulative contribution degree of the first 14 unmanned aerial vehicle flight risk factors is as high as 95.07%. The contribution degree of the last 8 factors is small, which can be regarded as redundant factors, so the first 14 factors: temperature, rainfall, wind speed, electromagnetic interference, obstacle approach rate, unmanned aerial vehicle failure, air collision, unmanned aerial vehicle speed, unmanned aerial vehicle size, pilot qualification, compliance, airspace type, population density, and ground loss are selected as the input data of the Bayesian model.

[0094] Step 3.2: Dynamic Bayesian network-based operational risk distribution assessment.

[0095] A Bayesian network (BN) is a graphical mathematical model for probabilistic reasoning. It describes the dependency relationship between variables using a directed acyclic graph. Given observed sample data or expert prior knowledge, BN (static Bayesian network) can use Bayesian probability theory to perform probabilistic reasoning on target variables and is widely used in problems involving multiple variables and uncertainty. Dynamic Bayesian network (DBN) introduces the concept of time sequence based on BN, and performs reasoning on dynamic time-varying random processes by describing the time sequence transition and causal relationship of random variables. In this application, the screened risk factor dataset is discretized as the input of the dynamic Bayesian network model, and the maximum likelihood estimation method is used to learn the conditional probability of the initial network and the transition network.

[0096] Specifically, the following steps are included:

[0097] Step 3.2.1: Construct a static Bayesian network for drone risk assessment.

[0098] This invention constructs a static risk factor network (BN) based on 14 risk factors with the highest contribution rates obtained through screening. First, Pearson correlation analysis is performed on each factor, and a correlation heatmap is plotted, as shown in Figure 6. Two factors with an absolute correlation coefficient greater than 0.1 exhibit a certain correlation. For example, the correlation coefficient between population density and ground loss is 0.33, indicating a positive correlation. This is because higher population density leads to greater ground loss when drones crash or cause other accidents. During drone flight, pilot qualifications, drone flight speed, and the type of airspace flown must comply with relevant laws and regulations. Electromagnetic interference, excessively high temperatures, wind speed, and rainfall may cause drone malfunctions such as communication failures, navigation system failures, and structural damage. Drone malfunctions, approaching obstacles, strong winds, and heavy rain may lead to mid-air collisions / crashes. Furthermore, the likelihood of a drone accident is also related to pilot qualifications; generally, highly qualified pilots have a stronger ability to operate drones in adverse conditions and are less prone to accidents. The degree of ground damage is related to the population density below the flight path, the likelihood of mid-air collisions / crashes, and the size of the drone. The combined risks of drone flight compliance, mid-air collisions / crashes, and ground damage constitute the operational risks of drones. Connecting these causal relationships using directed arrows yields a static BN structure as shown in Figure 7.

[0099] A Bayesian network is intuitively a directed acyclic graph, consisting of three elements: nodes and directed arcs, prior probabilities, and conditional probabilities. When applied to drone safety incident risk assessment, the nodes in the Bayesian network are represented by random variables (X1,…,X…). n This refers to the various risk factors (risk factors) that affect the risk of drone flight accidents. Assume P... a (X i Risk factor X i The probability of the parent node occurring in the model, and the risk factor X. i The conditional probability can be expressed as P(X) i |P a (X i If the joint probability distribution of the overall risk of drone flight is P(X1,…,X), then the probability distribution of the overall risk of drone flight is P(X1,…,X). n As shown in the following formula, this formula can be used to predict the probability of a drone flight safety risk event X.

[0100] Step 3.2.2: Construct a dynamic Bayesian network (DBN) for drone risk assessment.

[0101] DBN (Database-Based Network) is a probabilistic model that integrates the original network structure with temporal information, building upon a static Bayesian network to process time-series data. The DBN model uses directed edges to connect dynamic variables along the time axis, and mines temporal correlations through transition probabilities, thereby incorporating time factors to infer the probability of drone operational risks.

[0102] In the short term, risk factors such as temperature, wind speed, rainfall, drone speed, obstacle proximity rate, drone malfunction, mid-air collision, flight operation risk, flight restriction time, and ground loss are strongly correlated over time and vary widely. Their states at the previous and current moments continuously affect drone flight risk, exhibiting a cumulative effect over time. Therefore, the time factor is introduced into the BN model to analyze the combined states of risk factors such as temperature, wind speed, rainfall, drone speed, obstacle proximity rate, drone malfunction, mid-air collision, flight operation risk, compliance, and ground loss at different times.

[0103] This invention constructs a 2-time-slice DBN model, which infers the probability of UAV operation risk based on the state of each factor in the previous and current time, so as to reflect the probability of an accident occurring during the flight of the aircraft at the current time.

[0104] The static Bayesian Network (BN) is extended along the time axis to generate a DBN with two time slices, as shown in Figure 8. Parameters such as temperature, drone speed, obstacle approach rate, and wind speed in the DBN are dynamic variables; therefore, directed edges are used to connect adjacent time slice nodes to reflect the temporal changes of these variables. Other static variables do not require connections between time slices. To ensure the stability and realism of the model results, a 5-fold cross-validation method is used to train and test the dynamic Bayesian Network. Specifically, the dataset is divided into five mutually exclusive subsets of equal size using stratified sampling. Four subsets are used as the training set each time, and the remaining subset is used as the test set. Maximum likelihood estimation is used to learn the node parameters of the dynamic Bayesian Network. Then, the probability of drone accidents is predicted on the test set samples, and the drone operation risk is classified into different levels based on the probability of drone accidents.

[0105] Specifically, the DBN of drone flight safety risk is defined as (B1, B... → The initial Bayesian network is denoted by B1, B → The graph represents a combination of Bayesian networks across different time slices. Based on the initial distribution and the conditional distribution between adjacent time slices, expanding the dynamic Bayesian network to the T-th time slice yields a joint prediction probability of UAV flight safety risks spanning multiple time slices.

[0106] Since different three-dimensional honeycomb grids correspond to different geographical locations, the variable values of the risk factors corresponding to each grid are different, therefore, for each three-dimensional honeycomb grid, a risk probability value can be obtained according to the above calculation method, and this probability value is a risk prediction for the unmanned aerial vehicle flying through the grid.

[0107] The greater the predicted probability value of the safety event that may occur during the flight of the unmanned aerial vehicle, the higher the corresponding risk level, and the corresponding relationship between the predicted probability and the risk level in the present application is shown in Table 3.

[0108] Table 3 Corresponding relationship between predicted probability and unmanned aerial vehicle operation risk level

[0109] The present application utilizes the outstanding role of dynamic Bayesian networks in the analysis of uncertainty and correlation of causal relationships, and constructs an unmanned aerial vehicle operation risk assessment model, determines the unmanned aerial vehicle operation risk level according to the predicted probability, and the analysis results provide a strong basis for the risk prevention and control management and flight path planning of the unmanned aerial vehicle.

[0110] Step 3.3: Neighborhood grid risk value error compensation based on moving average method

[0111] The grid risk assessment value is affected by the unmanned aerial vehicle obstacle detection system, for example, the sensor device in the obstacle detection system may introduce errors in the process of mutual inductive device synthesis, voltage mutual inductive device secondary circuit voltage drop, etc., resulting in a certain deviation between the obstacle detection result and the actual situation, and further affecting the accurate assessment of the risk. These errors are difficult to detect and eliminate due to the limitations of the internal structure and measurement principle of the measuring device.

[0112] The present application adopts a moving average method to reduce the abnormal fluctuation level of the risk value in the grid, which takes the 6 hexagonal prisms sharing the side surface of the center hexagonal prism grid and the two hexagonal prism grids sharing the bottom surface and the top surface of the center hexagonal prism grid as the neighborhood, calculates the risk average value of the neighborhood hexagonal prism grid, and assigns the center hexagonal prism grid.

[0113] Each hexagonal prism is sequentially taken as a center hexagonal prism grid for average processing. Assuming that the 8 neighborhood hexagonal prism grids V1, V2, V3, …, V8 of the center hexagonal prism grid V o , then the risk value of the hexagonal prism grid V o after moving average processing is:

[0114] In the formula, R o is the risk value of the hexagonal prism grid V oR1, R2, R3, …, R8 are the risk values corresponding to the hexagonal prism grid V1, V2, V3, …, V8.

[0115] Step 3.4: regional weighted risk assessment

[0116] In actual air traffic control, there is a need to grasp the overall risk pattern and trend of the region, and there is also a need for more detailed regional micro risk feature insight. For example, for different control regions, such as above a certain residential area, above a certain hospital, above a certain important infrastructure, above a certain nature reserve, etc., the demand for grasping risk pattern and risk characteristics is different. According to the actual air traffic control application needs of different coarse and fine resolution, the present application designs a regional weighted risk index calculation method.

[0117] In the formula, w n is the risk weight of the nth three-dimensional honeycomb grid, s n is the risk value of the nth three-dimensional honeycomb grid, N is the number of three-dimensional honeycomb grids in the region, and V is the weighted risk index of the region.

[0118] In the present application, according to the unmanned aerial vehicle operation risk level, the three-dimensional honeycomb grid risk weight evaluation value is determined as shown in Table 4.

[0119] Table 4: Grid weighted risk assessment weight

[0120] The present application uses a moving average model to compensate for grid risk outliers, improving the safety of the planned flight path. According to the actual air traffic control application needs of different coarse and fine resolution, a regional weighted risk evaluation index calculation method is designed to highlight the unmanned aerial vehicle operation risk characteristics of different regions. In the present application, the flight risk value of each compensated honeycomb grid and the regional risk weighted index form a multi-scale grid risk value, so that the unmanned aerial vehicle operation risk can be comprehensively grasped from the single grid and regional level. For example, a large-scale grid can be a large area composed of many grids; a small-scale grid can be a small area composed of fewer grids, or a single grid.

[0121] Step 4: combined with multi-scale grid risk value and geographic space position, according to the grid coding mechanism to generate dynamic multi-scale low-altitude flight path three-dimensional honeycomb risk map.

[0122] The present application establishes a grid risk assessment coding mechanism to quantify the safety risk of each spatial grid for unmanned equipment flight, and uses the efficient calculation characteristics of the grid code to quickly obtain the low-risk grid that the unmanned equipment can pass through in the target area, and then plans a reasonable route.

[0123] Each grid risk assessment is uniquely identified by an identification code consisting of 25 bits of code, in order: administrative region code (6 bits), latitude and longitude code (8 bits), cell height (2 bits), cell diameter (2 bits), risk value (2 bits), serial number (5 bits). The coding rules are shown in FIG. 9, and the field descriptions are shown in Table 5.

[0124] Table 5: Field descriptions of the cell grid coding

[0125] FIG. 10 shows an example of a final generated three-dimensional cell risk map, which is an example of a risk map containing low-risk, lower-risk, and medium-risk areas.

[0126] Based on the same technical concept as the method embodiment, the application also provides a low-altitude air route risk map construction system based on a cell grid unit, comprising an airspace cell grid construction module, a grid flight risk assessment module, and an air route risk map construction module. The connection relationship of each module is shown in FIG. 1.

[0127] The airspace cell grid construction module is used to form a low-altitude airspace three-dimensional cell layered grid based on different types of airspace control requirements through multi-scale division of the airspace horizontal plane and multi-scale division of the airspace vertical plane. The multi-scale division of the airspace horizontal plane and the vertical plane height layering method are described in the foregoing method embodiment.

[0128] The grid flight risk assessment module is used to construct a dynamic Bayesian network-based unmanned aerial vehicle operation risk assessment model based on a cell three-dimensional grid unmanned aerial vehicle flight risk assessment index system, and to assess the multi-scale grid risk value.

[0129] The index system construction unit includes a multi-scale grid risk assessment unit.

[0130] The risk assessment index system construction unit determines the unmanned aerial vehicle flight risk assessment index system, which includes meteorological factors, flight restrictions, air risks, unmanned aerial vehicle conditions, obstacle conditions, ground conditions, pilot conditions, and legal compliance indicators, and each type of indicator contains one or more evaluation factors.

[0131] The multi-scale grid risk assessment unit: the risk factors of the unmanned aerial vehicle flight risk assessment index system are optimized through a feature screening algorithm, the occurrence probability of the unmanned aerial vehicle operation accident is inferred by using an unmanned aerial vehicle operation risk assessment model based on a dynamic Bayesian network according to the corresponding data set of the optimized risk factors, and the flight risk value of each cellular grid is obtained; the flight risk value of each cellular grid is compensated for error by using a moving average method, and each cellular grid is weighted to calculate a regional risk weighted index, and the flight risk value of each compensated cellular grid and the regional risk weighted index form a multi-scale grid risk value. The specific calculation method can be referred to the description of the method embodiment.

[0132] The air route risk map construction module is used for combining the multi-scale grid risk value with the geographical position information of the region, and generating a dynamic multi-scale low-altitude air route three-dimensional cellular risk map. The cellular grid coding dynamic index is designed, the identification code content includes administrative region code, current time, risk value and risk level and the like, and the efficient calculation characteristics of the grid unit are used to obtain the low-risk grid of the unmanned aerial vehicle in the target region, and then a reasonable route is quickly planned.

[0133] The application further provides a computer device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the program is executed by the processor to realize the steps of the low-altitude air route risk map construction method based on the cellular grid unit.

[0134] The application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by the processor to realize the steps of the low-altitude air route risk map construction method based on the cellular grid unit.

[0135] Those skilled in the art should understand that the embodiments of the application can be provided as a method, a system, or a computer program product. Therefore, the application can adopt a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0136] This invention is described with reference to flowchart illustrations of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each step in the flowchart, and combinations of steps in the flowchart, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more steps of the flowchart.

[0137] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more processes of a flowchart.

[0138] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more processes in the flowchart.

Claims

1. A method for constructing a low-altitude air route risk map based on a cellular grid cell, characterized by, The method comprises the following steps: Based on different types of airspace control requirements, through multi-scale division of the horizontal plane of the airspace and multi-scale division of the vertical plane of the airspace, a low-altitude airspace three-dimensional cellular layered grid is formed; According to a UAV flight related data set, a UAV flight risk assessment index system for the low-altitude airspace three-dimensional cellular layered grid is constructed, including: obtaining UAV related data from a meteorological department, a UAV flight control system, a UAV registration system, a UAV comprehensive supervision platform and a UAV obstacle detection and obstacle avoidance system, removing outliers and filling in missing values to obtain a UAV flight risk assessment data set; determining risk factors affecting UAV flight safety according to the UAV flight risk assessment data set, and forming a UAV flight risk assessment index system, wherein the UAV flight risk assessment index system comprises eight types of indexes, i.e., meteorological factors, flight restrictions, air risks, UAV conditions, obstacle conditions, ground conditions, pilot conditions and laws and regulations, and each type of index comprises one or more risk factors; The risk factors of the UAV flight risk assessment index system are optimized through a feature screening algorithm; According to the corresponding data set of the optimized risk factors, a UAV operation risk assessment model based on a dynamic Bayesian network is used to infer the probability of a UAV operation accident, and a flight risk value of each cellular grid is obtained; A sliding average method is used to compensate for errors of the flight risk value of each cellular grid, and a regional risk weighted index is calculated by weighting each cellular grid, and the compensated flight risk value of each cellular grid and the regional risk weighted index form a multi-scale grid risk value; The multi-scale grid risk value is combined with geographic location information, and a dynamic multi-scale low-altitude air route three-dimensional cellular risk map is generated according to a cellular grid coding index.

2. The method of claim 1, wherein, The multi-scale division method of the horizontal plane of the airspace is as follows: Based on the controlled airspace range, a point at the upper left corner of the corresponding area horizontal plane on the map is determined as the earth plane coordinate origin, and the coordinates of the point are the origin coordinate o(x0, y0). At the same time, the maximum length of the airspace range in the horizontal direction is estimated, and a plane cellular grid with a specification of n x m x l is established, wherein n is the number of plane cellular grids on the horizontal plane axis, m is the number of plane cellular grids on the vertical plane axis, and l is the minimum circumscribed circle diameter of a single plane cellular grid. On the plane cellular grid map coordinate system established with o(x0, y0) as the origin and a specification of n x m x l, an index G(v) and g(x, y) are respectively established for each plane cellular grid and the vertex of the grid, wherein v is the plane cellular grid number, G(v) represents the vth plane cellular grid, and g(x, y) represents the vertex index value of the plane cellular grid. Moreover, n x l is greater than the maximum length of the airspace range in the horizontal direction, (n+1) x l is less than the maximum length of the airspace range in the horizontal direction, m x l is greater than the maximum length of the airspace range in the vertical direction, and (m+1) x l is less than the maximum length of the airspace range in the vertical direction. The multi-scale division method of the vertical plane of the airspace is as follows: After the horizontal plane cellular grid is divided, six prisms are formed by extending h distance upward based on the six vertices of each plane cellular grid, and an index g(x, y, z) of the vertex of the regular hexagonal prism is established. By estimating the maximum height of the controlled airspace, a three-dimensional honeycomb grid coordinate system of m×n×h is established, wherein h is the height of a single three-dimensional honeycomb grid, m×h is greater than the maximum height of the controlled airspace, and (m+1)×h is less than the maximum height of the controlled airspace.

3. The method of claim 1, wherein, The UAV operation risk assessment model based on the dynamic Bayesian network is used to infer the UAV operation accident occurrence probability, and flight risk values of each honeycomb grid are obtained, including: Pearson correlation analysis is performed on the optimized risk factors, two risk factors with a Pearson correlation coefficient absolute value greater than a specified threshold are identified as having a correlation therebetween, a causal relationship therebetween is analyzed, the causal relationship is connected by a directional arrow, and a static Bayesian network is obtained; The static Bayesian network is extended on a time axis to generate a dynamic Bayesian network containing two time slices, a risk factor changing over time is used as a dynamic variable, adjacent time slice nodes of each dynamic variable are connected by a directional edge, and a UAV flight safety risk joint prediction probability across time slices is obtained according to a Bayesian formula; For each three-dimensional honeycomb grid, a corresponding UAV flight safety risk joint prediction probability is obtained by the above calculation method according to the actual situation of the risk factors thereof, and the probability value is a UAV flight risk value of the honeycomb grid.

4. The method of claim 2, wherein, The moving average method is used to compensate errors of the flight risk values, including: Each three-dimensional honeycomb grid is taken as a central hexagonal prism, and the following calculations are sequentially performed: six hexagonal prisms sharing a side surface with the central hexagonal prism grid and two hexagonal prism grids sharing a bottom surface and a top surface with the central hexagonal prism grid are taken as neighborhoods, an average risk value of the neighborhood grids is calculated, and the central hexagonal prism grid is assigned the average risk value.

5. The method of claim 1, wherein, The formula for calculating the regional risk-weighted index is as follows: In the formula, w n is the nth stereoscopic cellular grid risk weight, s n is the nth stereoscopic cellular grid risk value, which is a joint prediction probability of the UAV flight risk obtained through a dynamic Bayesian network, N is the number of stereoscopic cellular grids in the region, and V is the weighted risk index of the region.

6. The method of claim 1, wherein, The honeycomb grid coding index is composed of 25-bit coding, and the order is as follows: 6-bit administrative region code, 8-bit longitude and latitude coding, 2-bit honeycomb grid height, 2-bit honeycomb grid diameter, 2-bit multi-scale grid risk value, and 5-bit serial number.

7. A cellular grid cell based low altitude air route risk map construction system, characterized by, The method comprises the following steps: The airspace honeycomb grid construction module is configured to form a low-altitude airspace three-dimensional honeycomb hierarchical grid through horizontal plane multi-scale division and vertical plane multi-scale division based on different types of airspace control requirements; The grid flight risk assessment module comprises a risk assessment index system construction unit and a multi-scale grid risk assessment unit; the risk assessment index system construction unit is configured to construct a UAV flight risk assessment index system for the low-altitude airspace three-dimensional honeycomb hierarchical grid according to a UAV flight related data set; The multi-scale grid risk assessment unit is configured to optimize risk factors of the UAV flight risk assessment index system by a feature screening algorithm, infer a UAV operation accident occurrence probability by using a UAV operation risk assessment model based on a dynamic Bayesian network according to a corresponding data set of the optimized risk factors, and obtain flight risk values of each honeycomb grid; The moving average method is used to compensate errors of the flight risk values of each honeycomb grid, and a regional risk weighted index is calculated by assigning a weight to each honeycomb grid, and the flight risk values of each honeycomb grid after compensation and the regional risk weighted index form a multi-scale grid risk value. The route risk map construction module is configured to combine the multi-scale grid risk values with geographical position information, and generate a dynamic multi-scale low-altitude route three-dimensional cellular risk map according to a cellular grid coding index. The risk assessment index system construction unit constructs a risk assessment index system for the low-altitude airspace three-dimensional cellular hierarchical grid according to a UAV flight related data set, including: obtaining UAV related data from a meteorological department, a UAV flight control system, a UAV registration system, a UAV comprehensive supervision platform, and a UAV obstacle detection and obstacle avoidance system, eliminating outliers and filling in missing values to obtain a UAV flight risk assessment data set; determining risk factors affecting UAV flight safety according to the UAV flight risk assessment data set to form a UAV flight risk assessment index system, the UAV flight risk assessment index system including eight types of indexes of meteorological factors, flight restrictions, air risks, UAV conditions, obstacle conditions, ground conditions, pilot conditions, and laws and regulations, each type of index including one or more risk factors.

8. A computer device, comprising: The method comprises the following steps: a memory storing one or more programs; and a processor communicatively coupled to the memory and configured to execute the programs, when the programs are executed by the processor, the steps of the method for constructing a low-altitude route risk map based on a cellular grid unit according to any one of claims 1-6 are implemented.

9. A non-transitory processor-readable storage medium having stored thereon processor-executable instructions, the instructions being executable by a processor to cause the processor to perform operations comprising: The processor executable instructions are executed by the processor of the mobile terminal to implement the steps of the method for constructing a low-altitude route risk map based on a cellular grid unit according to any one of claims 1-6.

Images